1. Immunology and Inflammation
  2. Microbiology and Infectious Disease
Download icon

Development of antibody-dependent cell cytotoxicity function in HIV-1 antibodies

  1. Laura E Doepker
  2. Sonja Danon
  3. Elias Harkins
  4. Duncan K Ralph
  5. Zak Yaffe
  6. Meghan E Garrett
  7. Amrit Dhar
  8. Cassia Wagner
  9. Megan M Stumpf
  10. Dana Arenz
  11. James A Williams
  12. Walter Jaoko
  13. Kishor Mandaliya
  14. Kelly K Lee
  15. Frederick A Matsen IV
  16. Julie M Overbaugh  Is a corresponding author
  1. Human Biology Division, Fred Hutchinson Cancer Research Center, United States
  2. Public Health Sciences Division, Fred Hutchinson Cancer Research Center, United States
  3. Medical Scientist Training Program, University of Washington School of Medicine, United States
  4. Molecular and Cellular Biology Graduate Program, University of Washington and Fred Hutchinson Cancer Research Center, United States
  5. Department of Statistics, University of Washington, United States
  6. Department of Medicinal Chemistry, University of Washington, United States
  7. Department of Medicinal Microbiology, University of Nairobi, Kenya
  8. Coast Provincial General Hospital, Women’s Health Project, Kenya
Research Article
  • Cited 0
  • Views 699
  • Annotations
Cite this article as: eLife 2021;10:e63444 doi: 10.7554/eLife.63444

Abstract

A prerequisite for the design of an HIV vaccine that elicits protective antibodies is understanding the developmental pathways that result in desirable antibody features. The development of antibodies that mediate antibody-dependent cellular cytotoxicity (ADCC) is particularly relevant because such antibodies have been associated with HIV protection in humans. We reconstructed the developmental pathways of six human HIV-specific ADCC antibodies using longitudinal antibody sequencing data. Most of the inferred naive antibodies did not mediate detectable ADCC. Gain of antigen binding and ADCC function typically required mutations in complementarity determining regions of one or both chains. Enhancement of ADCC potency often required additional mutations in framework regions. Antigen binding affinity and ADCC activity were correlated, but affinity alone was not sufficient to predict ADCC potency. Thus, elicitation of broadly active ADCC antibodies may require mutations that enable high-affinity antigen recognition along with mutations that optimize factors contributing to functional ADCC activity.

eLife digest

Nearly four decades after the human immunodeficiency virus (HIV for short) was first identified, the search for a vaccine still continues. An effective immunisation would require elements that coax the human immune system into making HIV-specific antibodies – the proteins that can recognise, bind to and deactivate the virus.

Crucially, antibodies can also help white blood cells to target and destroy cells infected with HIV. This ‘antibody-dependent cellular cytotoxicity’ could be a key element of a successful vaccine, yet it has received less attention than the ability for antibodies to directly neutralize the virus. In particular, it is still unclear how antibodies develop the ability to flag HIV-infected cells for killing. Indeed, over the course of an HIV infection, an immune cell goes through genetic changes that tweak the 3D structure of the antibodies it manufactures. This process can improve the antibodies' ability to fight off the virus, but it was still unclear how it would shape antibody-dependent cellular cytotoxicity.

To investigate this question, Doepker et al. retraced how the genes coding for six antibody families changed over time in an HIV-carrying individual. This revealed that antibodies could not initially trigger antibody-dependent cellular cytotoxicity. The property emerged and improved thanks to two types of alterations in the genetic sequences. One set of changes increased how tightly the antibodies could bind to the virus, targeting sections of the antibodies that can often vary. The second set likely altered the 3D structure in others ways, potentially affecting how antibodies bind the virus or how they interact with components of the immune system that help to kill HIV-infected cells. These alterations took place in segments of the antibodies that undergo less change over time. Ultimately, the findings by Doepker et al. suggest that an efficient HIV vaccine may rely on helping antibodies to evolve so they can bind more tightly to the virus and trigger cellular cytotoxicity more strongly.

Introduction

A major concept underlying HIV vaccine research is that we can learn from the processes that lead to potent antibody responses in natural infection to inform immunogen design. For this reason, there have been several detailed studies of the evolutionary pathways of potent and broad neutralizing antibodies (bnAbs) to HIV (Bonsignori et al., 2017; Bonsignori et al., 2016; Doria-Rose et al., 2014; Landais et al., 2017; Liao et al., 2013; MacLeod et al., 2016; Simonich et al., 2019; Wu et al., 2015). These studies highlight the critical role of somatic hypermutation (SHM), particularly in complementary determining regions (CDRs) of antibody heavy chains, in driving the breadth and potency of HIV bnAbs (Kwong and Mascola, 2018). One recent study also highlighted that maturation of the antibody light chain, too, can be critical for breadth (Simonich et al., 2019). Collectively, these studies have provided important insights into the specific mutations that drive maturation of antibody lineages and specificities of bnAbs.

The keen interest in HIV bnAbs stems in part from a plethora of compelling studies that demonstrate the efficacy of bnAbs in protection of macaques from experimental SHIV infection (Pegu et al., 2017). The data are less convincing for sterilizing protection by HIV-specific non-neutralizing antibodies that mediate antibody-dependent cell cytotoxicity (ADCC) in the same experimental animal models (Fouts et al., 2015), where the functional interactions of antibodies are harder to test due to species differences (Bournazos et al., 2017). However, they have been implicated in protection from HIV-infection in humans in several settings. In the only partially effective HIV vaccine trial to date, antibodies that mediate ADCC were associated with protection, whereas neutralizing antibodies were not (Haynes et al., 2012). Moreover, in the setting of mother-to-child transmission, ADCC antibodies have been associated with both decreased transmission risk and disease progression in infants (Mabuka et al., 2012; Milligan et al., 2015). For these reasons, ADCC antibodies, which have the potential to eliminate infected cells, have been increasingly recognized as important to consider for vaccine design and prevention efforts (Lewis et al., 2017). Yet, nothing is known about the process of SHM required for HIV specificity and ADCC function of antibodies.

In light of the relevance of ADCC antibodies in human infections, we deep sequenced the antibody repertoire of an individual who developed potent ADCC antibodies to the HIV envelope (Env) gp120 and gp41 proteins. We used computational methods that were specifically developed to study antibody evolutionary pathways to infer the naïve precursors and resolve chronological lineage intermediates leading to six different mature ADCC-mediating IgG1 antibodies that have been previously described (Williams et al., 2015; Williams et al., 2019). Among the six lineages, most inferred naïve antibodies did not mediate ADCC. To develop ADCC functionality, inferred naive antibodies required CDR-localized SHM in one or both chains. In five of six cases, enhancement of ADCC potency to mature levels required mutations in framework regions (FWR), either alone or alongside additional CDR mutations. Interestingly, in one lineage, the developing antibodies first gained the capacity to bind HIV Env and then subsequently acquired ADCC activity due to additional SHM. Overall, binding affinity for Env and facilitation of ADCC were correlated, but we observed cases where binding affinity was similar amongst two antibodies within the same lineage but they differed in their ADCC capacity, suggesting that affinity alone is not sufficient for an antibody to mediate ADCC.

Results

Longitudinal sequencing of six ADCC HIV-1 antibody lineages

We previously reported the isolation of six monoclonal antibodies (mAbs), each with potent antibody-dependent cell cytotoxicity (ADCC) activity, from a sample collected 914 days post-infection (D914) from a clade A-infected Kenyan woman (Williams et al., 2015; Williams et al., 2019). All these mAbs derived from distinct B cell lineages, with four targeting two distinct gp41 Env epitopes (006, 016, 067, 072), and the other two targeting distinct gp120 Env epitopes (105, 157). One of the gp120 targeting mAbs (105) also had modest capacity to neutralize cell-free virus (Williams et al., 2015), whereas the others were non-neutralizing. To infer the ontogeny of these six mAbs, we used next-generation sequencing (NGS) to sequence full-length antibody variable region genes from five longitudinal blood samples collected from the subject, spanning from pre-infection (D-119) to over 4 years post-infection (D1512). Amongst the different antibody chains and timepoints sequenced, the sequencing libraries ranged widely in their sample coverage, between 6 and 409% with an average of 100% coverage, based on estimated B cell frequencies within the available peripheral blood mononuclear cell (PBMC) samples (Table 1). It is important to note that, despite the name ‘deep sequencing’, NGS-sampled datasets are relatively shallow compared to the entire repertoire. This is for two reasons: (1) each 10 mL PBMC sample is only ~0.22% of the subject’s whole-body circulating blood volume (~4.5 L), not counting lymphoid organs or peripheral tissue and (2) circulating human memory B cell repertoires fluctuate over time (Horns et al., 2019; Laserson et al., 2014), making it difficult to accurately track clonal B cell families. Despite these limitations inherent to studies of this type, our sequencing efforts successfully returned clonal sequences for all six lineages of interest.

Table 1
Longitudinal QA255 antibody variable region sequencing statistics.
Time pointLive PBMC count
(excluding non-viable)
PBMC viabilityAb chainReplicateRaw MiSeq readsProductive replicate-merged deduplicated sequencesEstimated sequencing coverage within sampled blood PBMCsEstimated sequencing coverage within QA255 d914 whole-body blood repertoire
D-1191.39E+0784.22%IgM1486,458200,25421%0.05%
IgG1604,237199,57048%0.11%
IgK1705,163292,54321%0.05%
IgL1406,267178,22113%0.03%
D4621.40E+0786.42%IgG1646,809504,029120%0.22%
2923,103
IgK157,096506,98936%0.08%
2967,693
IgL1201,566402,45629%0.06%
2586,568
D7913.90E+0697.50%IgG1126,955263,848226%0.22%
2776,578
3*522,308
IgK1527,031568,736146%0.22%
2934,780
3*1,057,984
IgL147,368212,92255%0.12%
2387,730
3*537,611
D11741.80E+06100.00%IgG1497,094220,740409%0.22%
2665,981
3*678,973
IgK1254,890415,601231%0.22%
2797,239
3*929,145
IgL1228,992307,415171%0.22%
2453,935
3*510,711
D15121.00E+0779.37%IgG1370,198155,26252%0.12%
IgK1408,498214,65121%0.05%
IgL193,84855,7856%0.01%
averages:293,689100%0.13%
  1. * Library replicate performed using unique molecular identifiers during cDNA synthesis step.

    Sequencing coverage was calculated for PBMCs using QA255 D914 B cell frequency statistics from the cell sort (Williams et al., 2015).

  2. Whole-body sequencing coverage was calculated assuming 10 ml of blood was sampled from a total of 4500 ml adult blood volume.

We identified sequences clonally related to the gamma heavy (VH) and lambda (VL) or kappa (VK) light chains of the six mature mAbs (Table 2; Figure 1A–B), inferred the most likely naive ancestors of each mAb chain, annotated clonal family V, (D), and J gene usage (Table 2), and calculated the degree of mutation in each lineage over time (Table 2; Figure 1C). Overall, the level of SHM in the mature D914 mAbs ranged from 3.3–11.5%, similar to our previous reporting of these lineages using different annotation methods (Williams et al., 2015; Williams et al., 2019). Incidentally, consensus among clonal sequences within families revealed potentially artifactual cloning mutations at the 5’ and 3’ ends of the variable regions in some of the previously characterized mature mAbs (Williams et al., 2015; Williams et al., 2019) and suggested, overall, that the consensus sequences were the most likely sequences for the antibodies. When directly compared, the sequence-corrected mAbs demonstrated equivalent epitope binding strength and ADCC activity to those previously studied (data not shown). Thus, we corrected the 5’ and 3’ ends of the mature D914 mAb variable regions to consensus sequences among each clonal family for use in this study (Supplementary file 1).

Table 2
Clonal sequence analyses for ADCC antibody lineages.

(A, B) Characteristics and statistics of heavy (A) and light (B) chain clonal families. Percent SHM was calculated as the mutation frequency at the nucleotide level compared to the predicted naïve allele, as determined by the per subject germline inference for QA255. Statistics calculated for individual timepoints within the context of the full subject repertoire were downsampled to 50–150K sequences for computational manageability. Light chain clonal family size statistics are reported but unreliable due to overclustering caveats that are explained in the Materials and methods. Per-timepoint clonal family statistics were excluded if ≤1 clonal heavy or light chain sequences were identified. #: The VH gene used in 067 and 157 lineages was determined to be an allele not cataloged in IMGT (http://www.imgt.org/): 1–69*11+A147G+C169Tl; the 157 lineage VK gene was also a new allele: 3–11*01+T5A+T9A+T36G+G84A. †: Percent SHM for 016 light chain lineage includes a 3-nt insertion in the CDRL3. SHM: somatic hypermutation; ints: intermediates. Longitudinal values for ‘No. of unique clonal sequences identified’ are plotted for heavy and light chains in Table 2; Figure 2—figure supplements 1, along with longitudinal values for ‘Average %SHM (nt)’ for heavy chains. Graphics displaying the most probable routes of antibody lineage maturation corresponding to the ‘No. of ints resolved in lineage path’ are available in Figure 4 and Figure 4—figure supplements 13.

