Information

Does a software exist to automatically design antibodies that target a given protein or antigen?

Does a software exist to automatically design antibodies that target a given protein or antigen?


We are searching data for your request:

Forums and discussions:
Manuals and reference books:
Data from registers:
Wait the end of the search in all databases.
Upon completion, a link will appear to access the found materials.

I'm looking for software to automatically design antibodies that target a given protein or antigen.


Much easier said than done!

If I had sufficient funds and time to pursue your goal what I would do is comb the Protein Data Bank for all the PDB files of interest. Surely you have some protein that interest you. Might consider doing the following:

  1. Figure out what family your protein belongs in, considering it's number and presence of motifs: coils, helices, beta sheets, etc…

  2. Do a literature review of any protein that has already been studied belonging to this family. So when a protein in this family catalyzes a reaction what are its flexible and restrained regions?

  3. Use a program like Autodock to dock your protein to a protein that its protein family might interact with.

  4. Calculate the interaction energies using Gromacs, NAMD, or Gaussian. You can do a zero time calculation of initial states or pursue a molecular dynamics approach as the proteins move randomly to their most stable configurations, or even a molecular mechanics approach. Which systems are most favorable?

  5. You got this far. Now what? But then none of these steps creates your candidates. It just gives you an idea of the physics involved in already structurally identified compounds. Yes, you can make point mutations in VMD and Pymol. But, if you want to add a new motif to an existing protein structure it will not work because more than 20 amino acids in a peptide do not fold properly.

To do some of the biological statistics, the research collaboratory of structural bioinformatics is right here: http://www.rcsb.org/pdb/home/home.do

You can parse through their folders with wget or curl in bash scripting at http://files.rcsb.org/

But then assuming you do figure out a candidate protein that doesn't exist yet. How are you going to make it in the lab? It appears that it would be better to do microbial studies. Maybe you can screen for a match with micro-array plates.


Structure and development of single domain antibodies as modules for therapeutics and diagnostics

Since their discovery just over 25 years ago, the single variable domain from heavy-chain-only antibodies plays a role in an increasing number of antibody-based applications. Structural and biophysical studies have revealed that the small, � kDa, single variable domain found in camelids displays versatility in target recognition. Such insight has served as the foundation to develop and engineer VHH domains with enhanced properties capable of targeting a range of therapeutically relevant protein antigens or low-molecular weight haptens. Furthermore, the modular nature of VHH domains allows them to be introduced into constructs that are simply not possible with conventional antibodies. Here, we review the structural and biophysical properties of VHH domains, highlight recent VHH-based therapeutics and diagnostics, and provide insight into VHH engineering that may pave the way to next-generation single domain antibody applications.

Impact statement

The development of novel antibody formats, beyond conventional antibodies, opens new possibilities in medical therapies, diagnostics, and general life science applications requiring affinity reagents. The camelid VHH domain, from heavy-chain-only antibodies, has emerged as a jack-of-all-trades module for novel affinity reagents. Applications include targeted cancer therapies, novel antimicrobial agents, conformation specific reagents, and tailor-made molecular switch entities. The breadth of unique uses for the VHH will continue to grow, opening new opportunities to treat and understand disease.


1. Introduction

SARS-CoV-2 uses its trimeric spike protein for binding to host angiotensin-converting enzyme 2 (ACE2) and for fusing with cell membrane to gain cell entry [1,2,3,4]. This is a multi-step process involving three separate S protein cleavage events to prime the SARS-2-S for interaction with ACE2 [2,3], and subsequent membrane fusion and cell entry. These processes involve different domains of the S protein interacting with host cell and other intracellular and extracellular components. Efficiency in each step could contribute to virulence and infectivity. Disrupting any of these steps could lead to medical cure.

The domain structure is very similar between SARS-S (UniProtKB: <"type":"entrez-protein","attrs":<"text":"P59594","term_id":"30173397","term_text":"P59594">> P59594) and SARS-2-S (UniprotKB: <"type":"entrez-protein","attrs":<"text":"P0DTC2","term_id":"1835922048","term_text":"P0DTC2">> P0DTC2). Both are cleaved to generate S1 and S2 subunits at specific cleavage sites ( Figure 1 A). S1 serves the function of receptor-binding and contains a signal peptide (SP) at the N terminus, an N-terminal domain (NTD), and receptor-binding domain (RBD). S2 ( Figure 1 A) functions in membrane fusion to facilitate cell entry, and it contains a fusion peptide (FP) domain, internal fusion peptide (IFP), two heptad-repeat domains (HR1 and HR2), transmembrane domain, and a C-terminal domain [2,3,5,6,7,8]. However, there are also significant differences between SARS-S and SARS-2-S. For example, the contact amino acid sites between SARS-S and human ACE2 (hACE2) [5,7,9,10] differ from those between SARS-2-S and hACE2 [11,12,13,14]. This may explain why some antibodies that are effective against SARS-S are not effective against SARS-2-S [4], especially those developed to target the ACE2 binding site of SARS-S [15]. In this article, numerous experiments on SARS-S are considered to facilitate comparisons and to highlight differences between the two.

Domain structure of SARS-S and SARS-2-S. (A) Key domains in SARS-S and SARS-2-S. SP, signal peptide NTD, N-terminal domain RBD, receptor-binding domain FP, fusion peptide IFP, internal fusion peptide HR, heptad repeats TM, transmembrane domain CT, cytoplasmic tail. The top and bottom numbers in each domain pertain to SARS-S and SARS-2-S, respectively. The red arrows indicate cleavage sites, and their numbers pertain to SARS-2-S (B) Alignment of SP between SARS-S (top) and SARS-2-S (bottom) (C,D) Alignment of two inter-domain segments (E) HR1 in SARS-S and SARS-2-S, together with the top view of a helix showing hydrophobic positions a and d on the same side (F) Hydrophobicity plot generated from DAMBE [16].


Results

Three systems were selected to test OptCDR's efficacy: a peptide from the capsid of hepatitis C (PDB: 1N64) ( Menez et al., 2003), fluorescein (PDB: 1FLR) ( Whitlow et al., 1995) and VEGF (PDB: 1CZ8) ( Chen et al., 1999). Several computational metrics are used to draw comparisons, including interaction energy defined as the minimized energy of the antigen–CDRs complex minus the energy of the CDRs and the energy of the antigen individually. It is approximated within CHARMM ( MacKerell et al., 1998) using the Van der Waals, electrostatics, bonds, angles, dihedral angles, improper dihedral angles and generalized Born with molecular volume integration implicit solvation energy functions. Contacts are defined as the number of CDR atoms within 3 Å of the antigen and polar contacts are determined using PyMOL ( DeLano, 2008). All computations were carried out on 3.0 GHz Intel Xeon processors with 4 GB of RAM. Each complete antibody library design was generated on its own processor and all were completed in <12 days of computations, with an average of about 9 days.

Hepatitis C capsid peptide

Hepatitis C is a virus that infects approximately 3.2 million people in the USA (http://www.cdc.gov/hepatitis/C/cFAQ.htm#statistics). The CDRs of antibody 19D9D6 (PDB: 1N64) ( Menez et al., 2003) that bind a peptide (residues 13–40) from the capsid of the hepatitis C virus with a Kd of 1.3 ± 0.1 nM are shown in Fig. 3A. This system was selected as a general test of OptCDR to generate promising design alternatives. We first examined the extent of improvements that can be achieved in the computationally accessible metrics of binding quality by only mutating the original antibody structure without altering the CDR canonical structures. The results obtained using IPRO show improvements in all three binding metrics (Table II). Interestingly, the mutations identified are confined to the CDRs with the fewest antigen contacts. This trend of predicting mutations in the periphery of the antibody binding pocket is consistent with a previous computational study that was experimentally validated ( Lippow et al., 2007). In this case, since 19D9D6 is already a high-affinity antibody, the results indicate that the dominant interactions in the center of the binding pocket are already effective and binding could only be improved further through repacking of the edges of the antibody–antigen interface.

Computational and experimental binding data for the antibodies

Antibody . Antigen . Interaction energy (kcal/mol) . Contacts . Polar contacts . Experimental Kd .
19D9D6 Hepatitis C capsid peptide −62.6 74 8 1.3 nM
IPRO affinity maturation of 19D9D6 Hepatitis C capsid peptide −78.1 to −80.1 81–87 14–15 NA
OptCDR designs Hepatitis C capsid peptide −88.2 to −104.8 88–115 18–23 NA
OptCDR design with antigen rearrangment Hepatitis C capsid peptide −123.6 to −175.8 88–112 23–31 NA
4-4-20 Fluorescein −49.5 22 4 0.7 nM
Boder et al. best design Fluorescein −78.7 23 4 48 fM
Fukuda et al. consensus design Fluorescein −67.8 22 5 ∼1 nM (wt 32 nM)
Fukuda et al. best design Fluorescein −70.4 24 3 0.88 nM (wt 32 nM)
Jermutus et al. consensus design Fluorescein −74.9 24 3 ∼37 pM
IPRO affinity maturation of 4-4-20 Fluorescein −77.8 17 1 NA
OptCDR designs Fluorescein −55.2 to −57.3 71–79 8–9 NA
PDB 1CZ8 VEGF-epitope 1 −110.8 86 21 0.11 nM
IPRO affinity maturation of 1CZ8 VEGF-epitope 1 −111.0 to −116.0 89–99 21–22 NA
OptCDR design VEGF-epitope 1 −82.0 to −109.0 73–110 11–25 NA
OptCDR design VEGF-epitope 2 −88.6 to −98.4 57–85 15 to 20 NA
OptCDR nanobodies VEGF-epitope 1 −82.2 to −92.6 79–86 20–22 NA
Antibody . Antigen . Interaction energy (kcal/mol) . Contacts . Polar contacts . Experimental Kd .
19D9D6 Hepatitis C capsid peptide −62.6 74 8 1.3 nM
IPRO affinity maturation of 19D9D6 Hepatitis C capsid peptide −78.1 to −80.1 81–87 14–15 NA
OptCDR designs Hepatitis C capsid peptide −88.2 to −104.8 88–115 18–23 NA
OptCDR design with antigen rearrangment Hepatitis C capsid peptide −123.6 to −175.8 88–112 23–31 NA
4-4-20 Fluorescein −49.5 22 4 0.7 nM
Boder et al. best design Fluorescein −78.7 23 4 48 fM
Fukuda et al. consensus design Fluorescein −67.8 22 5 ∼1 nM (wt 32 nM)
Fukuda et al. best design Fluorescein −70.4 24 3 0.88 nM (wt 32 nM)
Jermutus et al. consensus design Fluorescein −74.9 24 3 ∼37 pM
IPRO affinity maturation of 4-4-20 Fluorescein −77.8 17 1 NA
OptCDR designs Fluorescein −55.2 to −57.3 71–79 8–9 NA
PDB 1CZ8 VEGF-epitope 1 −110.8 86 21 0.11 nM
IPRO affinity maturation of 1CZ8 VEGF-epitope 1 −111.0 to −116.0 89–99 21–22 NA
OptCDR design VEGF-epitope 1 −82.0 to −109.0 73–110 11–25 NA
OptCDR design VEGF-epitope 2 −88.6 to −98.4 57–85 15 to 20 NA
OptCDR nanobodies VEGF-epitope 1 −82.2 to −92.6 79–86 20–22 NA

The various computational binding metrics are defined in the text at the start of the Results Section and all experimental Kd values were taken from the appropriate publications. VEGF-epitope 1 is the epitope bound by bevacizumab while epitope 2 is on the opposite side of VEGF. The nanobody designs have only three CDRs, all other antibodies have six.

Computational and experimental binding data for the antibodies

Antibody . Antigen . Interaction energy (kcal/mol) . Contacts . Polar contacts . Experimental Kd .
19D9D6 Hepatitis C capsid peptide −62.6 74 8 1.3 nM
IPRO affinity maturation of 19D9D6 Hepatitis C capsid peptide −78.1 to −80.1 81–87 14–15 NA
OptCDR designs Hepatitis C capsid peptide −88.2 to −104.8 88–115 18–23 NA
OptCDR design with antigen rearrangment Hepatitis C capsid peptide −123.6 to −175.8 88–112 23–31 NA
4-4-20 Fluorescein −49.5 22 4 0.7 nM
Boder et al. best design Fluorescein −78.7 23 4 48 fM
Fukuda et al. consensus design Fluorescein −67.8 22 5 ∼1 nM (wt 32 nM)
Fukuda et al. best design Fluorescein −70.4 24 3 0.88 nM (wt 32 nM)
Jermutus et al. consensus design Fluorescein −74.9 24 3 ∼37 pM
IPRO affinity maturation of 4-4-20 Fluorescein −77.8 17 1 NA
OptCDR designs Fluorescein −55.2 to −57.3 71–79 8–9 NA
PDB 1CZ8 VEGF-epitope 1 −110.8 86 21 0.11 nM
IPRO affinity maturation of 1CZ8 VEGF-epitope 1 −111.0 to −116.0 89–99 21–22 NA
OptCDR design VEGF-epitope 1 −82.0 to −109.0 73–110 11–25 NA
OptCDR design VEGF-epitope 2 −88.6 to −98.4 57–85 15 to 20 NA
OptCDR nanobodies VEGF-epitope 1 −82.2 to −92.6 79–86 20–22 NA
Antibody . Antigen . Interaction energy (kcal/mol) . Contacts . Polar contacts . Experimental Kd .
19D9D6 Hepatitis C capsid peptide −62.6 74 8 1.3 nM
IPRO affinity maturation of 19D9D6 Hepatitis C capsid peptide −78.1 to −80.1 81–87 14–15 NA
OptCDR designs Hepatitis C capsid peptide −88.2 to −104.8 88–115 18–23 NA
OptCDR design with antigen rearrangment Hepatitis C capsid peptide −123.6 to −175.8 88–112 23–31 NA
4-4-20 Fluorescein −49.5 22 4 0.7 nM
Boder et al. best design Fluorescein −78.7 23 4 48 fM
Fukuda et al. consensus design Fluorescein −67.8 22 5 ∼1 nM (wt 32 nM)
Fukuda et al. best design Fluorescein −70.4 24 3 0.88 nM (wt 32 nM)
Jermutus et al. consensus design Fluorescein −74.9 24 3 ∼37 pM
IPRO affinity maturation of 4-4-20 Fluorescein −77.8 17 1 NA
OptCDR designs Fluorescein −55.2 to −57.3 71–79 8–9 NA
PDB 1CZ8 VEGF-epitope 1 −110.8 86 21 0.11 nM
IPRO affinity maturation of 1CZ8 VEGF-epitope 1 −111.0 to −116.0 89–99 21–22 NA
OptCDR design VEGF-epitope 1 −82.0 to −109.0 73–110 11–25 NA
OptCDR design VEGF-epitope 2 −88.6 to −98.4 57–85 15 to 20 NA
OptCDR nanobodies VEGF-epitope 1 −82.2 to −92.6 79–86 20–22 NA

The various computational binding metrics are defined in the text at the start of the Results Section and all experimental Kd values were taken from the appropriate publications. VEGF-epitope 1 is the epitope bound by bevacizumab while epitope 2 is on the opposite side of VEGF. The nanobody designs have only three CDRs, all other antibodies have six.

