common, as only a subset of CTCF sites form in-sulated neighborhoods (8, 10, 24). CTCF cohesin–bound loops are largely preserved across celltypes (8, 9, 24), and a set of ~10,000 constitutiveCTCF-CTCF loops shared by GM12878 lympho-blastoid, Jurkat, and K562 (CML) cells (24) wereidentified for comparison (Fig. 4A, fig. S11, andtable S8). We used the International Cancer Gen-ome Consortium (ICGC) database—which containsdata for ~50 cancer types, ~2300 whole-genomesequence (WGS) samples, and ~13million uniquesomatic mutations—to examine the boundariesof these neighborhoods for somatic point muta-tions found in cancer genomes (table S9). Wefound a striking enrichment of mutations at theCTCF boundaries of constitutive neighborhoods(Fig. 4B, fig. S12A, and table S10) relative to re-gions flanking the boundary CTCF sites (±1 kbof the CTCF bindingmotif; P < 10−4, permutationtest) (fig. S12B), and in many instances thesecreated a change in the consensus CTCF bindingmotif (fig. S12C). Nonboundary CTCF sites didnot show such enrichment (Fig. 4B and figs. S12Dand S14). The genomes of esophageal and livercarcinoma samples were particularly enrichedfor boundary CTCF site mutations (Fig. 4, C andD, fig. S12, D and E, fig. S13, and table S10), andthere was no similar enrichment of mutations atthe binding sites of other transcription factors(fig. S15). In these cancers, a considerable frac-tion of the mutated neighborhood boundaryCTCF sites were affected by multiple mutations(≥3 mutations per site) [280/1826 (15%) in esoph-ageal carcinoma, 54/1030 (5%) in liver carcinoma](table S10), and recurrent mutations occurredmore frequently inneighborhoodboundary CTCFsites relative to nonboundary CTCF sites (fig. S16,A to C). The genes located within the most fre-quently mutated neighborhoods included knowncellular proto-oncogenes annotated in the CancerGene Census and other genes that have not beenassociated with these cancers (Fig. 4, E and F,and tables S11 and S12). Shown in Fig. 4, G andH,are two examples of proto-oncogene–containingneighborhoods where the activation of the genelocated in the neighborhood has been observedin the respective cancer type. These results sug-gest that somatic mutations of insulated neigh-borhood boundaries occur in the genomes ofmany different cancers.Our findings indicate that disruption of in-

sulated neighborhood boundaries can cause on-cogene activation in cancer cells. With maps of3D chromosome structure such as those describedhere, cancer genome analysis can consider howrecurrent perturbations of boundary elementsmay affect the expression of genes with roles intumor biology. Our understanding of 3D chro-mosome structure and its control is rapidly ad-vancing and should be considered for potentialdiagnostic and therapeutic purposes. Because con-trol of 3D chromosome structure involves bind-ing of specific sites by CTCF and cohesin, whichis affected by protein cofactors, DNA methyla-tion, and local RNA synthesis (25), advances inour understanding of these regulatory processesmay provide new approaches to therapeutics

that have an impact on aberrant chromosomestructures.

REFERENCES AND NOTES

1. B. Vogelstein, K. W. Kinzler, Nat. Med. 10, 789–799(2004).

2. B. Vogelstein et al., Science 339, 1546–1558 (2013).3. L. A. Garraway, E. S. Lander, Cell 153, 17–37 (2013).4. C. M. Croce, N. Engl. J. Med. 358, 502–511 (2008).5. M. H. Kagey et al., Nature 467, 430–435 (2010).6. J. H. Gibcus, J. Dekker, Mol. Cell 49, 773–782 (2013).7. D. U. Gorkin, D. Leung, B. Ren, Cell Stem Cell 14, 762–775

(2014).8. J. M. Dowen et al., Cell 159, 374–387 (2014).9. J. E. Phillips-Cremins et al., Cell 153, 1281–1295

(2013).10. X. Ji et al., Cell Stem Cell 18, 262–275 (2016).11. J. R. Dixon et al., Nature 485, 376–380 (2012).12. E. P. Nora et al., Nature 485, 381–385 (2012).13. S. A. Armstrong, A. T. Look, J. Clin. Oncol. 23, 6306–6315 (2005).14. P. Van Vlierberghe, A. Ferrando, J. Clin. Invest. 122,