Heavy chainLineageMature antibody statisticsClonal family (all timepoints merged, not downsampled)Clonal family statistics within repertoire (downsampled analyses)No. ints resolved in lineage path
Gene usage (partis)%SHM mature Ab# sampled clonal seqsAvg clonal family %SHMNo. unique clonal sequences identifiedSize: percent of sampled repertoireAverage %SHM (nt)
VHDHJHD-119D462D791D1174D1512D-119D462D791D1174D1512D-119D462D791D1174D1512
 0063–23*013–22*014*028.7710.8070000.002%7.61
 0164–34*016–13*015*029.97029.20581112090.13%0.05%0.01%9.610.312.614
 0671–69*11#1–1*016*038.13611.90351000.01%11.62
 0721–69*114–17*013*0211.551512.5011118325610.04%0.05%0.17%0.04%9.512.813.015.15
 1053–15*013–22*016*039.51934.901930000.04%52
 1571–69*11#1–26*016*036.51837.406165840.004%0.08%0.04%0.03%9.37.710.311.84
Light chainLineageMature antibody statisticsClonal family (all timepoints merged, not downsampled)Clonal family statistics within repertoire (downsampled analyses)No. ints resolved in lineage path
Gene usage (partis)%SHM mature Ab# sampled clonal seqsAvg clonal family %SHMNo. unique clonal sequences identifiedSize: percent of repertoireAverage %SHM (nt)
VKVLJK/LD-119D462D791D1174D1512D-119D462D791D1174D1512D-119D462D791D1174D1512
 0062–11*012*016.86736.11227427352620.14%7.17
 0161–51*022*019.9†15187.3299339523321360.03%0.02%0.11%6.67.17.711
 0672–11*013*023.320,3376.521496105340879467292.04%6.62
 0721–27*011*017.827966.24767197025433560.17%0.18%0.18%0.25%6.36.36.67.05
 1053–20*012*017.130,2868.03058488456071427124661.17%6.37
 1573–11*015*015.0188.20120150.24%0.07%5.95.37
Longitudinal characteristics of heavy and light chain clonal families.

(A, B) Number of clonal heavy (A) and light (B) chain sequences identified per sequenced timepoint per lineage. Caveats relating to light chain clusters are described in Methods. (C) Average variable region mutation for each clonal heavy chain family per timepoint. Dashed line at D914 indicates the timepoint of isolation of mature antibodies (Williams et al., 2015; Williams et al., 2019). SHM: somatic hypermutation.

mAbs constructed from D462 NGS sequences mediated ADCC

Each lineage varied in the time point(s) from which we identified clonal sequences, with most clonal VH sequences existing in the D462 datasets for four of the six lineages (Table 2A). The NGS sequences within each lineage that shared the highest nucleotide (nt) identity with the D914 mature VH variable regions, ranging from 88–99%, were found within the D462 or D791 timepoints (Supplementary file 2). As test cases, we synthesized antibodies with VH sequences derived from the D462 NGS results that were similar to the mature 006 and 016 mAbs (88.5% and 97.5% nt identity, respectively). The 006-NGSVH was paired with 006-NGSVL light chain that had 95.8% nt identity to the mature 006 VL (Figure 2—figure supplement 1A) and, since this option was not available for 016 VL, the 016-NGSVH was paired with a computationally-inferred light chain (016-2VL) that preceded the mature in development (Figure 2—figure supplement 1B). mAb functionalities were then assessed using the Rapid and Fluorometric ADCC assay, which uses primary PBMCs as target and effector cells (Gómez-Román et al., 2006); in this method, mAbs bind to antigen-coated target cells and facilitate lysis by effector cells. Both NGS-based mAbs demonstrated full ADCC function (Figure 2), suggesting that ADCC function likely developed within these lineages by D462 post-infection.

Figure 2 with 1 supplement see all
ADCC activity in antibodies made from NGS-sampled sequences.

RFADCC activity of antibodies that use NGS-sampled heavy chains that are most similar in sequence to the mature 006 or 016 antibodies. The 006-NGSVH was paired with an NGS-sampled light chain; the 016-NGSVH was paired with a computationally-inferred mature-like sequence 016-2VL. mAb pairings and their mutations are illustrated in Figure 2—figure supplement 1, with sequences available in Supplementary file 1. Data reflect two independent experiments. Source data is available in Figure 2—source data 1.

Figure 2—source data 1

Source data (both replicates) for RFADCC assessment of NGS mAbs, processed as detailed in Materials and methods.

https://cdn.elifesciences.org/articles/63444/elife-63444-fig2-data1-v2.csv

Inferred naive mAbs vary in antigen binding capability and ADCC function

Inferred naïve ancestor mAbs across lineages varied in their ability to bind HIV Env and facilitate ADCC function. Of the gp41-targeted lineages, only the inferred naive precursor of the 072 lineage bound gp41 (KD = 34.4 nM) and mediated detectable ADCC, although the activity was very low (Figure 3A, C). Both naive mAbs from the gp120-targeted lineages bound monomeric gp120 (KD = 23.4 nM for 105-0VH0VK and 41.1 µM for 157-0VH0VK; Figure 3B). Interestingly, despite the 105 naïve mAb having higher binding affinity compared to 157 naïve mAb, it did not mediate ADCC, while the 157 naive mAb mediated potent ADCC (Figure 3C).

Figure 3 with 1 supplement see all
Inferred naïve mAbs from six lineages vary in antigen binding capability and ADCC function.

(A–B) Binding kinetics of the inferred naive antibody (ligand) from each lineage to indicated concentrations of monomeric C.ZA.1197MB gp41 ectodomain (A) or BL035.W6M.C1 gp120 (B) (analyte). Best fitting lines (red) to a 1:1 binding model of ligand:analyte binding are shown. Data are representative of two independent experiments. (C) Positive control-normalized RFADCC activity of inferred naive antibodies compared to their respective mature antibodies. Normalization is described in Methods. Asterisks indicate indeterminate activity, as defined in Methods. Data are represented as mean ± SEM and reflect at least five independent experiments; source data are available in Figure 3—source data 1. 0: naive; m: mature. See Figure 3—figure supplement 1 for functional assessment of alternative naïve mAbs.

Figure 3—source data 1

Source data (all replicates) for RFADCC assessment of inferred naive mAbs (Figure 3C), processed as detailed in Materials and methods.

https://cdn.elifesciences.org/articles/63444/elife-63444-fig3-data1-v2.csv

Typically, the uncertainty in the computational inference of antibody lineage naive precursors is largely ignored, even though single amino acid differences in the naive mAb can result in differences in HIV Env binding and functional properties (Yuan et al., 2011). To account for inference uncertainty, we also tested ‘alternative naive’ mAbs for each lineage that were inferred not by partis (Ralph and Matsen, 2016b), but by linearham (Dhar et al., 2020), a second computational method that jointly models V(D)J recombination and evolutionary history. For 7 of 12 antibody chains, this alternative approach generated naïve sequences that differed from the original naïve sequences by one to four amino acids within the CDR3 and/or FWR4 regions (Supplementary file 1). For the remaining five, the alternative method inferred the same naive sequence as the original method. When tested, the alternative naïve sequences did not alter antigen binding or function in five of six antibody lineages when compared to the original naïve mAbs (Figure 3—figure supplement 1A). However, in the 157 lineage we found that the alternative naïve VK chain (157-AVK) differed from the original naïve (157-0VK) by two adjacent amino acids in the CDRL3 (Figure 3—figure supplement 1D). When 157-AVK was paired with either the original (157-0VH) or alternative (157-AVH) VH chains, it abolished gp120 binding and ADCC functionality (Figure 3—figure supplement 1B,C). Thus, there is some uncertainty about whether the true 157 lineage naïve precursor was specific for HIV and/or capable of mediating ADCC.

Characterization of lineage intermediates that confer HIV specificity and ADCC function

To reconstruct probable developmental routes that led to the mature D914 mAb heavy and light chains, we used Bayesian phylogenetic lineage inference, which emphasizes only the inferred ancestral intermediates that have high relative statistical confidence and infers chronological ordering of these likely ancestral sequences. This method was specifically developed to reconstruct antibody evolution using sparse data (Simonich et al., 2019). Among the 12 heavy and light chain lineages for the six mAbs, our methods resolved anywhere from 1 to 14 probable intermediate sequences that lay between the naive and the mature sequences (Table 2). Figure 4, showing 105VK development, exemplifies one pathway, although all pathways may be viewed in Figure 4—figure supplements 13. Of note, 8 of 12 of these computationally-inferred lineages were validated by the existence of NGS-sampled sequences that were identical either in nucleotide or amino acid sequence to at least one of the inferred intermediates (Figure 4, Figure 4—figure supplements 13).

Figure 4 with 3 supplements see all
Example of most probable routes of antibody maturation: 105 light chain.

Ancestral sequences (internal nodes within phylogenies) and their relative confidences were summarized for the 105 antibody light (kappa) chain from Bayesian clonal family phylogenies sampled from the associated posterior distribution for the 105-light clonal family of sampled NGS sequences. Results are displayed in this graphic illustrating multiple possible lineages of amino acid transitions and their relative confidences for 105 light chain development. Amino acid substitutions (arrows) connect the inferred naive sequence (white node, top) to the mature sequence (white node, bottom) via reconstructed ancestral intermediate sequences (red-shaded nodes). The red shading of nodes is proportional to the posterior probability that this ancestral sequence was present in the lineage. Low probability nodes were filtered out, resulting in some incomplete pathways within the graphics. For a given node, the blue shadings of multiple arrows arising from that node is proportional to the corresponding transition probability from that node to downstream nodes. Stars denote inferred sequences that were identical in nucleotide (green) or amino acid (yellow) sequence to sampled NGS sequences. ‘MID’ (red) denotes middle intermediate sequence. Labels with cyan highlighting denote sequences that were used in functionality assays (see also Supplementary file 1). See Figure 4—figure supplements 13 for probable maturation graphics for all 12 antibody chains of interest.

To focus our subsequent lineage studies on determining which heavy chain and/or light chain mutations enabled HIV Env binding and ADCC gain-of-function, we employed the following strategy to select inferred intermediates of interest (detailed fully in Materials and methods). For each chain, we chose lineage intermediates with high relative confidence that were chronologically near the middle of the inferred lineage, paired them together (see Materials and methods), and performed preliminary experiments to determine whether antigen binding or ADCC gain-of-function occurred before or after these ‘middle’ intermediates (Figure 4, Figure 4—figure supplements 13). If the middle intermediate had function, we focused on pre-middle inferred intermediates, if available within the computationally-inferred lineages. If the middle intermediate lacked function, we focused on post-middle inferred intermediates. Selected heavy and light chain sequences, were paired together in all possible combinations, roughly chronologically (i.e. increasing SHM and within per-chain resolved chronologies), to reveal per-chain mutations contributing to mAb function. Because we did not use paired sequencing methods, these pairings do not necessarily reflect true biological intermediates, but, instead, allowed us to identify ordered steps within each chain’s maturation that conferred HIV binding and/or ADCC activity. For clarity, we are reporting only the pairings that revealed steps in gain-of-function for the six lineages. Pertinent mAbs thus defined the mutations that conferred HIV binding and ADCC activity in each lineage as follows: gp41 antibody 006: mAb 006 (VH3-23, VL2-11) targets a discontinuous epitope that includes the C-terminal heptad repeat (Williams et al., 2019). Mutations in CDRH2 and CDRH3, totaling 1.4% VH SHM, were minimally required for antigen binding and ADCC activity by 006-1VH0VL in the 006 lineage (Figure 5A). The addition of six light chain mutations among CDRL1, CDRL3, and FWRL2 augmented gp41 binding and ADCC potency in 006-1VH1VL.

HIV specificity and ADCC development in gp41-targeted antibody lineages.

(Left) Mature-normalized RFADCC activity of antibody lineage members for lineages 006 (A), 016 (B), 067 (C), and 072 (D), colored by strength of functionality: no function or indeterminate (gray), low function (<50% of mature activity; light blue), high function (>50% of mature activity; dark blue). Intermediate sequences are numbered consecutively based on their position in the developmental pathway between naïve (0) and mature (m). Asterisks indicate indeterminate activity, as defined in Methods. Data are represented as mean ± SEM and reflect at least four independent experiments, including data presented in Figure 3 for naive and mature Abs to best account for assay variability and to compare intermediates directly to these antibodies. Source data for all replicates are available in Figure 5—source data 1. (Middle) Binding of lineage members (ligand) to monomeric C.ZA.1197MB gp41 ectodomain (analyte, 62.5 nM), measured by BLI. Data are representative of two independent experiments. Data are colored based on each antibody’s RFADCC functionality. (Right) mAb heavy and light chain pairings illustrating variable region amino acid substitutions with respect to the reference sequence at the top of each lineage: black lines indicate amino acid substitutions; gray shaded regions demarcate CDRs.