CDR–hepatitis C capsid peptide complexes. The capsid peptides are shown as red spheres, the CDRs are shown as orange ribbons, and CDR residues within 4 Å of the peptide are explicitly shown. All images are from the same perspective. (A) The natural antibody–peptide complex (PDB: 1N64). (B) An OptCDR design with no peptide conformational change. (C) An OptCDR design with peptide conformational change.

CDR–hepatitis C capsid peptide complexes. The capsid peptides are shown as red spheres, the CDRs are shown as orange ribbons, and CDR residues within 4 Å of the peptide are explicitly shown. All images are from the same perspective. (A) The natural antibody–peptide complex (PDB: 1N64). (B) An OptCDR design with no peptide conformational change. (C) An OptCDR design with peptide conformational change.

We next used OptCDR to design three sets of antibody CDRs to bind the peptide instead of relying on only adding point mutations to 19D9D6. We first assumed a conservative posture by imposing harmonic constraints that insured that the antigen conformation did not change significantly upon binding. The three generated designs exhibit highly diverse antigen locations/orientations, canonical structure selections and amino acid sequences, but all share the groove that is typically observed for peptide-binding antibodies. Significant improvements in all computational binding metrics are observed (Table II) over the case of using only mutations. Table III depicts the predicted lowest-energy amino acid sequences for the CDRs of the structure shown in Fig. 3B. By combining the predicted mutations, a library of CDRs can be generated (maximum library size = 1.1 × 10 14 ). By prioritizing mutations based on their binding scores, libraries of any smaller size can be culled from the original.

A library of hepatitis C capsid binding antibody CDRs

The CDR sequences and predicted mutations to them to form a library of up to 1.1 × 10 14 antibodies that can all bind the peptide from the capsid of hepatitis C.

A library of hepatitis C capsid binding antibody CDRs

The CDR sequences and predicted mutations to them to form a library of up to 1.1 × 10 14 antibodies that can all bind the peptide from the capsid of hepatitis C.

Finally, we removed the harmonic constraints on the antigen, allowing for its conformation to radically change in response to the interaction energy minimization step, and used OptCDR to generate three additional sets of antibody CDRs to bind the peptide. As seen in Table II, allowing conformational changes to the peptide in response to energy minimization led to additional improvements in interaction energy and polar contacts, although no effect on the number of contacts. It should be noted that the conformational changes were not forced or pre-specified, but came about as a response to interactions with the CDRs during the IPRO step (i.e. step 3) of OptCDR. In this case, the conformational changes were all between 4.2 and 5.2 Å RMSD from the initial peptide conformation. These results demonstrate how antigen conformational changes upon antibody binding may be an important contributor in informing antibody–antigen interactions. OptCDR allows for a user-specified presence, absence and modulation of the strength of any imposed harmonic constraints.

Fluorescein

We next turned our attention to the hapten fluorescein for a comparison with experimental results. Anti-fluorescein antibodies have been used as the test system by several experimental directed evolution efforts ( Boder et al., 2000 Jermutus et al., 2001 Fukuda et al., 2006) with different display technologies to improve the binding of an scFv derived from the anti-fluorescein antibody 4-4-20. This is a particularly interesting system because 4-4-20's affinity for fluorescein is near the affinity ceiling of the tertiary immune response (0.7 ± 0.3 nM) ( Boder et al., 2000). Boder et al. (2000) used yeast surface display to identify an scFv with 14 mutations exhibiting a ∼6500-fold improvement in Kd, while Fukuda et al. (2006) and Jermutus et al. (2001) used mRNA and ribosome display, respectively, to identify mutants with ∼30-fold improvement in Kd.

First, we used IPRO to evaluate these experimentally identified mutants: the best mutants identified by Boder et al. (2000) and Fukuda et al. (2006) and the consensus mutants identified by Fukuda et al. (2006) and Jermutus et al. (2001). Since some of the mutations were in framework regions, the entire variable domains were used to create the mutants, not just the CDRs. SwissParam (http://swissparam.ch/) was used to create the topology and parameter files needed in CHARMM for fluorescein. Once the mutant antibodies were modeled, we extracted their CDRs, so the calculated interaction energies could be directly compared with OptCDR results. Table II shows the computational and experimental improvements over the wild-type scFv of the four mutants. Note that the rank order of the mutants in terms of improvement over wild-type using Kd and interaction energy, as quantified in OptCDR, match. Furthermore, the H3 CDR in the Jermutus et al. (2001) mutant shows the highest RMSD from the wild-type structure, which matched the experimental observation of increased flexibility of this CDR due to the removal of a salt bridge.

Next, IPRO was used to computationally affinity mature the CDRs of antibody 4-4-20, leading to numerous mutations (40 total) in all CDRs except H2. The computational binding results are detailed in Table II. Although the improvement in interaction energy over the wild-type antibody is not quite as large as that of the best experimental mutant, it does surpass the calculated energies for all other mutants. We also used OptCDR to design two sets of CDRs to bind fluorescein. Their computational binding metrics are given in Table II, structures in Fig. 4 and amino acid sequences in Table IV. Interestingly, the interaction energies of the OptCDR designs do not reach the same levels as those of the computationally and experimentally affinity-matured versions of 4-4-20, but they do surpass the wild-type antibody. Both designs share a number of features that appear favorable to binding. First, in both cases, fluorescein is positioned within a deep cavity between the L3 and H2 CDRs on one side and the H3 CDR on the other. Both designs have long H3 CDRs folded mostly over the top of the fluorescein molecules to trap them in place, and it is this position of the H3 CDRs that lead to the notable increase in contacts between the experimental and OptCDR designs (Table II). For both designs, the edges of the cavity have polar residues with each fluorescein oxygen involved in at least one polar contact. Finally, the sides of both cavities are composed of aliphatic and aromatic residues stabilizing the core hydrophobic portion of fluorescein. We hypothesize that the H3 CDRs of the unbound designs are sufficiently flexible to allow fluorescein access to the binding pocket.

Two libraries of fluorescein-binding antibody CDRs

The amino acid sequences of the two OptCDR designed fluorescein-binding antibodies and the predicted mutations to the CDRs to form libraries of antibodies.

Two libraries of fluorescein-binding antibody CDRs

The amino acid sequences of the two OptCDR designed fluorescein-binding antibodies and the predicted mutations to the CDRs to form libraries of antibodies.

CDR–fluorescein complexes. Fluorescein is shown as cyan spheres with oxygens and hydrogens shown in red and white, respectively. The CDRs are shown as orange ribbons and all CDR residues within 4Å of fluorescein are explicitly shown. All the complexes are shown from the same perspective relative to the antibody binding pocket. (A) The structure of antibody 4-4-20. (B and C) The two OptCDR fluorescein-binding designs.

CDR–fluorescein complexes. Fluorescein is shown as cyan spheres with oxygens and hydrogens shown in red and white, respectively. The CDRs are shown as orange ribbons and all CDR residues within 4Å of fluorescein are explicitly shown. All the complexes are shown from the same perspective relative to the antibody binding pocket. (A) The structure of antibody 4-4-20. (B and C) The two OptCDR fluorescein-binding designs.

Vascular endothelial growth factor

Finally, OptCDR designs for binding VEGF are contrasted against an affinity-matured antecedent of the antibody medication bevacizumab ( Chen et al., 2001) to examine OptCDR's epitope targeting abilities. VEGF has been shown ( Willett et al., 2004) to promote tumor proliferation and growth. A number of anti-VEGF antibody-based drugs, including bevacizumab ( Chen et al., 2001), with low nanomolar affinity are available. The resolved structure (PDB: 1CZ8) of an affinity-matured bevacizumab antecedent shows VEGF situated within a pocket primarily formed by the heavy chain CDRs ( Chen et al., 1999). In all OptCDR results, a harmonic constraint was used to prevent the structure of VEGF from changing significantly during calculations.

IPRO-based computational affinity maturation of 1CZ8 led to only minimal improvements in the interaction energy, as well as the number of contacts and polar contacts, which is a testament to the thoroughness of the experimental affinity maturation that 1CZ8 has already undergone. When we used OptCDR to predict novel CDRs to bind the same epitope of VEGF targeted by 1CZ8, the best design, shown in Fig. 5B, has binding metrics that are comparable with the existing antibody (slightly greater interaction energy and a few more contacts and polar contacts). Thus, the range of binding metrics for OptCDR (Table II) only reaches the levels of 1CZ8, although the predicted structures are all notably different and most of which exhibit the planar binding pocket expected for protein-binding antibodies.

CDR–VEGF complexes. VEGF is shown as green spheres, the CDRs are shown as orange ribbons, and CDR residues that are within 4 Å of VEGF are explicitly shown. All images are from the same perspective relative to the antibody binding pocket. (A) The structure of PDB 1CZ8. (B) The best OptCDR design generated to bind the same portion of VEGF as bevacizumab using all six CDRs while (C) is the best design to bind an epitope on the opposite side of VEGF. (D) The best nanobody OptCDR design.

CDR–VEGF complexes. VEGF is shown as green spheres, the CDRs are shown as orange ribbons, and CDR residues that are within 4 Å of VEGF are explicitly shown. All images are from the same perspective relative to the antibody binding pocket. (A) The structure of PDB 1CZ8. (B) The best OptCDR design generated to bind the same portion of VEGF as bevacizumab using all six CDRs while (C) is the best design to bind an epitope on the opposite side of VEGF. (D) The best nanobody OptCDR design.

Next, we explored another set of OptCDR designs by targeting an epitope on the opposite side of VEGF from the portion bound by 1CZ8 and our other designs. To the best of our knowledge, this portion of VEGF is not recognized by cellular VEGF receptors or any designed antibodies. One of the predicted designs is shown in Fig. 5C and the computational binding metrics are given in Table II. Even though the obtained designs are very different as they target a completely different epitope of VEGF, the computational binding metrics achieved are quite similar in value. The results obtained demonstrate the efficacy of OptCDR to generate designs that bind VEGF with equivalent computational binding characteristics as the wild-type antibody with novel CDRs or targeting a different epitope of the VEGF molecule, alluding to the built-in redundancy of molecular recognition.

We decided to further explore VEGF-binding designs by focusing on only three out of the six CDRs, as in nanobodies. Nanobodies are single-domain proteins derived from the variable domain of heavy chains from a special subset of antibodies in camelids that lack light chains and thus have only three CDRs instead of six. OptCDR was used to generate nanobody CDR designs by considering only the H1, H2 and H3 CDRs. Despite the reduction in the number of structural degrees of freedom, OptCDR identified designs based solely on the heavy chain CDRs (Table II) that had similar computational binding metrics to the six CDR designs. This surprising finding is consistent with the experimental observation that nanobodies can have binding affinities that are equivalent to antibodies despite their smaller size. We believe that OptCDR achieved this through the selection of longer than typical canonical structures, especially for the H3 CDR, as is typical in experimental nanobodies. In addition, the absence of the light chain allows the antigen to assume positions and orientations that are normally prohibited due to steric clashes. A representative design plotted in Fig. 5D illustrates how the selection of longer structures for the H3 domain counteracts the loss of the light chain.


Discussion

The successful identification of two DP mutants, H-K30Q/H-E54H and L-N34D/L-H91S with affinity improved over or comparable to that of the WT D3h44 demonstrates the success of the double-point mutation strategy. The key process for selecting the four DP mutants from the 3D models of 59 sets of residue pairs was a visual inspection of the change in interactions between the WT and the DP mutant. The successful acquisition of two DP mutants also indicated that the analysis of the interactions between the mutated residues, antibody residues close to the mutated residues, the antigen, and surrounding water molecules is informative and useful for the design of novel cooperative DP mutants. This work also indicates that it may be possible to predict the binding affinity of designed antibodies based on information about their interactions.

The H-K30Q/H-E54H DP mutant exhibited a 1.5-fold increase in affinity (29 pM) over that of the WT D3h44 (45 pM). There are two possible reasons for the improvement in affinity of the DP mutant. One is the reinforcement of the interaction between the antigen and the antibody by the additional hydrogen bond between H-H54 and S202 of the antigen (Fig. 6b). The other is the replacement of weaker electrostatic and CH-O interactions between H-K30 and H-E54 of the WT (Fig. 6a) by the strong hydrogen bond between H-Q30 and H-H54 of the DP mutant (Fig. 6b). The interaction between H-K30 and H-E54 is not a strong ionic interaction, but a less strong electrostatic interaction, based on their distance and angle. These preferable interactions between the antigen and the DP mutant, and between the mutated residues, are a good example of the cooperative mutation produced by the double-point mutation strategy combined with interaction analysis.

(a) Interactions of H-K30 and H-E54 of the wild-type (WT). (b) Interactions of H-K30Q and H-E54H. (c) Interactions of L-N34 and L-H91 of the WT. (d) Interactions of L-N34D and L-H91S. The antigen and the heavy chain of the antibody are colored in magenta and green, respectively. A water molecule is depicted as a red sphere.