3398–3406 (2012).15. M. J. Fullwood et al., Nature 462, 58–64 (2009).16. Z. Tang et al., Cell 163, 1611–1627 (2015).17. D. Hnisz et al., Cell 155, 934–947 (2013).18. J. Lovén et al., Cell 153, 320–334 (2013).19. L. Brown et al., EMBO J. 9, 3343–3351 (1990).20. J. O’Neil, A. T. Look, Oncogene 26, 6838–6849 (2007).21. P. Van Vlierberghe et al., Blood 108, 3520–3529 (2006).22. J. Zhang et al., Nature 481, 157–163 (2012).23. R. Katainen et al., Nat. Genet. 47, 818–821 (2015).24. N. Heidari et al., Genome Res. 24, 1905–1917 (2014).25. C. T. Ong, V. G. Corces, Nat. Rev. Genet. 15, 234–246 (2014).26. C. G. Mullighan et al., Nature 446, 758–764 (2007).

ACKNOWLEDGMENTS

Supported by NIH grants HG002668 (R.A.Y.), CA109901 (R.A.Y.),NS088538 (R.J.), MH104610 (R.J.), and AI120766 (M.H.P.); an ErwinSchrödinger Fellowship (J3490) from the Austrian Science Fund(FWF) (D.H.); Ludwig Graduate Fellowship funds (A.S.W.); the LaurieKraus Lacob Faculty Scholar Award in Pediatric TranslationalResearch (M.H.P.); Hyundai Hope on Wheels (M.H.P.); and DanishCouncil for Independent Research, Medical Sciences, individualpostdoctoral grant DFF–1333-00106B and Sapere Aude ResearchTalent grant DFF–1331-00735B (R.O.B.). Work in the Dekker lab issupported by the National Human Genome Research Institute (R01HG003143, U54 HG007010, U01 HG007910), the National CancerInstitute (U54 CA193419), the NIH Common Fund (U54 DK107980,U01 DA 040588), the National Institute of General MedicalSciences (R01 GM 112720), and the National Institute of Allergyand Infectious Diseases (U01 R01 AI 117839). J.D. is an investigatorof the Howard Hughes Medical Institute. We thank R. Fitzgerald,S. Grimmond, and the ICGC Genome Projects ESAD-UK andOV-AU for permission to use genome sequence data. Data setsgenerated in this study have been deposited in the GeneExpression Omnibus under accession number GSE68978. TheWhitehead Institute filed a patent application based on this paper.R.A.Y. is a founder of Syros Pharmaceuticals, and R.J. is a founderof Fate Therapeutics.

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/351/6280/1454/suppl/DC1Materials and MethodsFigs. S1 to S16Tables S1 to S13References (27–71)

19 November 2015; accepted 18 February 2016Published online 3 March 201610.1126/science.aad9024

HIV-1 VACCINES

HIV-1 broadly neutralizing antibodyprecursor B cells revealed bygermline-targeting immunogenJoseph G. Jardine,1,2,3* Daniel W. Kulp,1,2,3* Colin Havenar-Daughton,3,4*Anita Sarkar,2,3,5* Bryan Briney,1,2,3* Devin Sok,1,2,3* Fabian Sesterhenn,1†June Ereño-Orbea,6 Oleksandr Kalyuzhniy,1,2,3 Isaiah Deresa,3,4 Xiaozhen Hu,1,3

Skye Spencer,1,3 Meaghan Jones,1,3 Erik Georgeson,1,3 Yumiko Adachi,1,2,3

Michael Kubitz,1,2,3 Allan C. deCamp,7 Jean-Philippe Julien,2,3,5,6,8 Ian A. Wilson,2,3,5,9

Dennis R. Burton,1,2,3,10 Shane Crotty,3,4,11‡ William R. Schief1,2,3,10‡

Induction of broadly neutralizing antibodies (bnAbs) is a major HIV vaccine goal.Germline-targeting immunogens aim to initiate bnAb induction by activating bnAbgermline precursor B cells. Critical unmet challenges are to determine whether bnAbprecursor naïve B cells bind germline-targeting immunogens and occur at sufficientfrequency in humans for reliable vaccine responses. Using deep mutational scanningand multitarget optimization, we developed a germline-targeting immunogen (eOD-GT8)for diverse VRC01-class bnAbs. We then used the immunogen to isolate VRC01-classprecursor naïve B cells from HIV-uninfected donors. Frequencies of true VRC01-classprecursors, their structures, and their eOD-GT8 affinities support this immunogen as acandidate human vaccine prime. These methods could be applied to germline targetingfor other classes of HIV bnAbs and for Abs to other pathogens.