Figure 5—source data 1

Source data (all replicates) for RFADCC assessment of Ab lineages (Figures 5 and 6), processed as detailed in Methods.

Duplicate columns accommodate for instances when two replicates were run on the same day.

https://cdn.elifesciences.org/articles/63444/elife-63444-fig5-data1-v2.csv

gp41 antibody 016: mAb 016 (VH4-34, VL1-51) targets a similar gp41 epitope as 006 (Williams et al., 2019). Detectable gp41 binding and ADCC function was demonstrated by the 016-1VH1VL mAb which contains mutations in heavy and light chain CDRs and FWRs (5.6% VH SHM and 3.6% VL SHM) (Figure 5B). A subsequent CDRL3 insertion of residue S94 (see details on chronology resolution of this insertion in Methods) allowed for augmented ADCC capacity (016-1VH2VL). Additional FWR and CDR mutations in the VH chain (016-2VH2VL) further strengthened gp41 binding and ADCC functionality.

gp41 antibody 067: mAb 067 (VH1-69, VL2-11) targets the C-C’ loop (Williams et al., 2019). For the 067 lineage, the 067-0VH1VL mAb incorporated a single CDRL3 mutation (0.3% VL SHM) and displayed detectable antigen binding; we could not determine whether there was meaningful ADCC function or not as levels were just above background (Figure 5C). To achieve more convincing ADCC function, two CDRH2 mutations were also required (1.0% VH SHM). Further mutations in either the heavy or light chain FWRs modestly increased ADCC potency to >50% that of the 067 mature mAb.

gp41 antibody 072: mAb 072 (VH1-69, VK1-27) targets an overlapping but distinct epitope as 067 (Williams et al., 2019). Unlike the other gp41-targeting lineages, the 072 lineage inferred naive mAb bound gp41 antigen and mediated weak ADCC, as aforementioned (Figures 3A,C and 5D). To gain ADCC potency, only mutations in the VH chain were necessary; mAbs 072-1VH0VK (0.8% VH SHM) and 072-2VH0VK (0.5% VH SHM) demonstrate that two substitutions in either the CDRH1 or CDRH2 increased activity, but, when combined in 072-3VH0VK, they were not additive or synergistic (Figure 5D). Instead, FWRH3 and FWRH4 mutations in mAb 072-4VH0VK were required for ADCC activity comparable to that of 072 mature mAb.

gp120 antibody 105: mAb 105 (VH3-15, VK3-20) targets a V3 linear epitope (Williams et al., 2015). The inferred 105 lineage naive and 105-0VH1VK mAbs bound gp120 but displayed only indeterminate levels of ADCC activity, as defined in Materials and methods (Figure 6A). 105-1VH1VK mAb demonstrates that heavy chain CDRH1 and CDRH2 mutations (1.9% VH SHM) were required for ADCC gain of function, even though these mutations did not affect binding affinity (Figure 6A). mAb 105-1VH1VK lacked full breadth, however, in that it mediated ADCC against clade A HIV strain BG505.W6M.B1, but not clade C CAP210.2.00.E8 (Figure 6—figure supplement 1A). ADCC breadth matching that of the 105 mature was achieved, instead, by 105-1VH2VK. Lastly, we note that the FWR4 mutation in the 105-1VH chain was likely a computational artifact and unlikely to affect mAb function, as this mutation is absent in the mature 105 mAb.

Figure 6 with 1 supplement see all
HIV specificity and ADCC development in gp120-targeted antibody lineages.

RFADCC activity (left), antigen binding to monomeric BL035.W6M.C1 (analyte, 62.5 nM) as measured by BLI (middle), and mAb heavy and light chain pairings (right) for the 105 and 157 lineages all displayed as in Figure 5. Asterisks indicate indeterminate activity, as defined in Methods. RFADCC data are represented as mean ± SEM and reflect at least four independent experiments, including data presented in Figure 3 for naïve and mature Abs to best account for assay variability. Source data for all replicates are available in Figure 5—source data 1. See also Figure 6—figure supplement 1 and Figure 6—source data 1 for RFADCC breadth data.

Figure 6—source data 1

Source data (all replicates) for RFADCC breadth assessment of Ab lineages 105 and 157.

Unlike the Maerials and methods section details for other RFADCC assays, data here are processed without normalization, just subtraction of background and negative-control signal from each sample’s activity.

https://cdn.elifesciences.org/articles/63444/elife-63444-fig6-data1-v2.csv

gp120 antibody 157: mAb 157 (VH1-69, VK3-11) targets a CD4-induced, C11-like epitope (Williams et al., 2015). As detailed in Figure 3—figure supplement 1 and further shown in Figure 6B, inferred naives 157-0VHAVK and 157-0VH0VK differed in their binding and ADCC capabilities, leaving us uncertain about whether this lineage was able to bind gp120 and facilitate ADCC since inception or gained these capacities following two CDRL3 mutations (Figure 6B). The 157-0VH0VK naive lacked full ADCC breadth, however. While this mAb demonstrated high (>50%) ADCC potency against two clade A HIV strains (Figure 6B and Figure 6—figure supplement 1B), it had low functionality against autologous QA255 transmitted founder virus and a clade C strain, and failed to mediate any ADCC against two other strains of clades C/D and B (Figure 6—figure supplement 1B).

Themes in ADCC development pathways among six lineages

Overall, the majority of lineages, regardless of epitope specificity, ultimately required substitutions in both the heavy and light chains to develop ADCC function >50% as potent as their respective mature mAb (Figure 7). In five of six lineages, detectable binding accompanied ADCC capacity, with lineage 105 being the exception that bound prior to developing ADCC function. Most notably, substitutions in CDRs were typically required for ADCC gain-of-function, while subsequent FWR substitutions, either alone or alongside additional CDR mutations, augmented this activity (Figure 7). Since the chronology of inferred ancestral sequences was only resolved on a per-chain basis, we note that heavy and light chain mutations likely co-occurred in an interlaced fashion. However, in two cases (016 and 105), boosted ADCC potency was conferred by specific mutations in a single chain that were resolved to have occurred subsequent to mutations already incorporated in the prior intermediate. In other cases, namely 006, 067, and 072, where we cannot know if FWR mutations occurred before or after critical CDR mutations, we can only conclude that the FWR mutations are required in addition to CDR mutations to boost ADCC potency. A graphical summary of the key developmental steps for each lineage is presented in Figure 7—figure supplement 1.

Figure 7 with 1 supplement see all
Locality of mutations required for gain of function.

Summary of the location of mutations critical for acquiring binding and activity functions in all five ADCC Ab lineages, respective to the previous step, that is building upon one another to gain functions. Lineages are summarized by row. Detectable ADCC function excludes lineage members with indeterminate ADCC activity, as defined in Materials and methods. Augmented ADCC activity indicates lineage members with activity most comparable to that of their respective mature. *157-0VHAVK required mutations for function while 157-0VH0VK did not; Env binding and detectable ADCC determinants were defined based on this comparison. fnl: functional. See Figure 7—figure supplement 1 for a more detailed summary graphic of key steps to attain function in all six lineages.

Antigen binding affinity and ADCC function correlate

In our detailed analyses of binding and ADCC activity, there were three lineages (067, 072, and 105) where we observed two mAbs from the same lineage having similar binding affinities but different ADCC capabilities. To visualize the relationship between binding affinity and ADCC potency, we plotted mAb binding affinities (KD) against corresponding RFADCC potencies and found that these functionalities were positively correlated in all lineages (Figure 8). It is notable that the antibodies in the 157 lineage that mediate potent ADCC (naive 157-0VH0VK and mature 157-mVHmVK) bound gp120 antigen with much lower affinity than antibodies with similar ADCC potency in the other five gp41- or gp120-targeted lineages. These data indicate that binding affinity alone does not dictate ADCC potency.

Antigen binding affinity and ADCC function correlate.

For antibodies within lineages 006, 016, 067, 072, and 105 (indicated by different colors), the log of reciprocal binding affinity, as determined by representative BLI data of least two independent experiments, is plotted against positive control-normalized RFADCC activity (the data shown in Figures 5 and 6). Antibodies that did not detectably bind antigen by BLI are excluded as none of these mediated ADCC. Dotted lines highlight comparisons between antibodies within the same lineage that have similar binding affinities, but different RFADCC activities. Source data for each mAb are available in Figure 8—source data 1.

Figure 8—source data 1

Source data for correlation of binding affinity (KDs determined by BLI, see Materials and methods) and average RFADCC function (Figures 5 and 6; Figure 5—source data 1) for each mAb.

https://cdn.elifesciences.org/articles/63444/elife-63444-fig8-data1-v2.csv

Linear epitopes of ADCC Ab lineages remain stable over time

Lineages 067, 072, and 105 target defined linear epitopes and longitudinal sequence data from subject QA255 was available to evaluate evolution within these epitopes. The 067 and 072 lineages target overlapping C-C’ loop epitopes (Williams et al., 2019), and, amongst 28 longitudinal QA255 Env sequences (Bosch et al., 2010), these epitopes remain stable between 21 dpi and 1729 dpi (Env sequences GenBank accessions MW383929-MW383956).

Lineage 105 targets the V3 loop and there were changes in this region among the QA255 Env sequences, although it was unknown if the sequence changes would impact binding of 105 to its epitope. To address this, we finely mapped the epitope and sites of escape for both the mature and naive mAbs using a phage-display deep mutational scanning (Phage-DMS) approach (Garrett et al., 2020). The mAbs were screened using a library displaying gp41 and V3 peptides from HIV strains BG505.W6M.C2 (clade A), BF520.W14M.C2 (clade A), and C.ZA.1197MB (clade C). Both naive and mature 105 lineage mAbs enriched (i.e. bound to) wild-type peptides generated in the backgrounds of BG505.W6M.C2 and C.ZA.1197MB Env, but not peptides of the BF520.W14M.C2 Env background (Figure 9—figure supplement 1). The differences in binding are likely explained by sequence differences at sites 307–309 in the V3 loop: IRI (bound) vs. VHL (not bound).

The phage display library also contained mutant peptides representing every possible single amino acid mutation in the context of all three envelope variants, allowing us to map mutations that enable escape from antibody binding. The main sites of selection indicative of escape centered around residues 308–316 spanning the sequence RIGPGQA (Figure 9A–B, Figure 9—figure supplement 2). The footprints between the naïve and mature mAbs were similar, with perhaps stronger selection observed in the naïve mAb at some positions such as 309 and 315. However, overall the data suggest that SHM and affinity maturation in this lineage did not affect the major epitope binding footprint.

Figure 9 with 2 supplements see all
Lineage 105 epitope largely remains stable over time.

(A–B) Heatmaps depict scaled differential selection: the relative effect of each possible substitution within the V3 region of Env (HXB2 numbering), compared to wild type BG505.W6M.C2 Env, on binding by (A) 105 inferred naïve mAb and (B) 105 mature mAb. Wild-type residues are indicated by black dots. Mutations enriched above the wild-type residue are colored blue while those depleted are colored red; color intensity reflects the relative amount of differential selection as indicated by the key. Data are the average of two biological replicates. See also Figure 9—figure supplement 1 for peptide enrichment plots and Figure 9—figure supplement 2 for scaled differential selection in C.ZA.1197MB Env. Figure 9—source code 1 and Figure 9—source datas 13 contain fold enrichment and scaled differential selection data and calculations. (C) Logo plots of the residues comprising the 105 lineage epitope amongst 28 total longitudinal QA255 Env sequences (GenBank MW383929-MW383956) using HXB2 numbering.

Figure 9—source code 1

R markdown file detailing and performing the analysis of Phage-DMS deep sequencing data in order to map antibody-targeted Env epitopes.

https://cdn.elifesciences.org/articles/63444/elife-63444-fig9-code1-v2.zip
Figure 9—source data 1

CSV file required for R analysis: gp41V3_PhageDMS_library_key.

This file maps the numbering between each Env strain and the HXB2 reference strain.

https://cdn.elifesciences.org/articles/63444/elife-63444-fig9-data1-v2.csv
Figure 9—source data 2

CSV file required for R analysis: Phage-DMS-105epitope-Rep1annotatedCounts.

This file contains raw counts of each sequence for each sample, with annotated information for each sequence and peptide (replicate 1).

https://cdn.elifesciences.org/articles/63444/elife-63444-fig9-data2-v2.csv
Figure 9—source data 3

CSV file required for R analysis: Phage-DMS-105epitope-Rep2annotatedCounts.