The other successful DP mutant was L-N34D/L-H91S, with a 1.8-fold increase in affinity (25 pM) over that of the WT (45 pM). This case is informative, because the DP mutation including residue 34, which does not contact with the antigen, improved the binding affinity. The increase in affinity may have occurred because the mutation provided additional weak interactions, such as CH–O, while preserving and/or replacing the interactions around L-34 and L-91. Two new CH-O interactions were formed: between L-D34 and L-S91 and between L-S91 and L-Y32 (Fig. 6d). Although a CH–O interaction is a weak interaction, two additional CH–O interactions contributed to the stabilization of the whole antibody/antigen complex and increased the affinity of the DP mutant (Fig. 6d). With respect to the interactions between the antigen and the mutated residues, the interaction networks of the WT/antigen complex (K169 of antigen, L-H91 and L-N34) were successfully replaced by the interaction networks (K169 of antigen, L-S91 and L-D34) of the DP mutant/antigen (Fig. 6c,d). Overall, a 1.8-fold increase in the affinity of the L-N34D/L-H91S DP mutant appeared to be achieved. This is another example of the cooperative mutations produced by the “DP mutation” strategy with the interaction analysis.

One of the unique features of the interaction analysis is the inclusion of weak interactions such as CH–O and CH–π. These interactions are recognized as important for the structure-based drug design of small molecules 19,20 as described in “Introduction”. In general, these weak interactions require small desolvation energy. On the contrary, high desolvation energy is required to form a strong hydrogen bond interaction. The balance between gain (formation of interaction) and loss (desolvation) of these weak interactions have attracted attention. In the case of this study, such weak interactions were frequently observed. For example, 14 and 16 CH–O interactions between mutated residues and their surrounding residues were found in the WT and L-N34D/L-H91S DP mutant, respectively. These numbers indicate that the effect of these weak interactions is not negligible, even though the individual interaction is weak.

Although two DP mutants with affinity improved over or comparable to the WT were successfully obtained, the current procedure for the selection of DP mutants is not perfect, because not all of the four DP mutants showed good affinity. For the H-A101S/H-A102H DP mutant, the affinity decreased to 150 pM, a 3.3-fold reduction compared to the WT. This reduction was probably due to the size of the mutated residues, especially H-H102. Based on the H-A101S/H-A102H mutant model, it was expected that mutations would increase favorable interactions, as described in the “Results” section. However, the mutation from the smaller H-A102 to the larger H-H102 triggers conformational changes of I152 of the antigen and L-Y49, by van der Waals collisions (Fig. S9). The instability stemming from the conformational changes may cause the 3.3-fold reduction of the affinity.

The H-Y33Q/H-A101R DP mutant showed complete loss of affinity. The main reason for the loss of the affinity appears to arise from the conformational preference of the mutated residues. The conformational preference of H-Q33 of the model is 3.4% of all of 54 conformations of Gln and that of the H-R101 model is 0.06% of all of 81 side chain conformations of Arg, according to the rotamer library of amino acids 28 (Fig. S10 and Table S3). Thus, the preferable interactions between H-Q33 and H-R101 observed in the model would rarely occur the probability is estimated at 3.4% × 0.06% = 0.002%.

The other key finding was that cooperative DP mutants could not be obtained when a typical procedure, the combination of improved SP mutants, was taken. In the case of the H-K30Q/H-E54H cooperative DP mutant (KD = 29 pM), the individual SP mutant H-K30Q showed a 2.2-fold reduction in affinity (KD = 97 pM) compared with the WT (KD = 45 pM), and the H-E54H SP mutant was not expressed. Using this approach, the chance to examine the affinity of the mutant with a combination of the residue with a 2.2-fold reduction of the affinity (H-K30Q) and the residue not expressed (H-E54H) is rare. In the L-N34D/L-H91S DP mutant (KD = 25 pM), the individual SP mutant L-N34D exhibited a 6.9-fold reduction in affinity (KD = 310 pM) and the affinity of the L-H91S SP mutant was observed to be 17,000 pM, 1/378 of that of the WT. The probability of developing a mutant with the combination of a residue with a 6.9-fold reduction (L-N34D) and a residue with a 378-fold reduction (L-H91S) is very low. A combination of the negative effects of an SP mutation—a reduction of affinity or no expression—and another SP mutation can result in a positive effect: an increase in affinity. Simultaneous cooperative mutations arising from the complementarity of two residues are thought to be important to this phenomenon.

Because of resource limitations, the number of mutants selected for experimental evaluation in the first round of the design was set to ten. Despite the generation of 3D model structures for 59 sets of residue pairs with 19 amino acids for each residue position, four DP mutants were selected, because they were the only DP mutants that had passed visual inspection of interactions. This low number suggested that the simultaneous mutation of two contacting residues with favorable interactions might be rare.

Another possibility is that not all of the potentially active mutants with favorable interactions had been pursued. In the evaluation of the mutant models, visual inspection was carried out on only the top 20 of 361 DP mutant models (every pair set of two amino acid positions, 19 × 19 = 361), based on the MM/GBVI binding energy value calculated using MOE 29 . The other mutant models were discarded. The number of models to be visually inspected was increased to a maximum of 80 if a mutant with favorable interactions did not exist among the top 20 models. Neither the total nor the electrostatic 12 MM/GBVI binding energy correlated with the change in experimental binding affinity (Fig. S11). Thus, the filtering of the mutant models according to MM/GBVI binding energy is not ideal, and there is a possibility that not all of the DP mutants with favorable interactions were subjected to further visual inspection. Although not ideal, filtering by MM/GBVI binding energy followed by visual inspection of the interactions appeared to be effective, given that the first round of design led to the identification of two DP mutants with improved or comparable affinity to that of the WT.

The affinity of the triple-point mutant L-N34D/L-H91S/H-N57Y was increased to 7.6 pM, a 5.9-, 3.2- and 2.0-fold increase over the affinity of the WT, L-N34D/L-H91S DP mutant, and H-N57Y SP mutant, respectively. The triple-point mutant was created to add the positive effects of two distantly located mutants, L-N34D/L-H91S and H-N57Y, which are 17 Å apart in 3D (Fig. 1 and Fig. S12). The 5.9-fold increase in affinity of the triple-point mutant was nearly equal to the product of the fold increase of the N34D/L-H91S DP mutant, 3.2, and that of H-N57Y SP mutant, 2.0. In this case, the assumption that the effect of the mutations of distant locations in 3D are independent each other and could be additive was validated. Although the starting point of the affinity of the WT was very high (45 pM), a 5.9-fold increase of the affinity of the triple-point mutant (7.6 pM) was achieved. This achievement reinforced the importance of 3D information in the design of mutants, because the addition of the positive effect of two mutants is thought to be possible only when the mutated residues are distant in 3D.

The change in the binding affinity, KD, is primarily governed by the change of the dissociation constant koff. In the design process, the 3D structures of the antigen/antibody complex were examined with the focus on short-range interactions around the mutation residues. It has been reported that the dissociation process is governed by short-range interactions between proteins 30 . This may be the possible explanation of our results on koff. In addition, the affinity changes derived from structure-based antibody design based on SP mutations are typically governed by change in koff 16 . The design of mutants based on the 3D structure of the antigen/antibody complex may not produce a significant effect on the association rate, kon. Design based on the complex, and which is intended to stabilize the complex, did affect the dissociation rate, koff. Thus, koff was improved if the design was successful and deteriorated if failed.

The stability of the mutants was not explicitly considered in the design process in the same way as kon and koff. Although some studies have reported that affinity and stability have a trade-off relationship 31 , the change of the melting temperature ∆Tm fell within the range of − 2 to + 2 °C for most of our mutants (Fig. 5). No correlation between the affinity change and ∆Tm was observed. The lack of correlation between the affinity and stability changes may indicate that the consideration of the preservation of the intra-antibody interactions in the design process contributed to preventing the mutants from significant reduction in stability.

Ten DP and SP mutants were prospectively designed using a double-point mutation strategy, with interaction analysis based on the 3D structure of the mutant models. Two DP and two SP mutants with an affinity improved over or comparable to that of the WT were successfully obtained. Of the four individual SP mutants, which involved the constituent residues of the two active DP mutants, three showed decreased affinity, and one was not expressed. These results indicated that the two active DP mutants are hard to be obtained using a SP mutation strategy. The triple-point mutant, which is a combination of the active DP and SP mutants, was designed and a further increase in affinity was achieved.

Although the DP mutation strategy with interaction analysis was successful, there are several issues still to be overcome. The most critical issue is the manual identification of the interactions, which is labor-intensive and time-consuming. Automating the identification of interactions based on the 3D structure of the mutation model is a current focus of our research, and the software for automatic identification of interactions has been prototyped. As preliminary test cases, this prototype software was applied to the ten mutants designed, and it reproduced the interactions obtained manually. The details of the automatic identification of interactions will be published elsewhere. Another important issue is the best way to evaluate the changes of interactions introduced by mutation. Currently the evaluation is not quantitative and is based on personal experience of the structure-based design. Further investigation of the relationships between the changes in affinity and changes of the interactions, based on data provided by the automatic interaction detection software and the automatic mutant modeling software, is required. Another issue is the modeling of the mutant/antigen complex including water molecules. Regarding the relationship between water rearrangement in the mutant models and affinity change, any significant relation between the extent of water rearrangement and the fold improvement in affinity was not observed (Fig. S13). Affinity improvement occurred in both cases where the water rearrangement is small (RMSD < 0.2 Å) and relatively large (RMSD > 0.6 Å). These observations indicate that the water rearrangement of the mutant models did not affect the results largely in our study. However, this may not be true in all cases. As an example of water rearrangement, it was found that a cavity produced by the mutation, which may accommodate a water molecule, was not fulfilled in some mutant models when all possible combinations of DP mutants were examined. Thus, the modeling of DP mutants involving water molecules is currently under consideration. For example, a short molecular dynamics simulation of solvated antigen/antibody complex with restraining protein atoms is an option for realizing water rearrangement in an appropriate manner. In addition, free-energy perturbation of DP and multiple mutants for affinity prediction is thought to be one of the directions for improving the accuracy of the affinity prediction 32 . Though we assumed that the structure of the overall antigen–antibody complex was retained upon mutation 33 , there might be the cases where DP and even SP mutation induces a large structural change, which is outside of the scope of the mutant modeling method we adopted in this study. Other methods may be necessary to study such a scenario. The inclusion of the effect of side chain flexibility is another issue, as seen in the failure of the design of the H-Y33Q/H-A101R DP mutant. After resolving the issues described above, we hope to extend the DP mutation strategy to a triple-point mutation strategy.


How immunity to respiratory syncytial virus develops in childhood, deteriorates in adults

The leading infectious cause of severe respiratory disease in infants, respiratory syncytial virus (RSV), is also a major cause of respiratory illness in the elderly. Approved vaccines do not yet exist, and despite the development of partial immunity following infection during childhood, individuals remain susceptible to RSV reinfection life-long. A comprehensive characterization of the antibody-response to RSV published on April 21st in PLOS Pathogens advances our understanding of the human immune response against RSV and has implications for vaccine design.

RSV is nearly ubiquitous, and most children are born with some protective immunity conveyed by maternal antibodies. As the maternal antibodies wane over time, infants become susceptible, and are often infected for the first time between nine months and two years of age.

Studies over the past three decades have explored the antibody responses before and after RSV infection in different age groups. We know that human antibodies that can mediate the destruction (or neutralization) of the virus target the two major proteins on the virus surface, namely the attachment protein G and the fusion protein F. However, which antibody combination conveys the best immune protection, and why RSV infections recur throughout life remain open questions.

To address them, Surender Khurana and colleagues from the US Food and Drug Administration in Silver Spring, USA, first performed a comprehensive and unbiased analysis of the human antibody response to the RSV F and G proteins in infants before and after RSV infection. They then characterized the changes in the response over time by analyzing antibodies from children, adolescents, and adults.

The blood of young infants, the researchers found, contains maternal antibodies that recognize several parts of both the F and G proteins. In older infants that had been infected with RSV, they saw a dramatic expansion in both quantity and diversity of the antibodies that recognized the G protein. Surprisingly, infection prompted only a modest increase in the antibody repertoire against the F protein. Looking at changes over time, the researchers found that the antibodies against the F protein continued to expand with age whereas those against the G protein weakened.

Because the G protein sequence varies between RSV strains, whereas the F protein is highly conserved among strains, some vaccines under development use only the more tractable F protein as a vaccine antigen. The results here--strong expansion of anti-G responses in infants following infection as well as strong anti-F responses but weakened anti-G responses in adults--suggest that such a vaccine design might be problematic. On the other hand, the fact that the strong anti-G responses seen in children target a relatively conserved region in the G protein suggest that variability in other parts of the G protein does not necessarily compromise G's utility as a vaccine antigen.

Taken together, the researchers say, their results suggest "an unlinked evolution of the antibody responses to F and G proteins in humans," and propose that "the significant drop in anti-G antibody levels in adults may be a factor in sustained susceptibility to RSV infections throughout life." Consequently, they state that their findings "imply the need to include G proteins in future RSV vaccines in order to boost the anti-G responses."


Study funding

The authors received financial support from the EPSRC, BBSRC, ERC, and the Frances and Augustus Newman Foundation. This work was supported by the programs “Investissements d'avenir” ANR-10-IAIHU-06, “Santé Publique France,” and grants from NIH (R01NS103848) and CDC (UR8/CCU515004). Collection of samples at Massachusetts General Hospital was funded by Prion Alliance. The funders played no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.


Background

Along with sanitation, vaccines are the most effective and economic public health tools for control of infectious disease [1]. However, vaccine development faces a number of challenges, such as overcoming the limited effectiveness of a number of vaccines, the need for frequent vaccine reformulation, as well as a complete lack of vaccines for some diseases. A central goal of vaccination is to generate long lasting and broadly protective immunity against target pathogens, but this goal is hampered by the variability of both the target pathogens and the human immune system [2]. Current practical solutions to the problem include polyvalent vaccines such as those being developed for dengue virus [3] or seasonal vaccine reformulation against influenza [4].