Development of an HIV vaccine is a globalhealth priority. Recent discoveries of po-tent broadlyneutralizingantibodies (bnAbs)that bind to relatively conserved epitopeson the HIV Env glycoprotein trimer and

protect against challenge in animal models have

reinvigorated vaccine design efforts to inducebnAbs (1). However, bnAbs have not been elicitedin standard animal models or humans.Germline targeting, a vaccine priming strat-

egy to initiate the affinity maturation of selectgermline-precursor B cells, has promise to initiate

1458 25 MARCH 2016 • VOL 351 ISSUE 6280 sciencemag.org SCIENCE

RESEARCH | REPORTSon July 23, 2020

http://science.sciencemag.org/

Dow

nloaded from

bnAb induction. The goals for a germline-targetingprime are to activate B cell precursors with bnAbpotential, select productive (bnAb-like) somaticmutations, and generate an expanded popula-tion of memory B cells that can be boosted andmatured subsequently to shepherd the responsefurther toward bnAb development (2, 3). For afew HIV bnAbs, next-generation sequencing ofantibody populations during bnAbdevelopment ininfected individuals has allowed bioinformaticinference of likely human germline precursors(4, 5). For most bnAbs, however, true human pre-cursors are not known but are usually approxi-mated by “germline-reverted” antibodies that useinferred germline V and J genes and retain ma-ture CDR3 (complementarity-determining region3) loops. Because CDR3 loops typically play a ma-jor role in antibody affinity and specificity, germline-reverted bnAbs are not known to be reliableproxies for true germline precursors.VRC01-class bnAbs are an important test case

for germline targeting, because they are amongthe most broad and potent of HIV bnAbs and be-cause their germline-reverted forms show no de-tectable affinity for HIV Env glycoproteins (6–10).Knock-inmice transgenic for a germline-revertedVRC01-class heavy chain responded to immuni-zation with the germline-targeting eOD-GT8 60-subunit self-assembling nanoparticle (60mer) butnot with native-like Env trimers, providing proofof principle that germline-targeting immunogenscan initiate a VRC01-class response if well-matchedB cells are present and competing B cells arestrongly reduced in frequency (2, 3). Here, weaddress further critical knowledge gaps thatobstruct the development of this (or any) germline-targeting immunogen as a human vaccine: Dothe targeted bnAb precursors exist in humans?What is the frequency and person-to-person varia-tion of germline-targeting immunogen–specificbnAb precursors? Can the germline-targeting im-munogen bind the targeted human bnAb pre-cursors in competition with other B cells in thefully complex human B cell repertoire? We exam-

ined these questions by developing new ex vivoapproaches and protein design methods.When we used the VRC01-class germline-

targeting immunogen eOD-GT6 (9) as bait toscreen human naïve B cells via a two-phasemultiple-validation methodology (11) (fig. S1),we failed to isolate VRC01-class B cells. We did,however, isolate non–VRC01-class naïve B cellswith Ab affinities as low as 120 mM for eOD-GT6(fig. S1). We therefore set out to develop an im-proved variant of eOD-GT6 with higher affinityand breadth for germline-reverted VRC01-classAbs, hypothesizing that such improvementsmighttranslate into improved affinity for diverse trueVRC01-class precursor Abs.To improve on eOD-GT6, we used yeast display

library screening coupled with next-generationsequencing (12). We screened a library of everypoint mutation at the 58 eOD:Ab interface posi-tions on eOD-GT7, a slightly improved version ofeOD-GT6 (11), against each of 29 VRC01-class Abs(18 germline-reverted and 11 mature bnAbs). Bymeasuring binding enrichments for each muta-tion and antibody (Fig. 1A and fig. S2), we iden-tified 12 positions in eOD-GT7 at which one ormore mutations were favorable (enriched by atleast a factor of 2) for binding to the majority (atleast 10 of 18) of germline-reverted bnAbs, andanother four positions at which one or more mu-tations were enriched by at least a factor of 1.25for binding to the vast majority (at least 17 of 18)of germline-reverted bnAbs (Fig. 1B). To identifycombinations of mutations predicted to conferthe greatest binding cross-reactivity, we thencreated a library encompassing all combinationsof a filtered set of the favorable mutations atthose 16 positions (13) (Fig. 1C). Upon screeningthis combinatorial library against the panel of29 VRC01-class Abs, we identified a sequence,eOD-GT8, predicted to have optimal breadthagainst the entire panel (Fig. 1C, figs. S3 and S4,and table S1).Relative to eOD-GT6, eOD-GT8 demonstrated