This file contains raw counts of each sequence for each sample, with annotated information for each sequence and peptide (replicate 2).

https://cdn.elifesciences.org/articles/63444/elife-63444-fig9-data3-v2.csv

We compared the longitudinal V3 sequence data (Bosch et al., 2010) with the detailed information on the amino acids important for binding defined by Phage-DMS. There was little evidence of strong selection pressure on the 105 epitope of Env between 21 dpi and 1729 dpi. The only sweeping mutation (H308R) emerged at 189 dpi (50% frequency) and swept the population by 560 dpi (Figure 9C). We infer that this substitution enabled the initial activation of the 105 lineage because this lineage prefers R308 over H308, as aforementioned, and it is known that 105 mAbs cannot mediate ADCC against autologous transmitted founder QA255.21P.A17 gp120 (Williams et al., 2015), which contains H308. Except for a subset of Q315R viruses present at 189 dpi and 560 dpi, all 308–316 residues remain stable until 1729 dpi, when substitutions are observed in 7/10 clones at either residue 308 or 316, with A316T present in 50% of 1729 dpi clones (Figure 9C). Overall, analyses of sequence variation in the 067, 072, and 105 lineages suggest limited selection pressure for escape from ADCC antibodies in this subject.

Discussion

Renewed interest in ADCC-capable antibodies as important to HIV vaccine responses has highlighted the need for a better understanding of their natural development (Forthal and Finzi, 2018). Here, we report the evolution of six ADCC antibody lineages within a single HIV-infected individual. The inferred naïve precursors of these ADCC lineages varied in their abilities to bind HIV antigen, a finding that agrees with studies of HIV-neutralizing antibody lineages (Stamatatos et al., 2017). To achieve potent ADCC activity, most lineages required mutations in both their heavy and light chains. There was also evidence that some of these changes contributed to greater breadth. Generally, ADCC function was achieved through mutations in CDRs, while increased ADCC potency required additional mutations in FWR. ADCC activity accompanied antigen binding in all lineages except one; in the exception, the V3/gp120-targeting 105 lineage demonstrated binding capacity prior to developing ADCC functionality. Although there was some evidence that binding affinity was not solely responsible for ADCC capability or potency, binding affinity and ADCC activity were largely correlated. In sum, this study presents six examples of developmental pathways taken by ADCC-mediating antibodies, highlighting that there are common themes in the ontogeny of HIV-specific ADCC antibodies, but that each evolutionary pathway has unique features.

Although not widely applied, antibody lineage studies require robust inference methods that account for statistical uncertainty if we are to be confident in their conclusions. This is highlighted in this study by the different properties displayed by two very similar inferred naïve precursors of the 157 lineage. The development of the 157 lineage was ambiguous: computational uncertainty in two CDRL3 residues resulted in multiple probable naive antibodies with different properties. In these different scenarios, the 157 naïve Ab was either fully capable of both antigen binding and potent ADCC, or it required two CDRL3 substitutions to achieve both functions. In the other five lineages, our use of multiple, tailored computational approaches allowed us to report results with high confidence, since the uncertainties present in computational inference of those lineages did not affect functional characteristics.

Three gp41-specific antibody lineages were not capable of binding HIV Env by their inferred naïve precursors, as measured by BLI; they developed this ability through mutation. The lack of detectable antigen binding by inferred naïve precursors is not altogether surprising as this has been observed for a number of HIV bnAbs (Stamatatos et al., 2017). Binding capacity was gained in two of the gp41-directed lineages (006 and 067) through limited mutations localized in the CDRs of single antibody chains (either heavy or light, respectively). The other gp41-directed lineage, 016, which targets a similar epitope as the 006 lineage, required many mutations amongst the CDRs and FWRs of both chains, highlighting that the gain of activity by different lineages can vary even if they are targeting the same epitope. As is true for all studies of this type, it is possible that our methods were not sensitive enough to detect low level binding or that, despite our best efforts, the inferred naïve sequences were inaccurate. Gp41-directed antibodies have been found to recognize self-antigens (Verkoczy and Diaz, 2014; Verkoczy et al., 2014) and this, or its response to another antigen, could explain how the 016 lineage emerged before gaining HIV-specificity.

We chose the RFADCC assay because RFADCC signal has been repeatedly correlated with disease outcome in humans (Dhande et al., 2018; Lewis et al., 2019; Mabuka et al., 2012; Madhavi et al., 2014; Milligan et al., 2015; Ruiz et al., 2017), thus making it relevant to understanding the development of potentially protective responses. Additionally, we previously demonstrated that the BL035.W6M.C1 strain of gp120 used here yields RFADCC results representative of activity against seven different strains of various clades (Milligan et al., 2015). There was evidence that the 072 lineage had ADCC function from inception, although the amount of ADCC was low. The remaining lineages required an average of 1.2% SHM to gain detectable ADCC function (ranging 0.1–4.3% SHM). For augmentation of ADCC function to >50% of mature activity, the lineages required an average of 2.2% SHM (range 0.6–4.8% SHM). NGS-sampled sequences that were present at 462 dpi displayed high function with 5.3% SHM (006 lineage), and developmental inferences for the 006-VH, 067-VH and 105-VH chains were informed by clonal sequences from only the 462 dpi time point, suggesting that the fully functional VH intermediates inferred for these lineages were likely present prior to 462 dpi. This is not surprising, as ADCC responses to HIV can develop within the first few months of infection (Dugast et al., 2014). Low mutation levels to achieve potent ADCC are promising for ADCC vaccine efforts, especially since these six characterized lineages have desirable characteristics for combating HIV, such as potent ADCC activity, cross-clade ADCC breadth, highly conserved epitopes, and potential to help clear infected cells (Williams et al., 2015; Williams et al., 2019). Moreover, these levels of mutation are easily attainable by vaccines (Schramm and Douek, 2018).

Importantly, most lineages required FWR mutations in one or both chains to achieve potent ADCC functionality, with or without additional CDR mutations. Although they can cost thermostability (Henderson et al., 2019), FWR mutations are known to contribute to the properties of HIV-specific bnAbs, predominantly by altering Ab conformation in ways that allow conserved epitopes to be bound with high affinity (Henderson et al., 2019; Kepler and Wiehe, 2017; Ovchinnikov et al., 2018). FWR residues contribute to the structure of the Ab and, though they do not usually directly contact the antigen, they can affect binding affinity indirectly (Foote and Winter, 1992). Furthermore, since they indeed affect Ab structure, FWR mutations could confer favorable Ab flexibility or an orientation that mediates better Fc receptor dimerization and ADCC potency (Acharya et al., 2014). FWR mutations in Fab ‘elbow’ regions can affect conformational flexibility and paratope plasticity during bnAb development (Henderson et al., 2019). Here, the 006-1VH lineage intermediate caused ADCC gain-of-function and it contained, among other mutations, a CDRH3 elbow mutation Y111F. Future structural studies are warranted to explore how FWR mutations affect Ab structure and function in these lineages.

Binding affinity has been implicated in the regulation of ADCC in several studies on tumor-specific Abs (Mazor et al., 2016; Tang et al., 2007). Here, while binding affinity correlated with ADCC function, it was not solely responsible for ADCC potency. In one lineage, binding capacity preceded detectable ADCC function. Notably, we found that half of the lineages contained clonally-related antibodies with similar binding affinity to one another but markedly different ADCC potencies. Conversely, mAbs within the two gp120-targeted lineages exemplify that there can be dramatic differences in binding affinity among mAbs that mediate equivalent, potent ADCC activity. Thus, ADCC function cannot be predicted based on affinity alone. These data support the notion that factors such as epitope specificity, binding mode, and other antibody-antigen interaction dynamics contribute to ADCC potency, as other studies have emphasized (Acharya et al., 2014; Orlandi et al., 2020; Tolbert et al., 2020).

Amongst Env sequences collected over ~4.5 years of HIV infection in subject QA255, we observed little directional escape in the C-C’ loop and V3 epitopes targeted by the 067, 072, and 105 antibody lineages, and only at later stages of infection. This finding could indicate that selective pressure exerted by ADCC antibodies is often not enough to drive escape, although there is evidence that escape can occur (Isitman et al., 2012). There are at least two possible explanations for limited escape from ADCC antibodies in an infected individual: one is that the activity of ADCC antibodies is insufficient to drive escape because they have limited impact on infection dynamics; the other is that the targeted killing of infected cells leads to less pressure for escape compared to the process of blocking virus entry mediated by neutralizing antibodies. If the latter is true, ADCC could be particularly useful components of vaccines responses and therapeutic tools.

Our study offers a granular understanding of how ADCC functionality developed in six human Ab lineages. The results of this study are relevant to rational vaccine design, especially as technologies for guiding Ab development continue to emerge through reverse vaccinology efforts (Burton, 2017; Rappuoli et al., 2016; Schramm and Douek, 2018). Additionally, the reliance on FWR mutations to achieve high ADCC functionality could inform mAb optimization strategies for therapeutic uses.