The majority of traditional vaccines provide protection through neutralizing antibodies and T cells alone rarely offer protection and prevention of diseases. However, they participate in reduction, control, and clearance of intracellular pathogens and have been linked with protective immunity against a number of viral pathogens [5–8]. The biggest success of immunological bioinformatics is the development of algorithms for prediction of peptide binding affinity to the human leukocyte antigen (HLA) - one of the rate limiting steps in T cell-based immune response [9]. Although current forms of these algorithms are highly accurate [10–12], the output alone is not enough to inform the selection of epitopes for therapeutic applications. In the conceptual framework for reverse vaccinology, Rino Rappouli described in silico predictions of immune epitopes from biological sequence data as a "naïve approach" when compared with experimental elucidation immunogenic peptides. Many parameters of a good vaccine target conferring efficient, lasting immunity, still remain to be considered after prediction of HLA binding: multiple rate-limiting steps of peptide pre-processing, confirming in vivo expression, considering dynamics of expression in different developmental stages and cellular environments, presence of epitope across pathogen population, response across host population, epitope stability over time, and others [13]. Here, we address the issue of variability by modifying the antigen selection step with a computational method for selecting multiple T cell targets from functionally homologous protein regions.

Traditionally, vaccine targets are selected from conserved regions in the genome of the pathogen in question, with the aim of conferring broad and lasting immunity. The first step is a variability analysis performed by calculating the frequency of nucleotides or amino acids on each position in a multiple sequence alignment (MSA) of homologous genes or proteins [14]. Regions, in which several consecutive residues show high conservation (typically >90% conservation is chosen as the threshold), are then further analyzed for immunogenic potential either by computational predictions, experimental testing, or a combination thereof. This systematic exclusion of low frequency variants when using traditional approaches [15–19] represents a major limiting factor, since immunogenic potential does not always correlate with the frequency in the viral population - both rare and common peptides can be immunogenic and valuable in vaccine constructs aiming for broad coverage [20].

Since the human immune system's evolution occurs on a significantly longer time scale than rapidly mutating pathogens [21], high selective pressure causes them to alter expression of some immunogenic antigens faster than the immune system can evolve to keep up with the changes [22]. The HLA binding affinity of a peptide relative to its frequency in a viral or malignant cell population is known as its targeting efficiency (TE). It has been shown that the TE of peptides varies in different organisms, and in some highly variable viruses it tends to be low [20]. Regions of high TE comprise peptides that are highly conserved, most likely owing to the protein's functional importance limiting the capacity of a pathogen to alter the protein while maintaining its fitness [23]. Regions of low TE comprise one or more peptides, potentially all of high HLA binding affinity, but each of them will have a low frequency in the pathogen population. For rapidly mutating viruses, such as RNA viruses [24] the selective pressure exerted on HLA-binding peptides, means that host immunity will often, and in some cases preferentially, target low frequency epitopes [20].

Selecting vaccine targets from protein regions with conserved HLA binding

We propose a novel method and visualization scheme for assessing the stability of protein regions for T cell target discovery, which takes the evolutionary relationship between HLA and pathogen epitopes into consideration. This method is based on analyzing columns of suitably sized sliding windows (from here on termed "blocks") from the rows of sequences in an MSA (Figure 1). An MSA of homologous protein sequences can be performed using a number of algorithms [25], and blocks of peptides of a given size (usually 8-11 amino acids long for HLA class I restricted epitopes and 13-25 amino acids long for HLA class II restricted T-cell epitopes) are extracted from each position in the alignment. The number of peptides in each block indicates the diversity of the block, for which Shannon entropy and consensus frequency can be calculated as informative metrics [26].

Extraction of blocks from MSA. Subdivision of an MSA into blocks of peptides, l amino acids in length. In this example, l = 9. Block 1 is highlighted in blue. Moving the sliding window to the right in the MSA in increments of one position will give all blocks blocks in the MSA.

In order to identify potential T cell targets, HLA binding affinities are predicted for all peptides in all blocks. Because blocks are extracted from aligned regions of homologous proteins, it is likely that the peptides within a given block display high sequence homology and the majority show similar HLA binding properties even when sequence variations exist. Similarly, the regions surrounding a block will be of high homology, thus increasing the likelihood that peptides from the same block will be processed and presented on the surface of target cells in a similar fashion [27].

Blocks of one or more peptides that are all predicted to bind to the same HLA alleles with similar affinity are potentially valuable targets for polyvalent vaccine designs. This allows for simultaneous immunization with several epitopes - a necessary tactic against highly mutating viruses in which mutations introducing drug resistance can occur within a single day [28, 29]. We previously used a rudimentary version of the block conservation analysis for vaccine target discovery in dengue virus (DENV) [30] and reported a 10-fold larger number of potential CD8 + vaccine target candidates as compared to an earlier benchmark study of DENV vaccine target candidates [31]. We here formalize the approach and present a software implementation. To further demonstrate the utility of block conservation, we performed an analysis of HLA class I epitopes in influenza A H7N9 hemagglutinin (HA). The software is integrated into a freely available web service at http://met-hilab.cbs.dtu.dk/blockcons/.


Abbreviations

Complementary Determined Region of the Antibody.

antigen-binding fragment of antibody that includes one complete light chain paired with one heavy chain fragment containing the variable domain and the first constant domain.

antigen-binding fragment of the antibody that includes the variable domain of the heavy chain.

antigen-binding fragment of antibody that includes variable domains of heavy and light chains.

antigen-binding fragment of the antibody that includes the covalently linked variable domains of the heavy and light chains.

sample standard deviation.

FP, TN, FN – true positives, false positives, true negatives, and false negatives, respectively.

Receiver Operating Characteristics.


Monoclonal antibodies targeting nonstructural viral antigens can activate ADCC against human cytomegalovirus

1 Division of Infection and Immunology, School of Medicine, Cardiff University, Cardiff, United Kingdom.

2 Cardiff Metropolitan University, Cardiff, United Kingdom.

3 Department of Medicine, University of Cambridge, Cambridge, United Kingdom.

4 University of Glasgow-MRC Centre for Virus Research, Glasgow, United Kingdom.

5 Cambridge Institute for Medical Research, University of Cambridge, Cambridge, United Kingdom.

Address correspondence to: Richard J. Stanton, Infection and Immunity, School of Medicine, Henry Wellcome Building, Heath Park, Cardiff CF14 4XN, United Kingdom. Phone: 44.0.7969.148916 Email: [email protected]

Authorship note: VMV and IM contributed equally to this work.

Find articles by Vlahava, V. in: JCI | PubMed | Google Scholar

1 Division of Infection and Immunology, School of Medicine, Cardiff University, Cardiff, United Kingdom.

2 Cardiff Metropolitan University, Cardiff, United Kingdom.

3 Department of Medicine, University of Cambridge, Cambridge, United Kingdom.

4 University of Glasgow-MRC Centre for Virus Research, Glasgow, United Kingdom.

5 Cambridge Institute for Medical Research, University of Cambridge, Cambridge, United Kingdom.

Address correspondence to: Richard J. Stanton, Infection and Immunity, School of Medicine, Henry Wellcome Building, Heath Park, Cardiff CF14 4XN, United Kingdom. Phone: 44.0.7969.148916 Email: [email protected]

Authorship note: VMV and IM contributed equally to this work.

Find articles by Murrell, I. in: JCI | PubMed | Google Scholar

1 Division of Infection and Immunology, School of Medicine, Cardiff University, Cardiff, United Kingdom.

2 Cardiff Metropolitan University, Cardiff, United Kingdom.

3 Department of Medicine, University of Cambridge, Cambridge, United Kingdom.

4 University of Glasgow-MRC Centre for Virus Research, Glasgow, United Kingdom.

5 Cambridge Institute for Medical Research, University of Cambridge, Cambridge, United Kingdom.

Address correspondence to: Richard J. Stanton, Infection and Immunity, School of Medicine, Henry Wellcome Building, Heath Park, Cardiff CF14 4XN, United Kingdom. Phone: 44.0.7969.148916 Email: [email protected]

Authorship note: VMV and IM contributed equally to this work.

1 Division of Infection and Immunology, School of Medicine, Cardiff University, Cardiff, United Kingdom.

2 Cardiff Metropolitan University, Cardiff, United Kingdom.

3 Department of Medicine, University of Cambridge, Cambridge, United Kingdom.

4 University of Glasgow-MRC Centre for Virus Research, Glasgow, United Kingdom.

5 Cambridge Institute for Medical Research, University of Cambridge, Cambridge, United Kingdom.

Address correspondence to: Richard J. Stanton, Infection and Immunity, School of Medicine, Henry Wellcome Building, Heath Park, Cardiff CF14 4XN, United Kingdom. Phone: 44.0.7969.148916 Email: [email protected]

Authorship note: VMV and IM contributed equally to this work.

Find articles by Aicheler, R. in: JCI | PubMed | Google Scholar

1 Division of Infection and Immunology, School of Medicine, Cardiff University, Cardiff, United Kingdom.

2 Cardiff Metropolitan University, Cardiff, United Kingdom.

3 Department of Medicine, University of Cambridge, Cambridge, United Kingdom.

4 University of Glasgow-MRC Centre for Virus Research, Glasgow, United Kingdom.

5 Cambridge Institute for Medical Research, University of Cambridge, Cambridge, United Kingdom.

Address correspondence to: Richard J. Stanton, Infection and Immunity, School of Medicine, Henry Wellcome Building, Heath Park, Cardiff CF14 4XN, United Kingdom. Phone: 44.0.7969.148916 Email: [email protected]

Authorship note: VMV and IM contributed equally to this work.

1 Division of Infection and Immunology, School of Medicine, Cardiff University, Cardiff, United Kingdom.

2 Cardiff Metropolitan University, Cardiff, United Kingdom.

3 Department of Medicine, University of Cambridge, Cambridge, United Kingdom.

4 University of Glasgow-MRC Centre for Virus Research, Glasgow, United Kingdom.

5 Cambridge Institute for Medical Research, University of Cambridge, Cambridge, United Kingdom.

Address correspondence to: Richard J. Stanton, Infection and Immunity, School of Medicine, Henry Wellcome Building, Heath Park, Cardiff CF14 4XN, United Kingdom. Phone: 44.0.7969.148916 Email: [email protected]

Authorship note: VMV and IM contributed equally to this work.

1 Division of Infection and Immunology, School of Medicine, Cardiff University, Cardiff, United Kingdom.

2 Cardiff Metropolitan University, Cardiff, United Kingdom.

3 Department of Medicine, University of Cambridge, Cambridge, United Kingdom.

4 University of Glasgow-MRC Centre for Virus Research, Glasgow, United Kingdom.

5 Cambridge Institute for Medical Research, University of Cambridge, Cambridge, United Kingdom.

Address correspondence to: Richard J. Stanton, Infection and Immunity, School of Medicine, Henry Wellcome Building, Heath Park, Cardiff CF14 4XN, United Kingdom. Phone: 44.0.7969.148916 Email: [email protected]

Authorship note: VMV and IM contributed equally to this work.

Find articles by Ladell, K. in: JCI | PubMed | Google Scholar | />

1 Division of Infection and Immunology, School of Medicine, Cardiff University, Cardiff, United Kingdom.

2 Cardiff Metropolitan University, Cardiff, United Kingdom.

3 Department of Medicine, University of Cambridge, Cambridge, United Kingdom.

4 University of Glasgow-MRC Centre for Virus Research, Glasgow, United Kingdom.

5 Cambridge Institute for Medical Research, University of Cambridge, Cambridge, United Kingdom.

Address correspondence to: Richard J. Stanton, Infection and Immunity, School of Medicine, Henry Wellcome Building, Heath Park, Cardiff CF14 4XN, United Kingdom. Phone: 44.0.7969.148916 Email: [email protected]

Authorship note: VMV and IM contributed equally to this work.

1 Division of Infection and Immunology, School of Medicine, Cardiff University, Cardiff, United Kingdom.

2 Cardiff Metropolitan University, Cardiff, United Kingdom.

3 Department of Medicine, University of Cambridge, Cambridge, United Kingdom.

4 University of Glasgow-MRC Centre for Virus Research, Glasgow, United Kingdom.

5 Cambridge Institute for Medical Research, University of Cambridge, Cambridge, United Kingdom.

Address correspondence to: Richard J. Stanton, Infection and Immunity, School of Medicine, Henry Wellcome Building, Heath Park, Cardiff CF14 4XN, United Kingdom. Phone: 44.0.7969.148916 Email: [email protected]

Authorship note: VMV and IM contributed equally to this work.

1 Division of Infection and Immunology, School of Medicine, Cardiff University, Cardiff, United Kingdom.

2 Cardiff Metropolitan University, Cardiff, United Kingdom.

3 Department of Medicine, University of Cambridge, Cambridge, United Kingdom.

4 University of Glasgow-MRC Centre for Virus Research, Glasgow, United Kingdom.

5 Cambridge Institute for Medical Research, University of Cambridge, Cambridge, United Kingdom.

Address correspondence to: Richard J. Stanton, Infection and Immunity, School of Medicine, Henry Wellcome Building, Heath Park, Cardiff CF14 4XN, United Kingdom. Phone: 44.0.7969.148916 Email: [email protected]

Authorship note: VMV and IM contributed equally to this work.

Find articles by Davison, A. in: JCI | PubMed | Google Scholar | />

1 Division of Infection and Immunology, School of Medicine, Cardiff University, Cardiff, United Kingdom.

2 Cardiff Metropolitan University, Cardiff, United Kingdom.

3 Department of Medicine, University of Cambridge, Cambridge, United Kingdom.

4 University of Glasgow-MRC Centre for Virus Research, Glasgow, United Kingdom.

5 Cambridge Institute for Medical Research, University of Cambridge, Cambridge, United Kingdom.

Address correspondence to: Richard J. Stanton, Infection and Immunity, School of Medicine, Henry Wellcome Building, Heath Park, Cardiff CF14 4XN, United Kingdom. Phone: 44.0.7969.148916 Email: [email protected]

Authorship note: VMV and IM contributed equally to this work.