superior affinity and breadth of binding togermline-reverted Abs (Fig. 1D and table S2).eOD-GT8 bound to all germline-reverted Abs inthe panel, whereas eOD-GT6 bound to only 8 of14 Abs with dissociation constants (KD) of lessthan 100 mM. For those eight germline-revertedAbs, the geometric mean affinity of eOD-GT8was higher than that of eOD-GT6 by a factor of2100; eOD-GT8 also had improved affinity (fac-tor of 3) for VRC01-class bnAbs. The tightest eOD-GT8 binding detected was for germline-revertedPGV20, with a KD of 508 fM (95% confidenceinterval, 234 to 943 fM) (Fig. 1D and fig. S5), afactor of 5900 improvement over eOD-GT6 (KD =3 nM) and a factor of 33 million improvementover the original eOD construct, eOD Base [KD =17 mM (9)]—a remarkable affinity improvementfor a protein-protein interface.To examine whether VRC01-class precursors

targeted by eOD-GT8 exist in humans, we per-formed epitope-specific B cell sorting from a poolof peripheral bloodmononuclear cells fromhealthy,HIV-seronegative donors. Epitope-specific B cellsbound tetramers of eOD-GT8 but not tetramers

of eOD-GT8-KO, a variant of eOD-GT8 with mu-tations abrogating binding by VRC01-class germline-reverted Abs. After sequencing immunoglobulin(Ig) genes from single sorted cells, we searchedfor VRC01-class antibody sequences—that is,those with a heavy chain that used VH1-2 alleles*02, *03, or *04 and a light chain with a 5–aminoacid CDR3 (9, 14). After sorting 2.4 million IgM+/IgG–/CD19+ B cells pooled from nine donors, werecovered a single GT8+/GT8-KO– Ab that qual-ified as a VRC01-class precursor. This Ab, VRC01c-HuGL1, bound to eOD-GT8 with a KD of 22 mMand had no detectable affinity for eOD-GT8-KO(fig. S6).To assess both the percentage of people who

possess VRC01-class germline precursor B cellsand the frequency of VRC01-class germline pre-cursor B cells within a given donor, we screenednaïve B cells from 15 healthy, HIV-seronegativedonors individually rather than pooled. For 7of 15 samples, we used the two-phase multiple-validation methodology that first assesses spec-ificity by probe binding in flow cytometry andthen confirms specificity and lack of polyreactiv-ity by single-cell secreted IgM (fig. S7); for eightsubsequent donors, we relied on sorting speci-ficity alone. For optimal cell sorting sensitivity,B cells were required to simultaneously bind twoeOD-GT8 probesmultimerized differently [trimer(“tri”) and streptavidin tetramers (“SA”)] whilenot binding eOD-GT8-KO-SA (Fig. 2A and fig. S7).For the 15 donors, the mean frequency of eOD-GT8tri+/SA+ B cells among 61.6 million naïve Bcells sortedwas 0.0056% (Fig. 2B). Strikingly, a vastmajority (84± 14%) of these eOD-GT8tri+/SA+B cellsdid not bind eOD-GT8-KO-SA (Fig. 2C), which sug-gests that naïveB cell reactivity to eOD-GT8 is highlyfocused to the CD4 binding site (CD4bs) (15).Paired heavy and kappa light chain sequences

were recovered from 173 eOD-GT8tri+/SA+/eOD-GT8-KO– B cells. All sequences were essentiallygermline, confirming the naïve B cell sorts. Half(50%) of these B cells were VH1-2, whereas only4% of control B cells from reference (16) wereVH1-2 (c2 = 29.9, P < 0.0001; Fig. 2D and fig. S8).Among these 87 VH1-2+ B cells, 26 had a lightchain CDR3 (L-CDR3) length of 5 amino acids, afactor of 85 enrichment relative to control B cells(c2 = 32.6, P < 0.0001; Fig. 2E). Twenty-five of the26 used the VH1-2*02 allele and one used VH1-2*04 (table S3); thus, 15% (26/173) of GT8tri+/SA+/eOD-GT8-KO– B cells were VRC01-class. In total,we identified 27 independent VRC01-class naïveB cells, including VRC01c-HuGL1.In addition to the VH1-2 alleles and critical 5–

amino acid L-CDR3, VRC01-class bnAbs possessseveral additional defining features, including aconsensusL-CDR3ofGln-Gln-Tyr-Glu-Phe (QQYEF).The majority of VRC01-class precursors we iso-lated contained a QQYxx partial VRC01-classconsensus motif that was significantly enrichedrelative to control B cells (67% versus 11%; c2 =8.2, P < 0.0001; Fig. 2F). Furthermore, 11% con-tained a QQYEx L-CDR3 motif (versus 1.5% ofcontrol B cells), one mutation away from a per-fect mature VRC01-class L-CDR3 (Fig. 2F). In ad-dition, the L-CDR1 loop is under strong selective