Materials and methods

Key resources table
Reagent type
(species) or
resource
DesignationSource or
reference
IdentifiersAdditional
information
Cell line
(human, female)
FreeStyle 293 F cellsInvitrogenCat#R790-07; RRID:CVCL_D603mAb production
Cell line (human, female)CEM.NKR cells; target cellsNIH AIDS Reagent ProgramCat#458; RRID:CVCL_X622RFADCC
Biological sample (human, female)QA255 PBMCs; D-119; D462; D791; D1174; D1512PMID:12060878Subject ID:QA255Prospective cohort of HIV-1 negative high-risk women in Mombasa, Kenya; longitudinal PMBC samples from 119 days pre-infection, 462 days post-infection(dpi), 791 dpi, 1174 dpi, and 1512 dpi
Antibody (human, female)006-0VH0VL; 006-1VH0VL; 006-1VH1VL; 016-0VH0VL; 016-1VH1VL; 016-1VH2VL; 016-2VH2VL; 067-0VH0VL; 067-1VH0VL; 067-0VH1VL; 067-1VH1VL; 067-2VH1VL; 067-1VH2VL; 072-0VH0VK; 072-1VH0VK; 072-2VH0VK; 072-3VH0VK; 072-4VH0VK; 105-0VH0VK; 105-0VH1VK; 105-1VH1VK; 105-1VH2VK; 157-0VHAVK; 157-0VH0VK (human monoclonal)This paperThis paper:Supplementary file 1Computationally-inferred human (subject QA255) monoclonal antibodies representing naive and intermediate lineage members of six ADCC-capable antibody lineages; RFADCC (1:100-1:500); BLI (8 ug mL-1)
Antibody (human, female)006-NGSVHNGSVL; 016-NGSVH2VL(human monoclonal)This paperThis paper:Supplementary file 1; BioProject:PRJNA639297NGS-sampled antibody chains paired into monoclonal human (subject QA255) antibodies; 016-2VL light chain was computationally-inferred and not directly NGS-sampled; RFADCC (1:100-1:500)
Antibody (human, female)006-mVHmVL; 016-mVHmVL; 067-mVHmVL; 072-mVHmVK; 105-mVHmVK; 157-mVHmVK (human monoclonal)This paper and PMIDs 30779811 and 26629541GenBank: MT791224-MT791235Functionally-isolated human (subject QA255) antibody chains; ‘mature’ antibodies are capable of ADCC function; some sequences were edited based on consensus NGS data as describe in this paper; RFADCC (1:100-1:500); BLI (8 ug mL-1)
Recombinant DNA reagent006-0VH; 006-0VL; 006-1VH; 006-1VL; 006-mVH; 006-mVL; 006-NGSVH; 006-NGSVL; 016-0VH; 016-0VL; 016-1VH; 016-1VL; 016-2VL; 016-2VH; 016-mVH; 016-mVL; 016-NGSVH; 067-0VH; 067-0VL; 067-1VH; 067-1VL; 067-2VH; 067-2VL; 067-mVH; 067-mVL; 072-0VH; 072-0VK; 072-1VH; 072-2VH; 072-3VH; 072-4VH; 072-mVH; 072-mVK; 105-0VH; 105-0VK; 105-1VK; 105-1VH; 105-2VK; 105-mVH; 105-mVK; 157-0VH; 157-AVK; 157-0VH; 157-0VK; 157-mVH; 157-mVKThis paper; GeneWizThis paper:Supplementary file 1Computationally-inferred or NGS-sampled human (subject QA255) IgG (VH), IgK (VK), and IgL (VL) antibody variable region sequences representing naïve, intermediate, and mature lineage members of six ADCC-capable antibody lineages
Recombinant DNA reagentHuman Igγ1 expression vectorPMID:17996249Addgene:80795
Recombinant DNA reagentHuman Igκ expression vectorPMID:17996249Addgene:80796
Recombinant DNA reagentHuman Igλ expression vectorPMID:17996249Addgene:99575
Sequence-based reagentNGS; NGS-sampled sequences; libraries; D-119; D462; D791; D1174; D1512This paperBioProject:PRJNA639297Longitudinal QA255 full-length antibody variable region IgG, IgM, IgK, and IgL sequences
Sequence-based reagentPrimersPMID:27525066; PMID:31097697see PMID:31097697Primers for amplification and sequencing of human IgM, IgG, IgK and IgL variable regions
Sequence-based reagentSmartNNNa; template switch adaptor primersPMID:27490633N/AAAGCAGUGGTAUCAACGCAGAGUNNNNUNNNNUNNNNUCTTrGrGrGrGrG
Peptide, recombinant proteingp120 (HIV strain BL035.W6M.C1 ‘BL035’)Immune Technology Corp.Cat#IT-001–115 p; GenBank:DQ208480.1BLI analyte (250 nM or 62.5 nM); RFADCC coat (1.5 µg per 10^5 cells)
Peptide, recombinant proteingp120 (HIV strain QA255.22P.A17 ‘QA255’)Cambridge BiologicsCat#01-01-1743; GenBank:MW383930RFADCC coat (1.5 µg per 10^5 cells)
Peptide, recombinant proteingp120 (HIV strainBG505.W6M.B1 ‘BG505’)Cambridge BiologicsCat#01-01-1028; GenBank:ABA61515.1RFADCC coat (1.5 µg per 10^5 cells)
Peptide, recombinant proteingp120 (HIV strain CAP210.2.00.E8 ‘CAP210’)Immune Technology Corp.Cat#IT-001-RC12p; GenBank:DQ435683.1RFADCC coat (1.5 µg per 10^5 cells)
Peptide, recombinant proteingp120 (HIV strain MK184.W0M.G3 ‘MK184’)Immune Technology Corp.Cat#IT-001–112 p; GenBank:DQ208487RFADCC coat (1.5 µg per 10^5 cells)
Peptide, recombinant proteingp120 (HIV strain SF162)Cambridge BiologicsCat#01-01-1063; GenBank:P19550.1RFADCC coat (1.5 µg per 10^5 cells)
Peptide, recombinant proteingp41 ectodomain (HIV strain C.ZA.1197MB)Immune Technology Corp.Cat#IT-001–0052 p; GenBank:AY463234.1BLI analyte (250 nM or 62.5 nM); RFADCC coat (1.5 µg per 10^5 cells)
Peptide, recombinant proteinPhage-DMS gp41 and V3 libraryPMID:33089110N/ADeep mutational scanning phage display library containing wildtype and mutant peptides that tile across the gp41 and V3 portions of three Envelope strains (BG505.W6M.C2, BF520.W14M.C2, and C.ZA.1197MB). Peptides are 31 amino acids long and contain every possible single amino acid mutation at the central position, with peptides overlapping by 30 amino acids.
Chemical compound, drugQ5 High-Fidelity Master MixNew England BioLabsCat#M0492S
Chemical compound, drugFreeStyle MaxThermo Fisher ScientificCat#16447500
Chemical compound, drugProtein G agarosePierceCat#20397
Chemical compound, drug293F FreeStyle Expression mediaInvitrogenCat#12338–026
Commercial assay or kitAllPrep DNA/RNA Mini KitQiagenCat#80204
Commercial assay or kitSMARTer RACE 5'/3' KitTakara Bio USACat#634858
Commercial assay or kitKAPA library quantification kitKapa BiosystemsCat#KK4824
Commercial assay or kit600-cycle MiSeq Reagent Kit v3IlluminaCat#MS-102–3003
Commercial assay or kitNextera XT 96-well index kitIlluminaCat#FC-131–1001
Commercial assay or kitAnti-human IgG Fc capture biosensorsPall ForteBioCat#18–5063
Software, algorithmFLASH v1.2.11PMID:21903629http://ccb.jhu.edu/software/FLASH/
Software, algorithmCutadapt 1.14 with Python 2.7.9PMID:23671333RRID:SCR_011841http://cutadapt.readthedocs.io/en/stable/
Software, algorithmFASTX toolkit 0.0.14Hannon Lab, Cold Spring HarborRRID:SCR_005534http://hannonlab.cshl.edu/fastx_toolkit/
Software, algorithmPartisPMID:26751373N/Ahttps://github.com/psathyrella/partis
Software, algorithmLinearhamPMID:32804924N/Ahttps://github.com/matsengrp/linearham
Software, algorithmFastTree 2PMID:20224823N/Ahttp://doi.org/10.1371/journal.pone.0009490.g003
Software, algorithmPrune.pyPMID:31097697N/Ahttps://github.com/matsengrp/cft/blob/master/bin/prune.py
Software, algorithmCFT (Clonal Family Tree)This paperN/Ahttps://github.com/matsengrp/cft/
Software, algorithmRevBayesPMID:27235697N/Ahttps://revbayes.github.io/; incorporated in Ecgtheow code
Software, algorithmEcgtheowPMID:31097697N/Ahttps://github.com/matsengrp/ecgtheow
Software, algorithmLocal BLAST for lineage-like NGS sequencesThis paperN/Ahttps://git.io/Je7Zp
Software, algorithmFlowJo v10TreeStarRRID:SCR_008520
Software, algorithmExcelMicrosoft OfficeRRID:SCR_016137
Software, algorithmForteBio’s Octet Software ‘Data Analysis 7.0’Pall ForteBioN/A
Software, algorithmPrism 8.0 cGraphPadRRID:SCR_002798
Software, algorithmRstudioRstudioRRID:SCR_000432

Human subject

Request a detailed protocol

Peripheral blood mononuclear cell (PBMC) samples were obtained between 1997 and 2002 from a female HIV-1 seroconverted subject, QA255, who was enrolled in a prospective cohort of HIV-1-negative high-risk women in Mombasa, Kenya (Lavreys et al., 2002). QA255 was 40 years old at the time of HIV infection (D0). Study participants were treated according to Kenyan National Guidelines; QA255 did not receive antiretrovirals at any point during the period in which samples were analyzed for this study. Antiretroviral therapy was offered to all participants in the Mombasa Cohort beginning in March 2004, with support from the President’s Emergency Plan for AIDS Relief. The infecting virus was clade A based on envelope sequence (Bosch et al., 2010). Approval to conduct this study was provided by the ethical review committees of the University of Nairobi Institutional Review Board, the Fred Hutchinson Cancer Research Center Institutional Review Board, and the University of Washington Institutional Review Board. Study participants provided written informed consent prior to enrollment.

Sample handling and RNA isolation

Request a detailed protocol

PBMCs stored in liquid nitrogen were thawed at 37°C, diluted 10-fold in pre-warmed RPMI and centrifuged for 10 min at 300 x g. Cells were washed once in phosphate-buffered saline, counted with trypan blue, centrifuged again, and total RNA was extracted from PBMCs using the AllPrep DNA/RNA Mini Kit (Qiagen, Germantown, MD), according to the manufacturer’s recommended protocol. RNA was stored at −80°C.

Sequencing of full-length antibody gene variable regions

Request a detailed protocol

Antibody sequencing was performed as previously described (Simonich et al., 2019; Vigdorovich et al., 2016). Library preparation was performed in technical replicate, as indicated in Table 1, using the same RNA isolated from each timepoint: D-119 (119 days prior to HIV infection), D462, D791, D1174, and D1512 post-infection. Briefly, RACE-ready cDNA synthesis was performed using the SMARTer RACE 5’/3’ Kit (Takara Bio USA, Inc, Mountain View, CA) using primers with specificity to IgM, IgG, IgK, and IgL, as previously reported (Simonich et al., 2019). One replicate each for D791 and D1174 were prepared using template switch adaptor primers that included unique molecular identifiers in the cDNA synthesis step: SmartNNNa 5’ AAGCAGUGGTAUCAACGCAGAGUNNNNUNNNNUNNNNUCTTrGrGrGrGrG 3’(Turchaninova et al., 2016), where ‘rG’ indicates (ribonucleoside) guanosine bases. cDNA was diluted in Tricine-EDTA according to the manufacturer’s recommended protocol. First-round Ig-encoding sequence amplification (20 cycles) was performed using Q5 High-Fidelity Master Mix (New England BioLabs, Ipswich, MA) and nested gene-specific primers. Amplicons were directly used as templates for MiSeq adaption by second-round PCR amplification (10–20 cycles), purified and analyzed by gel electrophoresis, and indexed using Nextera XT P5 and P7 index sequences (Illumina, San Diego, CA) for Illumina sequencing according to the manufacturer’s instructions (10 cycles). Gel-purified, indexed libraries were quantitated using the KAPA library quantification kit (Kapa Biosystems, Wilmington, MA) performed on an Applied Biosystems 7500 Fast real-time PCR machine. Libraries were denatured and loaded onto Illumina MiSeq 600-cycle V3 cartridges, according to the manufacturer’s suggested workflow.

Sequence analysis and naïve inference

Request a detailed protocol

Sequences were preprocessed using FLASH, cutadapt, and FASTX-toolkit as previously described (Simonich et al., 2019; Vigdorovich et al., 2016). The sequences from our NGS replicates were merged either cumulatively or on a per-timepoint basis, as appropriate, to achieve the highest depth possible for each analysis. Sequences were then deduplicated and annotated with partis (https://github.com/psathyrella/partis) using default options including per-sample germline inference (Ralph and Matsen, 2016a; Ralph and Matsen, 2016b; Ralph and Matsen, 2019; ). Sequences with internal stop codons or CDR3 regions that were out-of-frame or had mutated codons at the start or end were removed. Indel events were identified, tracked, and reversed for alignment purposes. We did not exclude singletons in an attempt to retain very rare or undersampled sequences. Sequencing run statistics are detailed in Table 1. Cumulative and per-timepoint datasets underwent clonal family clustering using both the partis unseeded and seeded clustering methods (Ralph and Matsen, 2016b).

Inference of unmutated common ancestor sequences and simultaneous clonal family clustering was performed using the seed clustering method along with previously-identified QA255 mature ADCC antibody sequences (Williams et al., 2015; Williams et al., 2019) as ‘seeds’. As an additional measure to ensure highest accuracy for inferred naïve sequences, the uncertainty on each inferred naive sequence was visualized both with the partis --view-alternative-annotations option and by comparing these results to the most likely naive sequences inferred by linearham software (see next section). The comparison revealed only the minor differences that would be expected based on linearham’s method, which uses a more detailed model to infer more accurate naive sequences for individual clonal families. For the unseeded clustering, each dataset was subsampled to 50–150K sequences for computational efficiency. For each dataset, three random subsamples were analyzed and compared to ensure that our subsampling was sufficient to minimize statistical uncertainties. Seeded analyses were not subsampled.

Due to the lack of D genes and much shorter non-templated regions in light-chain rearrangements, computational clustering analyses artifactually overestimate clonality in light chain families, that is they cluster together sequences that did not originate from the same rearrangement event, but which come from almost identical naïve rearrangements. This caveat, affecting all unpaired antibody chain sequencing studies, ultimately reduces accuracy in inferring true light chain clonal families and their maturation pathways.

Alternative naive inference

Request a detailed protocol

Alternative naïve sequences were inferred by applying a Bayesian phylogenetic Hidden Markov Model approach using the linearham software (https://github.com/matsengrp/linearham), using the partis-inferred clonal family clusters for each lineage as input (Dhar et al., 2020). Linearham samples naive sequences from their posterior distribution rather than providing a single naive sequence estimate, like other software programs do. The most probable linearham-predicted naïve sequences were compared to the most probable partis-predicted naïve sequences. In families with largely symmetric phylogenetic trees, the two methods return similar results, whereas with highly imbalanced trees, linearham is much more accurate.

Antibody lineage reconstruction

Request a detailed protocol

Antibody lineages were inferred as previously described (Simonich et al., 2019). In order to subsample large clonal families according to phylogenetic relatedness to the antibody chain of interest, initial phylogenetic trees of each QA255 clonal family was inferred using FastTree 2 (Price et al., 2010). This allowed clonal families to be reduced to the 75–100 sequences most relevant to inferring the lineage history of the antibody chain of interest (https://github.com/matsengrp/cft/blob/master/bin/prune.py), which was then analyzed with RevBayes (Höhna et al., 2016) using an unrooted tree model with the general time-reversible (GTR) substitution model. All settings for RevBayes runs were customized to ensure likelihood and estimated sample sizes were >100 for each lineage. MCMC iterations ranged from 10,000 to 200,000, with thinning frequencies between 10 and 200 iterations and number of burn-in samples between 10 and 190. All RevBayes runs were done in technical duplicate (i.e. specifying different starting trees) and duplicates were all confirmed to agree on lineage chronology. RevBayes output was summarized for internal node sequences (https://github.com/matsengrp/ecgtheow), resulting in summary graphics where relative confidence in unique inferred sequences and amino acid substitutions are represented by color intensity (Figure 4—figure supplements 13). For each lineage, inferred intermediate sequences found on the most probable lineage paths were selected for study (Supplementary file 1, Figure 4—figure supplements 13).

In the 016-VL lineage, we determined that a 3-nt insertion event occurred prior to the inferred_7_712 intermediate within this antibody’s most likely developmental pathway (Figure 4—figure supplement 1). Since most insertion-containing clonal NGS sequences (136 of 142) encoded serine at the insertion site (amino acid position 94), we inserted S94 into 016-2VL instead of N94 that would correspond to the mature 016-VL sequence. For thoroughness, the inferred_5_789 016-VL intermediate (which precedes inferred_7_712) was synthesized both without (016-1VL) and with (016-2VL) the S94 insertion.