Find articles by Wilkinson, G. in: JCI | PubMed | Google Scholar

1 Division of Infection and Immunology, School of Medicine, Cardiff University, Cardiff, United Kingdom.

2 Cardiff Metropolitan University, Cardiff, United Kingdom.

3 Department of Medicine, University of Cambridge, Cambridge, United Kingdom.

4 University of Glasgow-MRC Centre for Virus Research, Glasgow, United Kingdom.

5 Cambridge Institute for Medical Research, University of Cambridge, Cambridge, United Kingdom.

Address correspondence to: Richard J. Stanton, Infection and Immunity, School of Medicine, Henry Wellcome Building, Heath Park, Cardiff CF14 4XN, United Kingdom. Phone: 44.0.7969.148916 Email: [email protected]

Authorship note: VMV and IM contributed equally to this work.

1 Division of Infection and Immunology, School of Medicine, Cardiff University, Cardiff, United Kingdom.

2 Cardiff Metropolitan University, Cardiff, United Kingdom.

3 Department of Medicine, University of Cambridge, Cambridge, United Kingdom.

4 University of Glasgow-MRC Centre for Virus Research, Glasgow, United Kingdom.

5 Cambridge Institute for Medical Research, University of Cambridge, Cambridge, United Kingdom.

Address correspondence to: Richard J. Stanton, Infection and Immunity, School of Medicine, Henry Wellcome Building, Heath Park, Cardiff CF14 4XN, United Kingdom. Phone: 44.0.7969.148916 Email: [email protected]

Authorship note: VMV and IM contributed equally to this work.

1 Division of Infection and Immunology, School of Medicine, Cardiff University, Cardiff, United Kingdom.

2 Cardiff Metropolitan University, Cardiff, United Kingdom.

3 Department of Medicine, University of Cambridge, Cambridge, United Kingdom.

4 University of Glasgow-MRC Centre for Virus Research, Glasgow, United Kingdom.

5 Cambridge Institute for Medical Research, University of Cambridge, Cambridge, United Kingdom.

Address correspondence to: Richard J. Stanton, Infection and Immunity, School of Medicine, Henry Wellcome Building, Heath Park, Cardiff CF14 4XN, United Kingdom. Phone: 44.0.7969.148916 Email: [email protected]

Authorship note: VMV and IM contributed equally to this work.

1 Division of Infection and Immunology, School of Medicine, Cardiff University, Cardiff, United Kingdom.

2 Cardiff Metropolitan University, Cardiff, United Kingdom.

3 Department of Medicine, University of Cambridge, Cambridge, United Kingdom.

4 University of Glasgow-MRC Centre for Virus Research, Glasgow, United Kingdom.

5 Cambridge Institute for Medical Research, University of Cambridge, Cambridge, United Kingdom.

Address correspondence to: Richard J. Stanton, Infection and Immunity, School of Medicine, Henry Wellcome Building, Heath Park, Cardiff CF14 4XN, United Kingdom. Phone: 44.0.7969.148916 Email: [email protected]

Authorship note: VMV and IM contributed equally to this work.

Find articles by Stanton, R. in: JCI | PubMed | Google Scholar

Published February 15, 2021 - More info

Human cytomegalovirus (HCMV) is a ubiquitous pathogen that causes severe disease following congenital infection and in immunocompromised individuals. No vaccines are licensed, and there are limited treatment options. We now show that the addition of anti-HCMV antibodies (Abs) can activate NK cells prior to the production of new virions, through Ab-dependent cellular cytotoxicity (ADCC), overcoming viral immune evasins. Quantitative proteomics defined the most abundant HCMV proteins on the cell surface, and we screened these targets to identify the viral antigens responsible for activating ADCC. Surprisingly, these were not structural glycoproteins instead, the immune evasins US28, RL11, UL5, UL141, and UL16 each individually primed ADCC. We isolated human monoclonal Abs (mAbs) specific for UL16 or UL141 from a seropositive donor and optimized them for ADCC. Cloned Abs targeting a single antigen (UL141) were sufficient to mediate ADCC against HCMV-infected cells, even at low concentrations. Collectively, these findings validated an unbiased methodological approach to the identification of immunodominant viral antigens, providing a pathway toward an immunotherapeutic strategy against HCMV and potentially other pathogens.

Human cytomegalovirus (HCMV) establishes lifelong infection in the face of robust humoral and cell-mediated immune responses. The virus is a significant cause of morbidity and mortality in immunocompromised individuals such as transplant recipients and patients with HIV and following congenital infection. A vaccine against HCMV is considered to be the highest priority, particularly for the prevention of congenital disease ( 1 ), but none has been licensed. The standard for treatment is therefore antiviral agents, however, these are limited by toxicity and the emergence of resistant strains ( 2 ).

As an alternative, antibody (Ab) responses have been investigated as a basis for improved vaccines and immunotherapies ( 3 – 9 ). Several lines of evidence support a protective role for Abs in infection, including observational studies of natural immunity, which have documented a correlation between Ab titers and the prevention of intrauterine transmission ( 10 – 13 ). Moreover, the administration of hyperimmune globulin (HIG) can improve survival in patients undergoing solid organ transplantation ( 14 ), and Ab titers correlated with protection in vaccine trials ( 8 , 9 ). As a result, a range of Abs directed against virion envelope glycoproteins that are capable of neutralizing the entry of cell-free virus have been developed ( 5 , 15 , 16 ). However, neutralizing monoclonal antibodies (mAbs) administered as therapies have had only modest effects and/or failed to meet primary endpoints in clinical trials, namely, a reduction in viremia and/or the need for preemptive therapy ( 17 , 18 ).

One potential explanation for this lack of clinical efficacy lies in the biology of virus dissemination. Spread between individuals involves cell-free virus, which can be efficiently inhibited by neutralizing Abs. In contrast, dissemination within a host likely relies primarily on direct cell-to-cell spread ( 19 – 24 ), which is resistant to neutralizing antibodies ( 25 ), irrespective of the Ab repertoire of the donor ( 26 ). Thus, although classical neutralizing Abs may have a role in preventing transmission between people, they may be less effective in preventing the spread of virus within an individual. This is consistent with clinical trials of a subunit gB vaccine, in which protection correlated with Ab levels, but the induced antibodies did not exhibit overt neutralizing activity ( 27 , 28 ). We therefore sought to prioritize Ab-based immunotherapeutic approaches that could target infected cells directly.

NK cells are crucial for virus control in vivo ( 29 ). This fact is highlighted by the impressive arsenal of HCMV-encoded immune evasins that act in consort to suppress NK cell activation through the manipulation of ligands for activating and inhibitory receptors ( 30 , 31 ). However, in addition to working through these receptors, NK cells participate in Ab-dependent cellular cytotoxicity (ADCC) ( 32 , 33 ). ADCC involves the activation of NK cells upon engagement of Fc receptors (FcRs) on the NK cell surface, with the Fc portion of an Ab bound to a target cell. In vivo, HCMV infection is associated with a dramatic expansion of “adaptive” NK cells marked by the expression of CD94/NKG2C, and CD57 and by the loss of FcεR1γ ( 29 , 34 ). These cells are exceptionally efficient at mediating ADCC ( 35 – 38 ) and have been associated with protection from disease ( 35 , 39 – 41 ). Accordingly, ADCC may be an important mechanism of immune control during natural infection. In this scenario, Abs act as critical stimulators of cellular immunity, rather than acting through virus neutralization.

We were therefore interested in how ADCC operated in the context of an HCMV infection and whether it could be exploited for therapeutic use. We found that anti-HCMV Abs could activate NK cells early after HCMV infection, prior to the production of new virions, and that these Abs had a remarkable capacity to overwhelm the potent HCMV-encoded NK cell evasion mechanisms in vitro. We have previously exploited the power of proteomics to characterize viral and host gene expression during HCMV infection in unparalleled detail, revealing the ways in which the virus manipulates the host cell to promote survival, and to identify ways of counteracting the virus through antiviral restriction factors ( 33 , 42 – 47 ). Here, we combined this technique with functional immunological screening to identify the targets on the infected cell surface that mediate antiviral ADCC. Surprisingly, these techniques revealed that the optimal targets were not the structural glycoproteins that are traditionally assumed to be ADCC targets, but immune evasins that are expressed earlier during the viral life cycle. Their identification enabled us to isolate human mAbs directed against these targets that, once we had genetically engineered them, could activate NK cells in response to HCMV-infected cells. Thus, our technologies enabled the identification of optimal antigenic targets for the development of antiviral therapeutics, and the isolation of what we believe to be the first human mAbs targeting a single HCMV protein that are sufficient to mediate enhanced NK activation through ADCC, despite virus-encoded immune evasins. Our platform is therefore capable of generating novel antiviral immunotherapies that can efficiently activate antiviral cellular immunity.

HCMV-infected cells are susceptible to ADCC during the early phase of infection. We examined the ability of Cytotect (clinical-grade HIG pooled from donors with high anti-HCMV–neutralizing titers) to enhance NK cell activation in the presence of target cells infected with a HCMV strain (Merlin) expressing the complete repertoire of virally encoded immune evasins. Since adaptive NK cells are the primary mediators of ADCC in PBMCs from HCMV-seropositive donors ( 29 , 35 – 38 ), we examined the activation of CD56 + NK cells in the CD57 + and NKG2C + subsets, measuring degranulation via surface mobilization of CD107a. Both cell populations demonstrated a greater enhancement of degranulation when Ab was added, compared with the NKG2C – CD57 – cell population. However, in the majority of donors, we observed a large overlap between the CD57 + and NKG2C + cell populations, and the levels of degranulation were virtually indistinguishable between them. As NKG2C + NK cells are rarely present in uninfected individuals, and up to 4% of people do not harbor the corresponding gene (KLRC2), subsequent data were recorded for CD57 + NK cells.

Cytotect enhanced NK cell activation at a minimum concentration of 12.5 μg/mL and became progressively more potent as concentrations increased to 50 μg/mL, representing a relatively steep activation curve (Figure 1A). Experiments were capped at this maximum, because increased background activation was observed with higher concentrations of IgG Abs from HCMV-seronegative donors. Interestingly, efficacy was not dependent on NK cell stimulation, since equivalent results were obtained whether or not cells were preincubated with IFN-α (Figure 1, A and B). Given that HCMV actively represses the release of IFNs ( 48 ), this supports an important role for ADCC in rapidly activating NK cells against HCMV without a requirement for additional stimulations.

Characterization of ADCC-mediated NK cell activation against HCMV-infected fibroblasts. HFFFs immortalized with hTERT or similarly immortalized autologous SFs were infected with HCMV strain Merlin. Mock-infected HF-TERTs or SFs were included as controls. (A and B) Percentage of degranulation of CD56 + CD57 + NK cells among PBMCs in the presence of HF-TERTs infected for 48 hours with HCMV and different concentrations of either Cytotect or seronegative IgGs (Neg IgG). PBMCs were either untreated (A) or pretreated for 18 hours with IFN-α (B). (C and D) Percentage of degranulation of CD56 + CD57 + NK cells among PBMCs in the presence of HF-TERTs infected for 24 hours, 48 hours, or 72 hours with HCMV and either Cytotect or seronegative IgGs (each at 50 μg/mL). PBMCs were either untreated (C) or pretreated for 18 hours with IFN-α (D). (E and F) Percentage of degranulation of CD56 + CD57 + NKG2C + NK cells among PBMCs in the presence of HF-TERTs (E) or SFs (F) infected for 48 hours with HCMV and either Cytotect or seronegative IgGs (each at 50 μg/mL). Results are representative of at least 3 experiments. All data are shown as the mean ± SD of triplicate samples. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001, by 2-way ANOVA.

When the sensitivity of HCMV-infected cells to ADCC was investigated over the course of infection, we detected NK cell activation as early as 24 hours post infection (hpi), irrespective of preincubation with IFN-α, but this increased dramatically at 48 hpi (Figure 1, C and D) before decreasing slightly at 72 hpi. This reduction may be related to the expression at this later time point of viral FcRs and other NK inhibitors, which antagonize ADCC ( 32 , 45 , 49 ). HCMV antigens expressed on the cell surface by 48 hpi are therefore recognized by naturally occurring Abs and act as effective targets to drive ADCC. Importantly, HCMV has a slow replication cycle, with virions not produced in significant numbers until 72 hpi, so these observations highlighted a therapeutic opportunity to limit the dissemination of HCMV.

HCMV downregulates, but does not abrogate, the expression of endogenous HLA class I molecules. NK cell activation may therefore be influenced by interactions between residual HLA-I and killer immunoglobulin-like receptors (KIRs). To address this possibility, we investigated NK cell recognition of allogeneic and autologous targets in the context of ADCC. The potency of HCMV-encoded NK cell evasion functions is illustrated by the fact that uninfected autologous and allogeneic targets activated NK cells much more efficiently than did the corresponding HCMV-infected targets (Figure 1, E and F). However, in both cases, the inclusion of seropositive Abs overcame the strong protective effects of HCMV-encoded NK evasion functions to stimulate high levels of NK cell activation, irrespective of preincubation with IFN-α (Figure 1, E and F). Thus, the addition of anti-HCMV Abs was able to potently activate NK cells and overcome viral immune evasion prior to the production of new virions, irrespective of NK cell stimulation or engagement of HLA-I.

Antigens expressed on the cell surface at 48 hpi promote ADCC. ADCC has the potential to target infected cells during the early phase of the HCMV replication cycle. To determine which viral antigens primed ADCC, we reanalyzed data from our quantitative temporal viromics investigation of the HCMV-infected cell-surface proteome ( 45 ). We identified 3 clear kinetic classes of protein expression (Figure 2A). Ten proteins reached at least 25% of their maximal cell-surface levels by 24 hpi, and an additional 5 proteins reached at least 25% of their maximal levels by 48 hpi. Thus, a substantial number of viral proteins are trafficked to the cell membrane prior to the production of new virions. Furthermore, multiple proteins reached a maximal overall abundance equal to or higher than that of structural proteins expressed during the later phases of infection (Figure 2B). Therefore, targeting proteins expressed early during the viral life cycle is likely to be equally as effective as targeting later-expressed factors. An analysis of the partitioned abundance of each protein over time indicated that UL16, RL12, UL141, and US28 were expressed on the cell surface at 48 hpi, were among the most abundant viral proteins at this time point, and would therefore be potential candidates for ADCC targets (Figure 2C).

Identification of viral proteins on the plasma membrane that could prime ADCC. (A) Temporal profiles of viral proteins (n = 27) identified previously on the surface of cells infected with HCMV. Proteins were only included in the analysis if detected in experiments PM1 and PM2 and quantified by 2 or more peptides in experiment PM1 or experiment PM2. Data are shown for experiment PM2. Proteins are grouped on the basis of expression kinetics, indicating that greater than 25% of the maximal signal was reached by 24 hours (left), 48 hours (middle), or 72 hours (right). (B) Average total abundance of each surface-expressed viral protein measured using IBAQ. Error bars indicate ranges from experiments PM1 and PM2. (C) Partitioned IBAQ abundance of each surface-expressed viral protein over time. Average IBAQ abundance values in B were multiplied by the fractional abundance at each time point from A. (D) HF-TERTs transfected with the coxsackie-adenovirus receptor (HFFF-hCARs) were transduced with RAds expressing individual viral proteins. An identical vector lacking a transgene was used as a control. Surface-expressed proteins were isolated by aminooxy biotinylation followed by immunoprecipitation with streptavidin beads 48 hours after transduction. Western blots show detection of the C-terminal V5 tags engineered into each protein, with the exception of UL141, which was detected with a UL141-specific Ab. UL141 staining of the gel was performed separately but is overlaid on the same image. (E) Percentage of degranulation of CD56 + CD57 + NK cells among PBMCs in the presence of HFFF-hCARs, transduced as in D, and either Cytotect or seronegative IgGs (each at 50 μg/mL). Results are representative of 3 experiments. Data are shown as the mean ± SD of triplicate samples (E). *P < 0.05 and ****P < 0.0001, by 2-way ANOVA. ctrl, control.

On the basis of these results, we generated replication-deficient adenovirus (RAd) vectors expressing each of the 15 viral proteins that were reproducibly identified on the surface of HCMV-infected cells by 48 hpi (Figure 2D). Each RAd was then tested individually for its capacity to promote ADCC in the presence of pooled polyclonal HIG (Figure 2E). UL16, UL141, US28, RL11, and UL5 each induced a significant increase in NK cell activation that was dependent on the presence of Cytotect, indicating that these viral antigens could induce early-phase ADCC.

Abs directing ADCC can be isolated from human donors. To investigate whether the identified viral protein targets could mediate ADCC in the context of HCMV infection, we generated a series of mAbs. RL11 is an Fc-binding protein ( 50 ) that complicates both the production of specific Abs and the analysis of functional assays. US28 is a type 3 transmembrane protein, and thus the generation of US28-specific Abs would be less straightforward. Therefore, RL11 and US28 may not provide routine target antigens. Further, since UL5 was associated with only modest levels of NK cell activation, the type 1 membrane proteins UL16 and UL141 were prioritized. Sequences encoding the extracellular domains of each protein were cloned as modified constructs with a C-terminal 6xHis-tag (UL16) or a C-terminal Strep-tag (UL141) into separate RAd vectors for expression. The corresponding proteins were purified from cell supernatants via affinity chromatography, labeled with fluorochromes, and used as probes to stain IgG + B cells from a donor infected with HCMV. UL141-specific B cells were more numerous than UL16-specific B cells (Figure 3A). Single antigen-specific B cells were then flow-sorted into culture medium containing CD40L + feeders, IL-2, IL-4, IL-21, and B cell activating factor (BAFF) to generate plasma cells ( 51 ). All secreted mAbs were then screened against cells expressing UL16 or UL141. Both proteins contain an ER retention signal in the C-terminal cytoplasmic domain, which restricted cell-surface expression (Supplemental Figure 1A supplemental material available online with this article https://doi.org/10.1172/JCI139296DS1). To increase the sensitivity of this flow cytometry–based Ab screen, we increased the cell-surface abundance of target antigens by deleting this region (Supplemental Figure 1A). Screening 60 B cell supernatants against these proteins revealed that 9 bound UL141 and 5 bound UL16 (Supplemental Figure 1B).

Anti-UL16 and anti-UL141 mAbs can be isolated and cloned from seropositive donors. (A) IgG + B cells from a HCMV-seropositive donor were stained with fluorescently labeled UL16 or UL141 proteins to sort B cells expressing specific mAbs. FSC, forward scatter SSC, side scatter. (B and C) HFFF-hCARs were transduced with RAds expressing UL141 or UL16 lacking their ER retention signals. Cells were stained with the cloned human anti-UL141 or anti-UL16 mAbs and analyzed by flow cytometry. Cytotect was used as a positive control. (D) HFFF-hCARs were transduced with RAds lacking a transgene, or RAds expressing wild-type forms of UL141 or UL16. Samples were lysed, separated by SDS-PAGE, and analyzed by immunoblotting using human anti-UL16 or anti-UL141 mAbs. As a positive control, the UL16 lysate was stained with an anti-V5 Ab, and the UL141 lysate was stained with a murine anti-UL141 Ab. (E and F) HFFF-hCARs were transduced with RAds expressing wild-type forms of UL141 or UL16. Forty-eight hours later, they were stained with human anti-UL141 or anti-UL16 mAbs or Cytotect and then analyzed by flow cytometry.

B cell receptor (BCR) sequencing revealed that the predicted amino acid sequences of these mAbs were diverse and incorporated both κ and λ light chains, suggesting that Abs had the potential to target distinct epitopes (Supplemental Figure 2). We subcloned the variable domains of these BCRs into an expression plasmid that provided a human IgG1 backbone, with the specific purpose of optimizing the utility of the Ab fusion for ADCC. When expressed, these recombinant human mAbs retained their capacity to bind to UL141 and UL16 on the cell surface (Figure 3, B and C), but not to denatured antigen (Figure 3D), suggesting that all bind to conformational epitopes.

Anti-UL16 and anti-UL141 human mAbs activate ADCC when antigen is expressed in isolation. Although the mAbs bound to UL16 and UL141 when optimized for high expression on the cell surface (Figure 3, B and C), binding to the natural forms was not detectable by flow cytometry (Figure 3, E and F, Supplemental Figure 1A), indicating that very low levels of these proteins naturally traffic to the cell surface. Nevertheless, ADCC assays appeared more sensitive than flow cytometry, as the natural versions of both genes were able to induce ADCC with both Cytotect and mAbs (Figure 4, A and B).

Human anti-UL16 and anti-UL141 mAbs activate ADCC efficiently against adenovirally expressed UL16 and UL141. (AD) HFFF-hCARs were transduced with RAds expressing wild-type UL16 or UL141. An identical vector lacking a transgene was used as a control. (A) Percentage of degranulation of CD56 + CD57 + NK cells among PBMCs in the presence of transduced HFFF-hCARs and Cytotect (40 μg/mL), seronegative IgGs (40 μg/mL), or UL16-specific mAbs (each at 30 μg/mL). All 4 mAbs were included at equimolar concentrations in the mixture. (B) As in A for UL141. Five mAbs were included at equimolar concentrations in 1 mixture (B2, D3, G3, G4, and G11), and 8 mAbs were included at equimolar concentrations in another mixture (B2, C3, D3, E5, G2, G3, G4, and G11). (C) Percentage of degranulation of CD56 + CD57 + NK cells among PBMCs in the presence of transduced HFFF-hCARs and different concentrations of the tetravalent UL16-specific mAb mixture. (D) As in C for the pentavalent UL141-specific mAb mixture. (E and F) HF-TERTs were infected with HCMV strain Merlin. Mock-infected HF-TERTs were included as controls. (E) Percentage of degranulation of CD56 + CD57 + NK cells among PBMCs in the presence of infected HF-TERTs and Cytotect, seronegative IgGs, or the UL16-specific mAb mixture (each at 30 μg/mL). (F) As in E for UL141. Results are representative of at least 3 experiments. Data are shown as the mean ± SD of triplicate samples (AF). All experiments were performed 48 hours after transduction (AD) or infection (E and F). *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001, by 2-way ANOVA.

Each novel UL16 mAb was readily able to drive ADCC against fibroblasts expressing wild-type UL16 with an efficiency comparable to that observed with Cytotect (Figure 4A). The level of ADCC elicited by different anti-UL16 mAbs was remarkably similar, despite the diversity of their antigen binding (Fab) sequences. When the 5 mAbs were mixed together at equimolar concentrations, the ADCC effect was not enhanced beyond the level of each individual Ab. These findings suggested that each mAb targeted the same immunodominant epitope with similar efficiency, irrespective of diversity in the corresponding antigen-binding domains.

In contrast, only 2 of the UL141-specific mAbs were capable of mediating ADCC in isolation, and activation was extremely weak (Figure 4B). However, when all 8 purified Abs were mixed together at equal concentrations, ADCC was efficiently activated. Three of the Abs were prone to eliciting nonspecific activation against control infected cells, and therefore we tested a mixture of the other 5 Abs and found them to be equally capable of activating ADCC, but with reduced background levels (Figure 4B). The fact that anti-UL141 mAbs stimulated higher levels of degranulation when used as a mixture suggests that at least some of them bind to different epitopes on UL141. In dose-titration experiments against the corresponding targets, mixtures of UL16-specific or UL141-specific mAbs maximally activated NK cells at concentrations above 15 μg/mL (Figure 4, C and D), indicating greater efficacy compared with Cytotect (Figure 1, A and B).

Although these results were encouraging in terms of therapeutic development, pooled mAbs specific for UL16 or UL141 were unable to activate NK cells in the presence of targets infected with HCMV, even though Cytotect was effective (Figure 4, E and F). HCMV encodes 4 Fc-binding proteins (FcRs) (RL11, RL12, RL13, and UL119) that have the potential to antagonize ADCC. Accordingly, human IgGs bound cells infected with an HCMV-mutant strain lacking all 4 of these genes (HCMVΔFc) to a lesser extent than they bound cells infected with wild-type HCMV (Supplemental Figure 3A). However, NK cells were activated similarly under both conditions in the presence of Cytotect (Supplemental Figure 3B). The lack of efficacy of the specific Abs against HCMV-infected cells was therefore not caused by antagonism of ADCC by viral FcRs. It may reflect lower levels of protein on the cell surface during HCMV infection compared with RAd expression (Supplemental Figure 3C), or the concerted action of multiple virally encoded immune evasins that inhibit NK activation ( 30 ).

Ab engineering enables mAbs to activate ADCC against HCMV. A major advantage of cloned mAbs is that they can be manipulated to enhance different effector functions. We took advantage of this to optimize the ability of our mAbs to activate ADCC by introducing 2 amino acid sequence changes into the Fc region that had previously been shown to enhance binding to CD16 on NK cells ( 52 ). In line with previous data indicating that viral and host FcRs bind Fc in different ways ( 53 ), these modifications did not affect binding to viral FcRs (Supplemental Figure 3, D and E). Dose-titration experiments revealed that mixtures of engineered mAbs specific for UL16 or UL141 activated NK cells more potently and at much lower concentrations than did the corresponding unmodified mAbs (Figure 5, A and B) or Cytotect (Figure 5, C and D). As before, when tested separately, all of the mAbs against UL16 activated ADCC, and we observed no increase in activation when they were combined (Figure 5E). However, unlike the unmodified versions, all the modified UL141 mAbs activated ADCC individually (Figure 5F). Moreover, they retained the ability to show enhanced activation when used in combination, whether as a mixture of 5 or 8 mAbs (Figure 5F).

Optimized anti-UL16 and anti-UL141 mAbs activate ADCC efficiently against adenovirally expressed UL16 and UL141. HFFF-hCARs were transduced with RAds expressing wild-type UL16 or UL141. An identical vector lacking a transgene was used as a control. (A) Percentage of degranulation of CD56 + CD57 + NK cells among PBMCs in the presence of transduced HFFF-hCARs and different concentrations of native or Fc-engineered (modified) UL16-specific mAbs (tetravalent mixes). (B) As in A for UL141 (pentavalent mixes). (C) Percentage of degranulation of CD56 + CD57 + NK cells among PBMCs in the presence of transduced HFFF-hCARs and Cytotect, seronegative IgGs, or tetravalent mixes of native or Fc-engineered (modified) UL16-specific mAbs (native Abs each at 30 μg/mL Fc-engineered [modified] mAbs each at 1 μg/mL). (D) As in C for UL141 (pentavalent mixes). (E) As in C for individual Fc-engineered (modified) UL16-specific mAbs. (F) As in D for individual Fc-engineered (modified) UL141-specific mAbs. Results are representative of at least 3 experiments. All data are shown as the mean ± SD of triplicate samples. All experiments were performed 48 hours after transduction. ***P < 0.001 and ****P < 0.0001, by 2-way ANOVA. Mod, modified.

Next, we tested the efficiency of the mAbs in the context of HCMV infection both separately and in combination. Even in their modified form, the anti-UL16 mAbs were not able to reproducibly activate ADCC against HCMV-infected cells (Figure 6, A–C). In contrast, ADCC was efficiently achieved against HCMV using the modified anti-UL141 mAbs. Individually, we found that these mAbs only activated ADCC very weakly, but the combination of 5 Abs was successful at activating ADCC almost as effectively as Cytotect, despite being used at a 40-fold lower concentration (Figure 6, D and E). This effect was highly specific, because activation was not apparent when a virus lacking the cognate antigen was used (Figure 6F). Furthermore, these Abs were also capable of activating NK cells to secrete TNF-α and IFN-γ, indicating potent antiviral effector functions in the presence of targets infected with HCMV (Figure 6, G and H).