Identification of lineage-like NGS sequences

Request a detailed protocol

To determine if the computationally-inferred naïve and ancestor sequences were observed in the NGS data, we performed the following procedure for each lineage (implemented in a script found here: https://git.io/Je7Zp). A local BLAST database was created for each seeded clonal family and queried for sampled sequences that had high nucleotide sequence identity to lineage members using the ‘blastn’ command (Biopython package). E-value of 0.001 was used; other settings were default. Blastn matches for each lineage member were sorted according to their percent nt identity and alignment length. Sampled sequences with 100% nt or 100% aa identity in common with lineage members were noted (Supplementary file 1, Figure 4—figure supplements 13). To identify the closest sampled sequences to each VH mature sequence, blastn results per VH mature query were viewed and the highest percent aa identity match was selected.

ADCC-focused lineage determination

Request a detailed protocol

As outlined in our Results, we first selected ‘middle’ lineage intermediates for studying ADCC gain-of-function based on their moderate inclusion of amino acid substitutions (between naïve and mature sequences; Supplementary file 1) and high statistical confidence (Figure 4—figure supplements 13), along with, in some cases, their validation by sampled NGS sequences (Figure 4—figure supplements 13), and/or concentration of substitutions in complementarity-determining regions (CDRs) (Supplementary file 1). Middle intermediate chains were paired with partner naïve, middle, and mature chains and tested for antigen binding and ADCC function. Based on preliminary results using middle intermediates, ADCC-focused lineages were chosen from remaining pre-middle inferred intermediates or post-middle intermediates depending on whether the middle intermediate chain contributed to ADCC activity. If multiple relevant intermediate choices were available within the pertinent (early or late) portion of an inferred lineage (Figure 4—figure supplements 13), we selected sequences that had amino acid substitutions that were concentrated in CDRs (Supplementary file 1) and, whenever possible, we selected intermediates that were validated by NGS-sampled sequences, as aforementioned (Figure 4—figure supplements 13). For 067-VH, 067-VK, 072-VH, and 105-VH lineages, early mutations were implicated in ADCC gain-of-function, but we lacked early lineage resolution and therefore could not select early inferred intermediates to study. Instead, we selectively incorporated CDR substitutions from the middle intermediate sequence into the inferred naïve sequence. Non-conservative FWR3 substitutions were also incorporated into the 067-VH early lineage. For each lineage, intermediate sequences were numbered consecutively based on their position in the developmental pathway between naïve (0) and mature (m). Heavy and light chain pairings for lineage intermediates were based on chronology, which reflected levels of mutation. Each chain was paired with several partner chains for functional assessment.

Ultimately, developmental lineages featuring the minimal number of mAbs to study ADCC development were defined based on levels of mAb mutation and ADCC activity. For each lineage, the minimal antibodies included (1) the inferred naïve, (2) the most mutated intermediate(s) that lacked ADCC activity, if available, (3) the least mutated intermediate(s) that gained detectable ADCC function, and (4) the least mutated intermediate(s) that demonstrated ADCC activity comparable to the mature antibodies (>50%). Mature antibodies were always included as benchmark positive controls.

Cell lines

Request a detailed protocol

For antibody production: HEK 293 F cells (RRID:CVCL_D603; originally derived from female human embryonic kidney cells) were obtained from Invitrogen (Thermo Fisher Scientific, Waltham, MA, catalog #R790-07) and grown at 37°C in Freestyle 293 Expression Medium (Thermo Fisher Scientific, catalog #12338002) in baffle-bottomed flasks orbiting at 135 rpm. These cells were not further authenticated in our hands.

For RFADCC: CEM.NKR cells (RRID:CVCL_X622; originally derived from female human T-lymphoblastoid cells) were obtained from NIH AIDS Reagent Program (Germantown, MD, catalog #458) and grown at 37°C in RPMI 1640 media with added penicillin (100 U/mL), streptomycin (100 µg/mL), Amphotericin B (250 ng/mL), L-glutamine (2 mM), and fetal bovine serum (10%) (all from Thermo Fisher Scientific). These cells were not further authenticated in our hands.

Monoclonal antibody production

Request a detailed protocol

Antibody heavy and lightchain variable regions were synthesized as FragmentGENES (GENEWIZ, South Plainfield, NJ) and subsequently cloned into corresponding Igγ1, Igκ or Igλ expression vectors (Tiller et al., 2008). Equal ratios of heavy and light chain plasmids were co-transfected into HEK 293 F cells using FreeStyle MAX (Thermo Fisher Scientific, catalog #16447100) according to the manufacturer’s instructions. Column-based Pierce protein G (Thermo Fisher Scientific, catalog #20397) purification of IgG was done according to the manufacturer’s instructions.

Rapid and fluorometric ADCC (RFADCC) assay

Request a detailed protocol

The RFADCC assay was performed as described (Gómez-Román et al., 2006; Williams et al., 2019). In short, CEM-NKr cells (NIH AIDS Reagent Program, catalog #458) were double-labeled with PKH-26-cell membrane dye (Sigma-Aldrich, St. Louis, MO) and a cytoplasmic-staining dye (Vybrant CFDA SE Cell Tracer Kit, Thermo Fisher Scientific). The double-labeled cells were coated with clade A gp120 (BL035.W6M.Env.C1, Wu et al., 2006) or clade C gp41 ectodomain (C.ZA.1197MB) (Rousseau et al., 2006) for 1 hr at room temperature at a ratio of 1.5 μg protein: 1 × 105 double-stained target cells. For RFADCC breadth assays, additional HIV gp120 monomers of the following strains were: QA255.22P.A17, BG505.W6M.B1, CAP210.2.00.E8, MK184.W0M.G3, and SF162. All gp120 and gp41 proteins were sourced from Immune Technology Corp, New York, NY, or Cambridge Biologics, Brookline, MA, as specified in the Key Resources Table. Coated targets were washed once with complete RMPI media (RPMI supplemented with 10% FBS, 4.0 mM Glutamax, and 1% antibiotic-antimycotic, all from Thermo Fisher Scientific). Monoclonal antibodies were diluted in complete RPMI media to a concentration of 100–500 ng/mL and mixed with 5 × 103 coated target cells for 10 min at room temperature. PBMCs (peripheral blood mononuclear cells; Bloodworks Northwest, Seattle, WA) from an HIV-negative donor were added at a ratio of 50 effector cells: one target cell. The coated target cells, antibodies, and effector cells were co-cultured for 4 hr at 37°C then fixed in 1% paraformaldehyde (Affymetrix, Santa Clara, CA). Cells were analyzed by flow cytometry (Symphony I/II, BD Biosciences, San Jose, CA) and ADCC activity was defined as the percent of PKH-26+ CFDA- cells after background subtraction, where background (antibody-mediated killing of uncoated cells) was standardized to be 3–5% as analyzed using FlowJo software (FlowJo LLC, Ashland, OR). To mitigate differences in activity observed with different PBMC donors and between experiments, ADCC activity for each sample was normalized to monoclonal positive control mAbs: 167-D (NIH AIDS Reagent Program, Cat #11681) for gp41-targeted Abs and C11 (NIH AIDS Reagent Program, Cat #7374) for gp120-targeted Abs. The activity of an unrelated antibody, FI6v3, that recognizes influenza hemagglutinin protein was used to define the limit of detection. For each replicate experiment, samples were categorized in the following manner: positive (sample >2*FI6v3), indeterminate (1*FI6v3 ≤ sample ≤2*FI6v3), negative (sample <1*FI6v3). Experiments were excluded if FI6v3 signal was >10% of monoclonal positive control signal. Background (uncoated cells) and negative-control (1*FI6v3) signal was subtracted from each sample’s activity and the resultant values were averaged across experimental replicates, normalized to the respective lineage’s mature mAb activity, and plotted in Prism v8.0c (GraphPad, San Diego, CA). A designation of indeterminate was also assigned in cases where samples were negative in the majority of experimental replicates, but indeterminate or positive in any replicate(s). In such cases, average activity was reported as usual.

Biolayer interferometry

Request a detailed protocol

QA255 monoclonal antibody binding to monomeric gp120 or gp41 was measured using biolayer interferometry on an Octet RED instrument (ForteBio, Fremont, CA). Antibodies diluted to 8 µg mL−1 in a filtered buffer solution of 1X PBS containing 0.1% BSA, 0.005% Tween-20, and 0.02% sodium azide were immobilized onto anti-human IgG Fc capture biosensors (ForteBio). C.ZA.1197MB or BL035.W6M.C1 gp41 was diluted to 250 nM, or as indicated, in the same buffer (above) and a series of up to six, two-fold dilutions were tested as analytes in solution at 30°C. The kinetics of mAb binding were measured as follows: association was monitored for 180 s, dissociation was monitored for 180 s, and regeneration was performed in 10 mM Glycine HCl (pH 1.5). For experiments in which dilution series were run, binding-affinity constants (KD; on-rate, Kon; off-rate, Kdis) were calculated using ForteBio’s Data Analysis Software 7.0. Responses (nanometer shift) were calculated and background-subtracted using double referencing against the buffer reference signal and non-specific binding of biosensor to analyte. Data were processed by Savitzky-Golay filtering prior to fitting using a 1:1 model of binding kinetics.

Epitope mapping with phage-DMS

Request a detailed protocol

Profiling of escape mutations was done as previously described (Garrett et al., 2020). We utilized a deep mutational scanning (DMS) phage display library containing wildtype and mutant peptides that tile across the gp41 and V3 portions of three Envelope strains (BG505.W6M.C2, BF520.W14M.C2, and C.ZA.1197MB). Peptides in the library are 31 amino acids long and contain every possible single amino acid mutation at the central position, with peptides overlapping by 30 amino acids. Immunoprecipitation was performed in technical duplicate by incubating 10 ng of antibody sample with 1 mL of gp41/V3 Phage-DMS library at a concentration representing 200,000 pfu/mL of each unique clone. Antibody and phage were allowed to form complexes by rocking overnight at 4°C in 96-deep-well plates that had been pre-blocked with 3% BSA in TBST. To isolate phage-antibody complexes, 20 uL each of Protein A and Protein G Dynabeads were added and incubated at 4°C for 4 hr. Beads were magnetically separated, washed 3x with 400 µL wash buffer (150 mM NaCl, 50 mM Tris-HCl, 0.1% [vol/vol] NP-40, pH 7.5), resuspended in 40 µL of dH2O before lysis (95°C for 10 min), and stored at −20°C. PCR and deep sequencing of the enriched phage was done as previously described. Briefly, sequences were amplified to add Illumina barcodes (two rounds) and then pooled and sequenced on an Illumina MiSeq with 1 × 125 bp single-end reads (BioProject accession PRJNA685289). Enrichment and scaled differential selection were calculated as previously described, with analysis performed and plots generated using RStudio (Figure 9—source code 1, Figure 9—source datas 13). Two biological replicates were run, using the same antibody preps but distinct peptide library preps.

Data and code availability

Request a detailed protocol

The QA255 longitudinal antibody deep sequencing datasets (Table 1) and Phage-DMS sequencing data sets generated during this study are publicly available at BioProject SRA, accessions PRJNA639297 [https://www.ncbi.nlm.nih.gov/bioproject/PRJNA639297/] and PRJNA685289 [https://www.ncbi.nlm.nih.gov/bioproject/PRJNA685289]. The inferred antibody variable region sequences generated in this study have not been deposited in GenBank because computationally-inferred sequences are not accepted, but they are available in Supplementary file 1. GenBank accession numbers for mature QA255 antibody are MT791224-MT791235 and QA255 Envelope sequences are MW383929-MW383956. GenBank accession numbers for HIV Env variants are available in the Key Resources Table. The custom code generated or used in this study for antibody lineage determination is publicly available on GitHub: prune.py (https://github.com/matsengrp/cft/blob/master/bin/prune.py), ecgtheow (https://github.com/matsengrp/ecgtheow), CFT (https://github.com/matsengrp/cft), and Blast validation (https://git.io/Je7Zp). Code for Phage-DMS analysis is provided as Figure 9—source code 1.

Quantification and statistical analysis

Request a detailed protocol

Where applicable, raw data were normalized and/or averaged across replicates using Microsoft Excel. Plots were generated using GraphPad Prism version 8.0 c. Relevant experimental details, such as use of biological and technical replicates, can be found in figure legends. For RFADCC plots, bars with error bars represent mean and SD. RFADCC experimental exclusion criteria are detailed within the RFADCC method section above. For Phage-DMS, calculations were performed as previously described (Garrett et al., 2020), using RStudio; Phage-DMS analysis source code is available in Figure 9—source code 1.

Data availability

Sequencing data have been deposited in BioProject SRA under the accession codes PRJNA639297 and PRJNA685289. Data generated and analyzed in this study are included in the manuscript and supporting files. Source data files have been provided for Figures 1, 2, 4, 5, 7, and 8.

The following data sets were generated
    1. Doepker L
    2. Overbaugh J
    (2020) NCBI BioProject
    ID PRJNA639297. Subject QA255 antibody sequencing.
    1. Garrett M
    2. Overbaugh J
    (2020) NCBI BioProject
    ID PRJNA685289. Development of antibody-dependent cell cytotoxicity function in HIV-1 antibodies.