Anti-UL141–optimized Abs activate ADCC efficiently against HCMV. HF-TERTs were infected with HCMV strain Merlin (AH) or Merlin ΔUL16 ΔUL141 (C and F). Mock-infected HF-TERTs were included as controls. (A) Percentage of degranulation of CD56 + CD57 + NK cells among PBMCs in the presence of infected HF-TERTs and different concentrations of Fc-engineered (modified) UL16-specific mAbs (tetravalent mixture). (B) Percentage of degranulation of CD56 + CD57 + NK cells among PBMCs in the presence of infected HF-TERTs and Cytotect (40 μg/mL), seronegative IgGs (40 μg/mL), or Fc-engineered (modified) UL16-specific mAbs tested individually or in combination (each at 1 μg/mL). (C) Percentage of degranulation of CD56 + CD57 + NK cells among PBMCs in the presence of infected HF-TERTs and Cytotect (40 μg/mL), seronegative IgGs (40 μg/mL), or the tetravalent mixture of Fc-engineered (modified) UL16-specific mAbs (each at 1 μg/mL). Activity was tested against HF-TERTs infected with Merlin or Merlin ΔUL16 ΔUL141. (D) As in A for UL141 (pentavalent mixture). (E) As in B for UL141. (F) As in C for UL141. (G) Percentage of intracellular TNF-α production by CD56 + CD57 + NK cells among PBMCs in the presence of infected HF-TERTs and Cytotect (50 μg/mL), seronegative IgGs (50 μg/mL), or Fc-engineered (modified) UL141-specific mAbs tested individually or in combination (each at 1 μg/mL). (H) As in G for IFN-γ. Results are representative of at least 3 experiments. Data are shown as the mean ± SD of triplicate samples (AH). Experiments were performed 48 hours after infection (AF). *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001, by 2-way ANOVA.

Finally, we examined the ability of our mAbs to promote direct killing of cells. Measuring short-term cytotoxicity using chromium-release assays revealed that a mixture of 5 modified anti-UL141 Abs led to a substantial increase in NK-mediated cell death when UL141 was expressed in isolation (Figure 7A), or when fibroblasts were infected with HCMV (Figure 7B). This effect was not restricted by cell type, because we obtained similar results when HCMV infected epithelial cells were used (Figure 7C). Furthermore, our defined Abs markedly outperformed Cytotect in these assays, despite being used at a lower concentration. Interestingly, unlike in degranulation assays (Supplemental Figure 3B), when we performed cytotoxicity experiments, the viral FcRs did limit cell death, since killing was significantly enhanced in their absence (Figure 7B). However, this effect was more pronounced with Cytotect than with our engineered mAbs. Thus, Ab engineering to enhance NK cell activation may also improve function by overcoming viral countermeasures. We also investigated the ability of the UL141 mAbs to promote the control of virus using a recently developed 10-day viral dissemination assay (VDA), which captures the effects of both cytotoxic and noncytotoxic virus control in a fully autologous system (Figure 7, D and E, and refs. 54 , 55 ). The UL141 mAbs demonstrated a striking ability to enhance NK-mediated virus control in this assay, confirming that they can act as powerful effectors for long-term control of virus infection, even at low effector/target (E/T) ratios.

Anti-UL141–optimized Abs mediate efficient killing of HCMV-infected cells. (AC) 51 Cr release into the supernatant was used as a measure of the ability of NK cells to kill target cells. Targets were mixed with ex vivo–purified NK cells as effectors at a an E/T ratio of 20:1, and 51 Cr release was measured 4 hours later. Seronegative IgG (50 μg/mL), Cytotect (50 μg/mL), or a mixture of 5 Fc-engineered (modified) UL141-specific mAbs were included as indicated. Targets were HF-CARs infected with RAd vectors expressing UL141 (RAd-UL141), or lacking a transgene (RAd Ctrl) (A) HFFF mock infected or infected with wild-type HCMV (HCMV) or HCMV lacking the viral FcRs (ΔFc) (B) or ARPE19 mock infected or infected with wild-type HCMV (C). For ARPE19 infection, cells were infected by coculturing with purified fibroblasts for 24 hours and then sorted to purity. All experiments were performed 48 hours after infection. (D and E) HCMV expressing mCherry linked to an immediate early gene (UL36), and EGFP linked to a late gene (UL32) were used to infect SFs at a low MOI. Autologous NK cells were then added alone or together with a control mAb or the mixture of 5 modified anti-UL141 mAbs (each at 1 μg/mL). Eight to 10 days later, the percentage of infected cells demonstrating expression of immediate early (D) or late (E) viral proteins were measured by flow cytometry for mCherry or EGFP, respectively, and normalized to the percentage of infected cells in the absence of NK cells. Results are representative of at least 2 experiments. Data are shown as the mean ± SD of triplicate samples. **P < 0.01, ***P < 0.001, and ****P < 0.0001, by 2-way ANOVA.

Multiple human anti-HCMV mAbs have been developed that target virus neutralization as their mechanism of action ( 5 , 17 , 18 , 56 – 58 ). Although these mAbs offer advantages over HIG, in that they are defined products with a specific activity, the highly cell-associated nature of clinical HCMV strains and the intrinsically greater resistance to neutralization of cell-to-cell spread in comparison with cell-free entry mean that their ability to prevent intra-host spread may be limited ( 25 , 26 ). In contrast, Ab-mediated activation of cellular immunity does not suffer from these limitations and has been implicated in the control of multiple different viruses, including West Nile virus, smallpox virus, herpes simplex virus, influenza virus, yellow fever virus, Ebola virus, and Epstein-Barr virus. It also correlates with control of HIV in both vaccination and natural infection ( 59 , 60 ) and is thought to underlie the efficacy of numerous antitumour Abs in clinical development ( 61 , 62 ). There is thus considerable interest in exploiting this powerful mechanism of control across multiple pathogens and diseases. However, this requires mapping of the antigens that optimally activate ADCC and production of cloned human mAbs capable of mediating ADCC. Our demonstration that plasma membrane proteomics and functional immunology can be combined to identify novel ADCC targets not only opens up a fuller understanding of natural immunity against HCMV that can now be exploited for therapeutic benefit, but is also applicable to exploiting Ab-mediated activation of cellular immunity in other infectious diseases, and potentially even cancer.

As a virus that persists lifelong, HCMV faces major challenges in avoiding being cleared by the immune response and, as a result, has evolved an exceptionally broad range of techniques to limit immune activation ( 30 , 31 ). The study of these has revealed details about the underlying functioning of the immune system, but also shows that the virus poses a particular challenge to the development of methods to activate antiviral immunity. It is therefore all the more impressive that our technologies enabled the development of Abs capable of reversing the ability of viral immune evasins to inhibit NK cell activation, even when the HCMV strain expressed the complete repertoire of genes present in a clinical isolate ( 19 , 20 , 33 ). In addition to encoding functioning immune evasins, it seems likely that HCMV has evolved to restrict cell-surface expression of viral proteins in order to minimize ADCC. As a result, the extreme sensitivity of mass spectrometry was required in order to identify viral cell-surface antigens. Nevertheless, although cell-surface antigen levels were extremely low, it is clear that ADCC had evolved to be extraordinarily sensitive, with Ab engineering enabling strong NK activation to occur despite Ab binding being undetectable by flow cytometry, underscoring the potential of our pipeline to produce highly effective Abs. The strict species specificity of CMVs and the fact that our primary targets (UL16 and UL141) are not conserved in mouse or rat CMV, and show only 32% homology in rhesus CMV, preclude efficacy testing of our Abs in animal models. Future work will be required to demonstrate both safety and efficacy in humans.

The choice of cell-surface antigen is likely to be an important parameter that defines the efficacy of mAbs that activate ADCC. Surprisingly, the antigens that we identified as mediating ADCC were not the classical viral structural proteins that ADCC studies have traditionally focused on. Our previous proteomics analysis defined 5 temporal classes of viral gene expression ( 45 ), with examples from multiple classes found on the infected cell surface. However, targeting those present 48 hpi offers a number of advantages. ADCC activity with polyclonal IgG from seropositive donors was as high at this time point as it was later in infection, implying that many of the antigens that prime ADCC-mediated control in healthy individuals are present within 48 hours. New virions have not yet formed, increasing the chances that cells will be killed before the virus can spread, and the abundance of the proteins we targeted was among the highest of any viral protein, at any time point. In addition, by focusing on nonstructural proteins, there was limited risk of inadvertently enhancing disease through Ab-dependent enhancement (ADE) of infection ( 63 ). Although we prioritized UL16 and UL141, US28 or RL11 may also be useful targets if suitable Abs can be generated, although at present this is not simple. Abs targeting US28 in particular could be important, since US28 is expressed during latency and there is evidence that polyclonal Abs targeting this protein can lead to the destruction of latently infected monocytes via neutrophil-mediated ADCC ( 64 ). Finally, our target antigens were chosen on the basis of their ability to activate ADCC with HIG. Some of the other cell-surface proteins that we identified may also mediate ADCC effectively, but if they do not induce high Ab levels during natural infection, they would have remained silent in our functional assays. For these proteins, murine immunization strategies could be used to generate additional ADCC-capable mAbs. Likewise, it is possible that some potential targets were missed by our mass spectrometry strategy if the peptides they generated ionized poorly.

It is notable that all of the targets identified in the present study are immune evasion genes. Among its many roles, US28 acts as a cytokine sink on the cell surface ( 65 ). UL141 reduces the cell-surface expression levels of CD112 and CD155 ( 66 , 67 ), which are ligands for the activating NK cell receptor DNAM1, as well as TRAIL receptors ( 68 ), while UL16 reduces cell-surface levels of ULBP1–3 and MICB, which bind to the activating NK cell receptor NKG2D ( 69 , 70 ). It may be that both UL141 and UL16 traffic to the cell surface to scavenge their targets. Accordingly, if viral mutants arose in vivo to evade Ab recognition, infected cells might become more susceptible to NK cell–mediated immune control, which in turn would hinder the widespread selection of such mutants. The use of multiple Abs targeting the same antigen could also limit the selection of viral escape mutations. The sequences of both UL141 and UL16 are well conserved among clinical HCMV isolates, suggesting that Abs targeting them could control a broad range of virus strains ( 71 , 72 ).

Cloned mAbs offer major advantages over polyclonal products such as HIG. They are defined products with consistent specificity over time, and molecular engineering can be used to optimize functionality for specific purposes. As a result, our mAbs activated ADCC at concentrations over 40-fold lower than that of Cytotect, something that may significantly enhance effectiveness in vivo ( 73 ). Furthermore, the generation of anticancer immunotherapies has resulted in the development of multiple different Ab optimizations, which now become amenable to deployment against HCMV. This includes “arming” Abs with drugs or toxins, or converting them into bispecific or trispecific NK engagers to enhance ADCC efficacy even further ( 74 ). In addition to ADCC, surface-bound Abs can also activate phagocytosis, complement, and T cells ( 75 ) and can lead to an adaptive cellular response by binding to FcRs on DCs. The induction of such mechanisms, in addition to ADCC, has been shown to be effective at mediating tumour control ( 61 , 62 ), and modifications exist to further optimize these activities ( 76 ). Thus, the development of our mAbs provides a platform with which multiple aspects of the immune system can be armed, increasing efficacy in vivo even further. As well as opening up the possibility of exploiting optimized Abs for passive infusion, the cell-surface targets that we have identified could also be considered as part of a vaccine strategy. For example, by vaccinating with UL141 protein, it may be possible to generate a polyclonal anti-UL141 Ab response, which could provide enhanced immunity via Fc-mediated effector functions. In this context, it will be important to determine the efficacy of ADCC Abs in controlling HCMV infection in individuals exhibiting different repertoires of NK cell subsets, including in those who are HCMV seropositive or seronegative and in individuals with larger or smaller numbers of adaptive NK cells.

In conclusion, we have developed a methodological pipeline combining proteomics with functional immunology, single-cell cloning, and molecular engineering that identified novel therapeutic targets revealed that “classical” cell-surface antigens were not necessarily the optimal targets avoided potential issues with ADE and produced Abs capable of binding targets and activating cellular immunity, despite the presence of multiple immune evasins and despite the fact that target expression levels can be too low to detect by flow cytometry. We anticipate that our approach will be generically applicable to other pathogens and tumors, both in terms of passive immunization and vaccine design, with broad implications for immunotherapeutic strategies beyond HCMV. However, here we used it to demonstrate that ADCC is an extraordinarily potent effector mechanism for activating NK cells against HCMV-infected cells. We have identified multiple cell-surface targets for the development of novel antiviral immunotherapies or vaccination strategies that can activate ADCC, and we have generated what we believe to be the first human Abs targeting a single HCMV antigen that are sufficient to activate ADCC. Together, we believe these results open the path for the development of novel immunotherapeutic strategies that can activate multiple different arms of cellular immunity and enable enhanced control of HCMV in vivo.

Cells. Human fetal foreskin fibroblasts (HFFFs), HFFFs immortalized with human telomerase reverse transcriptase (HF-TERTs) ( 77 ), HF-TERTs transfected with the coxsackie adenovirus receptor (HFFF-hCARs) ( 78 ), TERT-immortalized healthy donor skin fibroblasts (SFs), and 293 TREX cells (Thermo Fisher Scientific) were grown under standard conditions in DMEM (Thermo Fisher Scientific) supplemented with 10% FCS, penicillin (100 U/m), and streptomycin (100 μg/mL). Expi293F suspension cells (Thermo Fisher Scientific) were maintained in a humidified, shaking incubator at 150 rpm, 37°C, and 8% CO2 and were grown in Gibco Expi293 Expression Medium (Thermo Fisher Scientific). Ms40L low cells were a gift from Garnett Kelsoe (Duke University, Durham, North Carolina, USA) and David Baltimore (Caltech, Pasadena, California, USA) ( 79 , 80 ). They were kept in DMEM supplemented as above with the addition of 50 μM β-mercaptoethanol.