References

Decision letter

  1. Satyajit Rath
    Senior and Reviewing Editor; Indian Institute of Science Education and Research (IISER), India
  2. Stephen Kent
    Reviewer; The University of Melbourne, Australia
  3. Felix Horns
    Reviewer; Stanford University, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

The manuscript uses a time series of bleeds from a single HIV-infected individual with low neutralising but high antibody-dependent cellular cytotoxicity (ADCC) antibody activity to identify six temporal lineages culminating in six distinct ADCC-capable antibodies. Using variable region repertoire sequencing, computational sequence reconstruction of antibody intermediates, and measurements of affinity and ADCC efficiency, the authors provide insights into how binding affinity and ADCC capability likely change during antibody clone diversification over time. Similar to studies that have investigated HIV-specific neutralising antibodies, this analysis provides clues to the evolutionary pathways of ADCC-functional antibody development.

Decision letter after peer review:

Thank you for submitting your article "Development of antibody-dependent cell cytotoxicity function in HIV-1 antibodies" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by Satyajit Rath as the Senior and Reviewing Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Stephen Kent (Reviewer #1); Felix Horns (Reviewer #3).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

We would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). Specifically, when editors judge that a submitted work as a whole belongs in eLife but that some conclusions require a modest amount of additional new data, as they do with your paper, we are asking that the manuscript be revised to either limit claims to those supported by data in hand, or to explicitly state that the relevant conclusions require additional supporting data.

Our expectation is that the authors will eventually carry out the additional experiments and report on how they affect the relevant conclusions either in a preprint on bioRxiv or medRxiv, or if appropriate, as a Research Advance in eLife, either of which would be linked to the original paper.

Summary:

The manuscript analyses the origins and development of six HIV-specific ADCC-mediating antibodies from a single HIV clade A-infected subject over time, beginning from seroconversion to 4 years post-infection. Using NGS-based variable region repertoire sequencing, computational sequence reconstruction of antibody intermediates, and measurements of affinity and ADCC efficiency, the authors provide insights into how binding affinity and ADCC capability likely change during antibody clone diversification over time. Similar to studies that have investigated HIV-specific neutralising antibodies, this analysis provides clues to the evolutionary pathways of ADCC-functional antibody development. While antibody binding was correlated with ADCC activity, antibodies with similar affinity exhibited ADCC activity with different potencies. Mutations in complementary determining regions were needed for ADCC activity, with additional mutations in framework regions generally providing enhanced potency. The study was rigorously conducted, and addresses an important question with well-supported conclusions. Some issues that need to be addressed are identified below.

Essential revisions:

1) The ordering of mutations is not fully resolved and thus some claims about ordering are not well supported. As an illustrative example, Figure 3A shows that several mutations were acquired between 0VH0VL and 1VH0VL that improved both ADCC and affinity. The ordering of these mutations and thus the ordering of improvements to ADCC and affinity are not resolved. Similarly, further mutations in 1VH1VL dramatically improved both ADCC and affinity. But again the ordering is not resolved. Similar arguments can be made for most of the antibody clones in Figure 3 and Figure 4. Thus, it seems misleading to claim that mutations first improve affinity, then later mutations improve ADCC, as implied in the Abstract. Several places elsewhere in the text also imply that the ordering of mutations has been resolved (e.g. "Thereafter, six light chain mutations among CDRL1, CDRL3, and FWRL2 augmented gp41 binding and ADCC potency in 006-1VH1VL"). Is there any evidence supporting this ordering? In the absence of evidence, it seems quite likely that these mutations happened in an interlaced fashion. This should be clarified in the text.

2) The fidelity of pairing between heavy and light chains is a key issue and should be clarified. In this study, the heavy and light chain sequences are reconstructed separately based on the mature antibody sequences. The experimental results hinge significantly on correct identification of corresponding heavy and light chains. The results would be more easily interpreted and perhaps better supported if the authors could estimate the fidelity of this strategy of computational pairing, possibly using a "ground truth" from paired heavy-light chain sequencing data sets. Otherwise, the authors should acknowledge this caveat, ideally in the main text.

3) Again, because the authors have reconstructed the mutation histories of heavy and light chains separately, we have no direct information about the co-occurrence of heavy and light chain mutations. Thus, the pairing of heavy and light chain sequences and the mutations they carry is arbitrary and may not represent actual combinations of mutations that were present together in a true intermediate. Ideally, the authors would validate or estimate the fidelity of the relative timing or co-occurrence of the mutations. This could perhaps be achieved using a paired heavy-light chain sequencing data set, as above. Otherwise, the authors should clarify in the text that the reconstructed intermediates (e.g. 1VH 1VL, 1VH 2VL, and 2VH 2VL in Figure 3B) are arbitrarily paired and may not represent the combinations of mutations in true intermediates.

4) A major concern is the lack of analysis of the native antibody Fc region over time given the central nature of this part of the antibody to ADCC functions. It may that there are no changes or no insights that can be drawn from this. However that would be an important finding in of itself and show the importance of the Fab analyses. Such an analysis would make the manuscript much stronger.

5) The first time point sampled is quite late at over 400 days. Comments about putative changes earlier would be useful.

6) There is interest in potential escape from ADCC antibodies as an argument about the pressure applied by these responses – with their evolution studies and knowledge of the epitopes the authors may be able to draw inferences on ADCC escape, and comments regarding this issue would be useful.

7) It would strengthen the argument to include assessment of the breadth of ADCC activity and how it fits into the evolutionary pathways.

https://doi.org/10.7554/eLife.63444.sa1

Author response

Essential revisions:

1) The ordering of mutations is not fully resolved and thus some claims about ordering are not well supported. As an illustrative example, Figure 3A shows that several mutations were acquired between 0VH0VL and 1VH0VL that improved both ADCC and affinity. The ordering of these mutations and thus the ordering of improvements to ADCC and affinity are not resolved. Similarly, further mutations in 1VH1VL dramatically improved both ADCC and affinity. But again the ordering is not resolved. Similar arguments can be made for most of the antibody clones in Figure 3 and Figure 4. Thus, it seems misleading to claim that mutations first improve affinity, then later mutations improve ADCC, as implied in the Abstract. Several places elsewhere in the text also imply that the ordering of mutations has been resolved (e.g. "Thereafter, six light chain mutations among CDRL1, CDRL3, and FWRL2 augmented gp41 binding and ADCC potency in 006-1VH1VL"). Is there any evidence supporting this ordering? In the absence of evidence, it seems quite likely that these mutations happened in an interlaced fashion. This should be clarified in the text.

To address points 1, 2, 3 and 5, that raise concerns about claims of developmental chronology and pairing choices amongst heavy and light chain intermediates, we’d like to offer clearer explanations of our approaches (here) as well as amendments to the text based (listed per point).

For longitudinal antibody sequencing, we prioritized depth rather than per-sequence accuracy because we already had “ground truth” pairing information from our six biologically-isolated ADCC antibodies and our focus was to infer likely ancestral sequences (nodes) between naïve and mature sequences within each clonal phylogeny. Our choices of sequencing and analysis methods were also driven by the characteristics of our unique, precious, and old samples: they derive from the pre-ART time period, which is important but means they have been in -80°C storage for ~20 years. With more abundant samples, we would have been eager to perform multiple sequencing approaches, such as both bulk unpaired and paired sequencing, but that was not the case here.

To achieve the highest depth possible with our longitudinal samples, we performed unpaired bulk sequencing with a well-established protocol rather than paired sequencing, which severely curtails depth and was not a well-practiced approach at the time this study began. As described and validated in a previous study of ours (Simonich et al., 2019), we designed our analysis to handle the challenges coming from sparsely-sampled noisy data. Specifically, we used Bayesian phylogenetics to infer ancestral sequences while allowing for uncertainty in the tree, rather than using ancestral sequence reconstructions from a single maximum-likelihood tree, which does not acknowledge tree uncertainty that can be quite significant. The chronology (ordering) of these ancestral sequences is resolved as clearly as the sequencing data and analyses allow; some pathways are resolved to 1-2 AA substitutions per ordered step, while others lack fine resolution and big leaps containing many AA substitutions occur between inferred ancestral sequences. The resolution of each lineage is viewable in Figure 4—figure supplements 1-3, with a main text example now included as Figure 4.

A strength of our approach is that we developed specific tools to enable confident reconstruction of the ordering of ancestral sequences (and, thus, the mutations occurring between them) even though our biological samples are quite old. In other words, we did not use off-the-shelf computational methods developed for other biological problems and adapt them for antibody sequence studies. This is exciting, indeed; if we had had more samples, we wouldn’t have had to work so hard to this end. All in all, with the bulk longitudinal sequencing data we acquired, we made as many confident conclusions about which mutations came before others as we could and those intermediate sequences are what we paired together into mAbs and tested.

Since this approach inferred the most likely development routes taken by individual antibody chains (heavy and light, independently), our inferred ancestral sequences are indeed unpaired, as emphasized by the reviewers’ points 1, 2, and 3. As outlined below, we added text to more explicitly explain that our experimental pairings do not reflect true biological intermediates. Indeed, we did not set out to define with certainty the paired intermediates that led to antibody ADCC function. Rather, our goal was to define biologically-informed minimal mutations (ideally in the form of a highly-likely computationally inferred ancestral sequence) that are needed for antibody lineages to gain ADCC function. Ultimately, we used the sequencing data and computational analysis to learn what is probable in terms of antibody development over time and, in the end, the findings from this approach are what drive this story – namely that both CDR and FWR mutations are needed to achieve potent ADCC and binding affinity is not exclusively responsible for ADCC potency. Moreover, this is the first study of stepwise ADCC antibody development ever, and we find it frankly remarkable that we found themes among the six lineages under investigation.

Changes to the text in response to points 1, 2, and 3:

– Clarified explanations of our methods which determined the chronology of highly probable ancestral sequences within single chains (heavy or light), though not each individual mutation per chain (Introduction, Results, Discussion and Materials and methods).

– Amended text that was inaccurately implying chronology between heavy and light chain mutations (Abstract, Results and Discussion).

2) The fidelity of pairing between heavy and light chains is a key issue and should be clarified. In this study, the heavy and light chain sequences are reconstructed separately based on the mature antibody sequences. The experimental results hinge significantly on correct identification of corresponding heavy and light chains. The results would be more easily interpreted and perhaps better supported if the authors could estimate the fidelity of this strategy of computational pairing, possibly using a "ground truth" from paired heavy-light chain sequencing data sets. Otherwise, the authors should acknowledge this caveat, ideally in the main text.

These concerns are addressed above in response to points 1, 2, 3, and 5, including the list of changes to the text.

3) Again, because the authors have reconstructed the mutation histories of heavy and light chains separately, we have no direct information about the co-occurrence of heavy and light chain mutations. Thus, the pairing of heavy and light chain sequences and the mutations they carry is arbitrary and may not represent actual combinations of mutations that were present together in a true intermediate. Ideally, the authors would validate or estimate the fidelity of the relative timing or co-occurrence of the mutations. This could perhaps be achieved using a paired heavy-light chain sequencing data set, as above. Otherwise, the authors should clarify in the text that the reconstructed intermediates (e.g. 1VH 1VL, 1VH 2VL, and 2VH 2VL in Figure 3B) are arbitrarily paired and may not represent the combinations of mutations in true intermediates.

These concerns are addressed above in response to points 1, 2, 3, and 5, including the list of changes to the text.

4) A major concern is the lack of analysis of the native antibody Fc region over time given the central nature of this part of the antibody to ADCC functions. It may that there are no changes or no insights that can be drawn from this. However that would be an important finding in of itself and show the importance of the Fab analyses. Such an analysis would make the manuscript much stronger.

Based on literature in this field, the Fc region of antibodies (actually the whole constant region, part of which is included in the Fab portion), remains stable in human antibodies. Specifically, the constant region in antibodies lacks mutations because the AID (activation-induced cytidine deaminase) enzyme responsible for somatic hypermutation does not gain access to the constant regions of Ig genes (Longerich et al., 2005; Peled et al., 2008) even though AID is required for class switch recombination, which takes place upstream of the 5’ ends of each constant heavy chain gene (Muramatsu et al., 2000; Stavnezer, Guikema and Schrader, 2008). With these notions in mind, we chose to perform full length variable region sequencing, which included enough sequence into the constant region to determine antibody isotype. This approach allowed us to determine the effect of AID-mediated somatic hypermutation on ADCC development in antibody lineages and, indeed, our data demonstrate that variable region mutations affect ADCC function even though ADCC activity is an Fc-mediated process.

However, we hear and completely agree with your point that ADCC activity is Fc-mediated and information about the Fc portions of these antibodies is of interest (sequence, binding angle, isotype and subclass). We have clarified that all these antibodies are IgG1 subclass in the Introduction. Despite the abovementioned dogma about constant regions remaining unmutated through antibody lineage evolution, we regret that we lack enough samples to fully sequence the constant regions of these lineages, because, if we found mutations in these regions, it would indeed be an important and extremely surprising and novel finding.