Viruses. All viruses were derived from a bacterial artificial chromosome (BAC) containing the complete wild-type HCMV genome, with the exception of RL13 and UL128, since the absence of these genes enhances stability in fibroblasts ( 20 , 81 ). Mutations were engineered using either recombineering or en passant mutagenesis, as described previously ( 20 , 82 – 85 ). The primer sequences are listed in Table 1. Viruses were generated by transfection of BACs ( 20 ) into HF-TERTs and titrated on HFFFs. All modifications were sequence verified prior to BAC transfection, and all viruses were sequenced at the whole-genome level following reconstitution to exclude the occurrence of second-site mutations ( 86 ).

Primer sequences used in this study

RAds were generated as described previously ( 84 ). They were as follows: RAd-Ctrl (no exogenous protein-coding region) RAd-UL141ΔER (expressing UL141 carrying a deletion of the cytoplasmic tail and an exogenous signal peptide containing an HA tag after the cleavage site) RAd-UL16ΔER (expressing UL16 carrying a deletion of the cytoplasmic tail and an exogenous signal peptide containing an HA tag after the cleavage site) RAd-sUL141 (expressing the UL141 extracellular domain with a C-terminal Strep-tag) RAd-sUL16 (expressing the UL16 extracellular domain with a C-terminal 6xHis-tag) RAd-UL141 (expressing the native form of UL141 ref. 67 ) and RAd-UL16 (expressing the native form of UL16). RAds expressing other HCMV proteins have been described previously ( 84 ), and all contained a C-terminal V5 epitope tag. All RAds were propagated by transfection of the relevant plasmids into 293 TREX cells as described previously ( 84 ).

Proteomics. Data originally published by Weekes et al. ( 45 ) were reanalyzed to estimate the absolute abundance of each cell-surface viral protein. To be included in this analysis, proteins required quantitation, in both experiments PM1 and PM2, of 2 or more peptides in at least 1 of the 2 experiments. Overall, this included 27 of 29 of the viral proteins we originally measured. Experiment PM1 examined cells infected with strain Merlin in biological duplicates at 0 hours, 24 hours, 48 hours, and 72 hours. Reanalysis was based on the mean values for each time point. Experiment PM2 examined cells infected with the same HCMV strain in single replicates at 0 hours, 6 hours, 12 hours, 18 hours, 24 hours, 48 hours, 72 hours, and 96 hours. In reanalysis, the mean values for time point 0 were used, and infection with irradiated HCMV at 12 hours was excluded from analysis. In Figure 2A, for experiment PM2 data, the proteins were grouped according to when greater than 25% of the maximum signal was reached. Abundance for each protein was normalized to a maximum of 1, as described previously ( 45 ). For Figure 2B, the method of intensity-based absolute quantification (IBAQ) was adapted from the original description ( 87 ) to estimate the relative abundance of each of the 27 viral proteins. The maximum MS1 precursor intensity for each quantified peptide was determined, and a summed MS1 precursor intensity for each protein across all matching peptides was calculated, considering data for experiments PM1 and PM2 separately. Intensities were divided by the number of theoretical tryptic peptides from each protein between 7 and 30 amino acid residues in length to give estimated IBAQ values. For each of experiments PM1 and PM2, the estimated IBAQ values were divided by the sum of all values to give the normalized IBAQ values. The average and range of the normalized IBAQ values for each protein are shown in Figure 2, B and C. To determine the proportion of the average normalized IBAQ values that arose at each time point of infection, the IBAQ values were adjusted in proportion to the normalized tandem mass tag (TMT) values shown in Figure 2A.

Protein purification and labeling. Soluble UL141 and UL16 were produced in HFFF-hCARs transduced with RAd-sUL141 or RAd-sUL16, respectively, over a 10-day period at a MOI of 40 PFU/cell. Supernatants were collected and purified using Strep-Tactin (IBA GmbH) or HisTrap HP Columns (GE Healthcare). Both proteins were subjected to buffer exchange in PBS and fluorescently labeled using the Alexa Fluor 647 Protein Labeling Kit (Thermo Fisher Scientific).

Ab isolation. PBMCs were isolated from a healthy HCMV-seropositive donor, and IgG + memory B cells were isolated using an IgG + Memory B Cell Isolation Kit (Miltenyi Biotec). The enriched B cells were stained for 30 minutes at 4 o C with 2 μg/mL Alexa Fluor 647–labeled protein (soluble UL141 or UL16) and flow sorted using a BD FACSAria III (BD Biosciences). Single cells were sorted into individual wells containing Ms40L low feeder cells, 10% FCS, 5% human AB serum, IL-4 (10 ng/mL), BAFF (10 ng/mL), IL-21 (10 ng/mL), and IL-2 (50 ng/mL) in a final volume of 100 μL (all cytokines were from Peprotech). Cultures were supplemented with an additional 100 μL of the same medium 1 week later. Two weeks after coculturing, 50 μL supernatant from each of the single-cell colonies was screened by flow cytometry for binding to UL141 (RAd-UL141ΔER) and UL16 (RAd-UL16ΔER). RNA was extracted from the cells that were positive for binding using the RNEasy Plus Kit (QIAGEN). The Ab sequence was determined by nested reverse transcription PCR (RT-PCR) as described previously ( 88 ). Sequences were analyzed by the IgBLAST tool to identify the V and J composition of the heavy and light chains, then PCR amplified using specific primers and cloned separately into an expression plasmid containing a human IgG1 constant domain, provided by Patrick Wilson (University of Chicago, Chicago, Illinois, USA) ( 88 ).

Ab engineering. S239D and I332E modifications were introduced into the Fc region of each mAb by Gibson assembly ( 52 ). The 2 fragments of the plasmid, containing overlapping regions with the desired modifications, were generated using the following primer sequences: 5′-GGGGGACCGGACGTCTTCCTCTTCCCCCCA-3′ and 5′-GGTTTTCTCCTCGGGGGCTGGGAGGG-3′, or 5′-AGGAAGACGTCCGGTCCCCCCAGGAG-3′ and 5′-CAGCCCCCGAGGAGAAAACCATCTCCAAAGCCA-3′. The resulting fragments were assembled using the NEBuilder HiFi DNA Assembly Cloning Kit (New England Biolabs).

Ab production and purification. Expi293F suspension cells were pelleted, resuspended at 20 × 10 6 cells/mL, and transfected with the relevant light and heavy chain plasmids at a ratio of 70:30 (1.25 μg/10 6 cells of total plasmid DNA) using polyethylenimine (PEI) diluted in ultrapure water (3.75 μg/10 6 cells) and 0.1% Pluronic F-68 ( 89 ). Transfected cells were cultured for 3 hours and subsequently diluted to 10 6 cells/mL with Expi293 Expression Medium containing forskolin (10 μM). Ab-containing supernatants were collected 7 days after transfection.

Both mAbs and Abs from the serum of seronegative donors were purified as described previously ( 88 ). Briefly, supernatants were filtered through a 0.45 μm syringe filter and incubated overnight at 4˚C with protein G agarose beads. The following day, the bead-supernatant reactions were transferred to room temperature for 2 hours and then centrifuged at 3000g for 10 minutes. The beads were transferred to a chromatography column, washed with 5 resin-bed volumes of 1 M NaCl, and eluted twice with 2.5 resin-bed volumes of PBS. Abs were eluted into Tris-HCl, pH 9.0, with 2.5 resin-bed volumes of glycine buffer, pH 2.8 (Pierce, Thermo Fisher Scientific), ensuring that the final pH was approximately 7.0. The Abs were subsequently subjected to buffer exchange against PBS.

CD107a Assays. Degranulation assays were based on the flow cytometric detection of CD107a ( 90 ). PBMCs were rested overnight in RPMI supplemented with 10% FCS, penicillin (100 U/mL), streptomycin (100 μg/mL), and l -glutamine (2 mM) in the absence or presence of IFN-α (1000 U/mL). HF-TERTs (allogeneic) or SFs (autologous) were plated in DMEM without FCS and infected the following day with HCMV (MOI = 5 PFU/cell). The medium was replaced 24 hpi with DMEM containing 10% FCS. Assays were performed 48 hpi unless stated otherwise. Targets were harvested using TrypLE Express (Gibco, Thermo Fisher Scientific), preincubated for 30 minutes with the relevant Ab preparations, and mixed with PBMCs at an E/T ratio of 10:1 in the presence of GolgiStop (0.7 μL/mL, eBioscience) and anti–CD107a–PerCP-Cy5.5 (clone H4A3, BioLegend). Assays were performed in triplicate in U-bottomed, 96-well plates at a final volume of 200 μL/well. Background activation was determined in wells containing effectors without targets. Cells were incubated for 5 hours, washed in cold PBS, and stained with LIVE/DEAD Fixable Aqua (Thermo Fisher Scientific), anti–CD3-BV711 (clone UCHT1, BioLegend), anti–CD56-BV605 (clone 5.1H11, BioLegend), anti–CD57-APC (clone HNK-1, BioLegend), and anti–NKG2C-PE (clone 134591, R&D Systems). In some experiments, cells were also fixed and permeabilized using Cytofix/Cytoperm (BD Biosciences) and stained with anti–TNF-α–BV421 (clone MAb11, BioLegend) and anti–IFN-γ–PE–Cy7 (clone B27, BioLegend). Data were acquired using an Attune NxT Flow Cytometer (Thermo Fisher Scientific) and analyzed with Attune NxT software or FlowJo software, version 10 (Tree Star). All assays were repeated with samples from multiple donors. When used directly ex vivo, NK cells from different donors can vary significantly in the magnitude of their responses, thus, only experiments where results showed consistent patterns between donors are included. Donors included both HCMV-seropositive and -seronegative individuals.

Chromium release cytotoxicity assays. Assays were performed as previously described ( 91 ). In brief, targets were incubated with 150 μCi sodium chromate ( 51 Cr) for 1 hour, washed and allowed to leach for 1 hour, and then incubated with purified NK cells and Abs. After 4 hours, supernatants were removed and mixed with scintillation fluid (Optiphase HiSafe 3, PerkinElmer), before reading the cpm in a MicroBeta 2 (PerkinElmer). Maximum lysis was generated using 2.5% Triton X-100. Specific lysis was calculated as follows: (sample cpm – spontaneous cpm)/(maximum cpm – spontaneous cpm).

Viral dissemination assays. Assays were performed as previously described ( 54 ). Briefly, SFs were infected at a MOI of 0.05 with a virus containing a P2A-mCherry cassette after UL36, and an EGFP tag directly fused to UL32. At 24 hpi, purified ex vivo (NK Isolation Kit, Miltenyi Biotec) autologous NK cells were added at a range of E/T ratios, in the presence or absence of Abs. After 8–10 days, nonadherent cells were washed off and discarded, and adherent cells were trypsinized, fixed in 4% PFA and analyzed by flow cytometry for mCherry and/or EGFP expression. To determine levels of the NK-mediated control, the percentage of fluorescent cells in the presence of Ab and NK cells was normalized to the percentage of fluorescent cells in the presence of Ab alone.

Immunoblotting. HFFF-hCARs were transduced with RAd-UL141 or RAd-UL16 (MOI = 5 PFU/cell) for 48 hours. Whole-cell lysates were collected and boiled in reducing-denaturing Nu-PAGE lysis buffer (Thermo Fisher Scientific), separated by electrophoresis in Criterion TGX Gels (Bio-Rad) and transferred onto nitrocellulose membranes (GE Life Sciences). Membranes were blocked in TBS-T buffer with 5% dried nonfat milk and stained with either anti-V5 (clone CV5-Pk1, Bio-Rad) or anti-actin (A2066, MilliporeSigma) Abs. Proteins were visualized with SuperSignal West Pico PLUS Chemiluminescent Substrate (Thermo Fisher Scientific) and imaged on a GBOX-Chemi-XX6 gel documentation system (Syngene) operating GeneSys software.

Study approval. Healthy adult donors provided written informed consent for the collection of venous blood samples and dermal fibroblasts according to the principles of the Declaration of Helsinki. Study approval was granted by the Cardiff University School of Medicine Research Ethics Committee (reference number 16/52).

Statistics. Statistical significance was determined using a 1- or 2-way ANOVA as appropriate, with Sidak’s post tests. A P value of 0.05 or less was considered significant.

VMV, IM, RJA, KL, DAP, AJD, GWGW, MRW, ECYW, and RJS designed experiments. VMV, IM, LZ, RJA, EL, MRW, KLM, NMS, MPW, and RJS performed experiments and analyzed data. VMV, DAP, AJD, GWGW, MPW, ECYW, and RJS wrote the manuscript.

This work was supported by funding from the Wellcome Trust (100326/Z/12/Z, 204870/Z/16/Z, 108070/Z/15/Z) and the Medical Research Council (MRC) (MR/S00971X/1, MR/L008734/1, MR/P001602/1, MC_UU_12014/3, MR/S00081X/1. LZ was funded by a grant from Kymab, however, Kymab had no part in the design, execution, or analysis of the experiments reported in this study. The graphical abstract was created using BioRender.

Conflict of interest: The antibodies described in this manuscript are the subject of a patent application titled “Anti-viral therapeutics” (patent no. GB2101066.5).

Copyright: © 2021, Vlahava et al. This is an open access article published under the terms of the Creative Commons Attribution 4.0 International License.


Watch the video: Zákaz neočkovaných na pracovišti i testování v části škol, naznačil Hamáček zpřísnění (July 2022).


Comments:

  1. Thurl

    I am sorry, that I interfere, but it is necessary for me little bit more information.

  2. Trevls

    The post made me think * left to think a lot * ...

  3. Kazikree

    I'm finite, I apologize, but it doesn't quite come close to me.

  4. Asadel

    There is something in this and I think this is a great idea. I completely agree with you.

  5. Mataur

    Instead of criticism write the variants.

  6. Brak

    In it something is. Thanks for the help in this question. All ingenious is simple.

  7. Brazshura

    Congratulations, I think this brilliant idea



Write a message