5) The first time point sampled is quite late at over 400 days. Comments about putative changes earlier would be useful.

In the above response to points 1, 2, 3, and 5, we clarify the extent of our abilities to predict ordered changes based on clonal antibody sequences from various time points. In response to this comment, specifically, we have added time point information to the NGS hits (green and yellow stars) within pathways presented in Figure 4—figure supplements 1-3. Thus, readers may infer which inferred ancestral sequences were possibly present at/after distinct time points post-infection. Though we cannot confidently dissect the order of mutations that occurred earlier than day 462 post-infection any further than we already have, we have added text in the Discussion to note that three VH clonal families (006, 067, and 105) contain only D462 NGS sequences (Table 2) and thus all developmental inferences for these three families likely occurred prior to D462 (Discussion).

6) There is interest in potential escape from ADCC antibodies as an argument about the pressure applied by these responses – with their evolution studies and knowledge of the epitopes the authors may be able to draw inferences on ADCC escape, and comments regarding this issue would be useful.

We have added data to the Results section on the (lack of) evidence for longitudinal mutational Env escape from lineages 067, 072, and 105, which all have linear peptide epitopes. The 067 and 072 epitopes are entirely stable between 21 dpi and 1729 dpi. For lineage 105, we finely mapped the epitope of both the inferred naïve mAb and the mature mAb, finding that the residue positions comprising the epitope remain stable, and that there is little evidence for strong selective pressure on these residues between 21 dpi and 1729 dpi. In adding the additional section on 105 lineage epitope mapping, we have also added Meghan Garrett as an author of the paper for her contributions to this work. Unfortunately, lineages 006, 016, and 157 target discontinuous epitopes, so we cannot comment on escape here without significant additional experimentation.

Changes to the text: Results, Discussion, Materials and methods, Figure 9, Figure 9—figure supplements 1-2, Figure 9—source data 1, Figure 9—source data 1-3.

7) It would strengthen the argument to include assessment of the breadth of ADCC activity and how it fits into the evolutionary pathways.

We have added an assessment of developing ADCC breadth in the two gp120-targeted lineages as possible, given their mature breadth profiles (Williams et al., 2015). Specifically, we added data for RFADCC against two (lineage 105) or five (lineage 157) additional HIV strains, including the autologous transmitted founder virus (lineage 157 only). Changes to the text include the addition of Figure 6—figure supplement 1 and in the Results, Discussion and Materials and methods. The results indicate that full breadth developed after initial ADCC activity by both lineages. These data indeed strengthen the study; thank you for the suggestion. We maintain confidence in BL035 gp120 as the primary RFADCC coat protein for gp120-targeted antibodies in this study because we previously demonstrated that this strain is representative of RFADCC results among seven different gp120s of various clades (Milligan et al., 2015). This experimental choice is now clarified in the Discussion.

Unfortunately, RFADCC breadth studies are not possible for gp41-targeted lineages. There are very limited gp41 coat proteins available for experimental use and early attempts with MN gp41 yielded no RFADCC signal. As of yet, C.ZA.1197MB gp41 ectodomain is the only effective RFADCC coat protein available for gp41-targeted lineages.

References:

Longerich, S., Tanaka, A., Bozek, G., Nicolae, D., and Storb, U. (2005). The very 5' end and the constant region of Ig genes are spared from somatic mutation because AID does not access these regions. J Exp Med, 202(10), 1443-1454. doi:10.1084/jem.20051604Muramatsu, M., Kinoshita, K., Fagarasan, S., Yamada, S., Shinkai, Y., and Honjo, T. (2000). Class switch recombination and hypermutation require activation-induced cytidine deaminase (AID), a potential RNA editing enzyme. Cell, 102(5), 553-563. doi:10.1016/s0092-8674(00)00078-7Peled, J. U., Kuang, F. L., Iglesias-Ussel, M. D., Roa, S., Kalis, S. L., Goodman, M. F., and Scharff, M. D. (2008). The biochemistry of somatic hypermutation. Annu Rev Immunol, 26, 481-511. doi:10.1146/annurev.immunol.26.021607.090236Stavnezer, J., Guikema, J. E., and Schrader, C. E. (2008). Mechanism and regulation of class switch recombination. Annu Rev Immunol, 26, 261-292. doi:10.1146/annurev.immunol.26.021607.090248

https://doi.org/10.7554/eLife.63444.sa2

Article and author information

Author details

  1. Laura E Doepker

    Human Biology Division, Fred Hutchinson Cancer Research Center, Seattle, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Supervision, Validation, Investigation, Visualization, Methodology, Writing - original draft, Project administration, Writing - review and editing
    Contributed equally with
    Sonja Danon
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-4514-5003
  2. Sonja Danon

    Human Biology Division, Fred Hutchinson Cancer Research Center, Seattle, United States
    Contribution
    Validation, Investigation, Visualization, Writing - original draft, Project administration, Writing - review and editing
    Contributed equally with
    Laura E Doepker
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5399-7081
  3. Elias Harkins

    Public Health Sciences Division, Fred Hutchinson Cancer Research Center, Seattle, United States
    Contribution
    Data curation, Software, Formal analysis, Validation, Visualization, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
  4. Duncan K Ralph

    Public Health Sciences Division, Fred Hutchinson Cancer Research Center, Seattle, United States
    Contribution
    Data curation, Software, Formal analysis, Validation, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
  5. Zak Yaffe

    1. Human Biology Division, Fred Hutchinson Cancer Research Center, Seattle, United States
    2. Medical Scientist Training Program, University of Washington School of Medicine, Seattle, United States
    Contribution
    Validation, Investigation, Writing - review and editing
    Competing interests
    No competing interests declared
  6. Meghan E Garrett

    1. Human Biology Division, Fred Hutchinson Cancer Research Center, Seattle, United States
    2. Molecular and Cellular Biology Graduate Program, University of Washington and Fred Hutchinson Cancer Research Center, Seattle, United States
    Contribution
    Software, Validation, Investigation, Visualization, Methodology, Writing - original draft
    Competing interests
    No competing interests declared
  7. Amrit Dhar

    1. Public Health Sciences Division, Fred Hutchinson Cancer Research Center, Seattle, United States
    2. Department of Statistics, University of Washington, Seattle, United States
    Contribution
    Data curation, Software, Formal analysis, Validation, Writing - review and editing
    Competing interests
    No competing interests declared
  8. Cassia Wagner

    Medical Scientist Training Program, University of Washington School of Medicine, Seattle, United States
    Contribution
    Formal analysis, Validation, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-9934-7578
  9. Megan M Stumpf

    Human Biology Division, Fred Hutchinson Cancer Research Center, Seattle, United States
    Contribution
    Formal analysis, Investigation, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-8085-3094
  10. Dana Arenz

    Human Biology Division, Fred Hutchinson Cancer Research Center, Seattle, United States
    Contribution
    Validation, Investigation, Writing - review and editing
    Competing interests
    No competing interests declared
  11. James A Williams

    Department of Medicinal Chemistry, University of Washington, Seattle, United States
    Contribution
    Supervision, Validation, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
  12. Walter Jaoko

    Department of Medicinal Microbiology, University of Nairobi, Nairobi, Kenya
    Contribution
    Resources, Data curation
    Competing interests
    No competing interests declared
  13. Kishor Mandaliya

    Coast Provincial General Hospital, Women’s Health Project, Mombasa, Kenya
    Contribution
    Resources, Data curation
    Competing interests
    No competing interests declared
  14. Kelly K Lee

    Department of Medicinal Chemistry, University of Washington, Seattle, United States
    Contribution
    Resources, Supervision, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
  15. Frederick A Matsen IV

    Public Health Sciences Division, Fred Hutchinson Cancer Research Center, Seattle, United States
    Contribution
    Resources, Software, Supervision, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0607-6025
  16. Julie M Overbaugh

    1. Human Biology Division, Fred Hutchinson Cancer Research Center, Seattle, United States
    2. Public Health Sciences Division, Fred Hutchinson Cancer Research Center, Seattle, United States
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition, Writing - original draft, Project administration, Writing - review and editing
    For correspondence
    joverbau@fredhutch.org
    Competing interests
    Reviewing editor, eLife
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-0239-9444

Funding

National Institutes of Health (R37 AI038518)

  • Julie M Overbaugh

National Institutes of Health (R01 HD103571)

  • Julie M Overbaugh

National Institutes of Health (R01 GM113246)

  • Frederick A Matsen IV

National Institutes of Health (R01 AI146028)

  • Frederick A Matsen IV

National Institutes of Health (T32 AI07140)

  • Laura E Doepker

National Institutes of Health (T32 AI083203)

  • Zak Yaffe

National Institutes of Health (P30 AI027757)

  • Duncan K Ralph

Howard Hughes Medical Institute (Faculty Scholar grant)

  • Frederick A Matsen IV

Simons Foundation (Faculty Scholar grant)

  • Frederick A Matsen IV

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank the participants, staff, Scott McClelland, and Ludo Lavreys for continued efforts to oversee the Mombasa cohort. We acknowledge Brian Oliver, Vladimir Vigdorovich, and D Noah Sather for technical assistance with NGS library preparation and sequence processing. We thank Vrasha Chohan, Haidyn Weight, and Tucker Price for their help with antibody production and testing.

Ethics

Human subjects: Approval to conduct this study was provided by the ethical review committees of the University of Nairobi Institutional Review Board, the Fred Hutchinson Cancer Research Center Institutional Review Board (protocol 7776), and the University of Washington Institutional Review Board; Clinical Trial Management System Number RG1000880. Study participants provided written informed consent prior to enrollment.

Senior and Reviewing Editor

  1. Satyajit Rath, Indian Institute of Science Education and Research (IISER), India

Reviewers

  1. Stephen Kent, The University of Melbourne, Australia
  2. Felix Horns, Stanford University, United States

Publication history

  1. Received: September 25, 2020
  2. Accepted: January 8, 2021
  3. Accepted Manuscript published: January 11, 2021 (version 1)
  4. Version of Record published: February 15, 2021 (version 2)

Copyright

© 2021, Doepker et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 699
    Page views
  • 99
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Immunology and Inflammation
    2. Medicine
    Daniela Fraccarollo et al.
    Research Article Updated

    Immature neutrophils and HLA-DRneg/low monocytes expand in cancer, autoimmune diseases and viral infections, but their appearance and immunoregulatory effects on T-cells after acute myocardial infarction (AMI) remain underexplored. We found an expansion of circulating immature CD16+CD66b+CD10neg neutrophils and CD14+HLA-DRneg/low monocytes in AMI patients, correlating with cardiac damage, function and levels of immune-inflammation markers. Immature CD10neg neutrophils expressed high amounts of MMP-9 and S100A9, and displayed resistance to apoptosis. Moreover, we found that increased frequency of CD10neg neutrophils and elevated circulating IFN-γ levels were linked, mainly in patients with expanded CD4+CD28null T-cells. Notably, the expansion of circulating CD4+CD28null T-cells was associated with cytomegalovirus (CMV) seropositivity. Using bioinformatic tools, we identified a tight relationship among the peripheral expansion of immature CD10neg neutrophils, CMV IgG titers, and circulating levels of IFN-γ and IL-12 in patients with AMI. At a mechanistic level, CD10neg neutrophils enhanced IFN-γ production by CD4+ T-cells through a contact-independent mechanism involving IL-12. In vitro experiments also highlighted that HLA-DRneg/low monocytes do not suppress T-cell proliferation but secrete high levels of pro-inflammatory cytokines after differentiation to macrophages and IFN-γ stimulation. Lastly, using a mouse model of AMI, we showed that immature neutrophils (CD11bposLy6GposCD101neg cells) are recruited to the injured myocardium and migrate to mediastinal lymph nodes shortly after reperfusion. In conclusion, immunoregulatory functions of CD10neg neutrophils play a dynamic role in mechanisms linking myeloid cell compartment dysregulation, Th1-type immune responses and inflammation after AMI.

    1. Genetics and Genomics
    2. Immunology and Inflammation
    Hyeogsun Kwon et al.
    Research Article

    Mosquito immune cells, known as hemocytes, are integral to cellular and humoral responses that limit pathogen survival and mediate immune priming. However, without reliable cell markers and genetic tools, studies of mosquito immune cells have been limited to morphological observations, leaving several aspects of their biology uncharacterized. Here, we use single-cell RNA sequencing (scRNA-seq) to characterize mosquito immune cells, demonstrating an increased complexity to previously defined prohemocyte, oenocytoid, and granulocyte subtypes. Through functional assays relying on phagocytosis, phagocyte depletion, and RNA-FISH experiments, we define markers to accurately distinguish immune cell subtypes and provide evidence for immune cell maturation and differentiation. In addition, gene-silencing experiments demonstrate the importance of lozenge in defining the mosquito oenocytoid cell fate. Together, our scRNA-seq analysis provides an important foundation for future studies of mosquito immune cell biology and a valuable resource for comparative invertebrate immunology.