SCIENCE sciencemag.org 25 MARCH 2016 • VOL 351 ISSUE 6280 1459

1Department of Immunology and Microbial Science, TheScripps Research Institute, La Jolla, CA 92037, USA. 2IAVINeutralizing Antibody Center, The Scripps Research Institute,La Jolla, CA 92037, USA. 3Center for HIV/AIDS VaccineImmunology and Immunogen Discovery, The Scripps ResearchInstitute, La Jolla, CA 92037, USA. 4Division of VaccineDiscovery, La Jolla Institute for Allergy and Immunology, LaJolla, CA 92037, USA. 5Department of Integrative Structuraland Computational Biology, The Scripps Research Institute,La Jolla, CA 92037, USA. 6Program in Molecular Structureand Function, Hospital for Sick Children Research Institute,Toronto, Ontario M5G 0A4, Canada. 7Vaccine and InfectiousDisease Division, Statistical Center for HIV/AIDS Researchand Prevention (SCHARP), Fred Hutchinson Cancer ResearchCenter, Seattle, WA 98109, USA. 8Departments ofBiochemistry and Immunology, University of Toronto,Toronto, Ontario M5S 1A8, Canada. 9Skaggs Institute forChemical Biology, The Scripps Research Institute, La Jolla,CA 92037, USA. 10Ragon Institute of MGH, MIT, and Harvard,Cambridge, MA 02129, USA. 11Division of Infectious Diseases,Department of Medicine, University of California San DiegoSchool of Medicine, La Jolla, CA, USA.*These authors contributed equally to this work. †Present address:Institute of Bioengineering, École Polytechnique Fédérale deLausanne, CH-1015 Lausanne, Switzerland. ‡Correspondingauthor. E-mail: schief@scripps.edu (W.R.S.); shane@lji.org (S.C.)

RESEARCH | REPORTSon July 23, 2020

http://science.sciencemag.org/

Dow

nloaded from

pressure during VRC01-class bnAb affinity matu-ration tominimize clasheswith gp120 (6, 17). VRC01-class bnAb L-CDR1 loops generally become veryshort (2 to 6 amino acids) through deletion, orretain a germline length of 6 amino acids andadd flexible glycines (17). Of the 27 VRC01-classprecursors isolated by eOD-GT8, 23 used Vkgenes containing L-CDR1 loops of 6 or 7 aminoacids (Fig. 2G), thus confirming potential to de-velop into VRC01-class bnAbs. Indeed, 17 of theVRC01-class naïve B cells had Vk genes used inknown VRC01-class bnAbs (Fig. 2H). At least 24of the VRC01-class precursors hadH-CDR3 lengthsof 10 to 19 amino acids (Fig. 2I) (18), consistentwith knownVRC01-class bnAb lengths of 10 to 19

amino acids. Thus, not only are the eOD-GT8isolated naïve B cells highly enriched for VRC01-class core characteristics of VH1-02 and a 5–aminoacid L-CDR3, they possess further refined se-quence attributes of VRC01-class bnAbs.Combining data from the 15 donors analyzed

individually, the overall frequency of recoveredVRC01-class precursors was 1 in 2.4million naïveB cells (Fig. 2J), consistent with both our firstpooled sort and a previous bioinformatically es-timated range (17). The observed counts were con-sistent with a Poisson distribution with constantfrequency of 1 in 2.4 million (Fig. 2K) (11), whichsuggests that VRC01-class precursors occur at aconsistent rate among 96% of humans possess-

ing the necessary VH1-2 alleles (9). Adults havean estimated 1010 to 1011 B cells, and lymphnodeseach have ~50 million B cells, of which ~65 to75% are naïve B cells (19). Thus, our results indi-cate that VRC01-class precursor B cells are rela-tively common in humans: At least 2700 to 31,000eOD-GT8–reactive VRC01-class naïve B cells arelikely present in nearly all potential human vac-cine recipients,with~15 suchB cells in each lymphnode, at any given time (20).The KD values of 24 isolated VRC01-class pre-

cursors for monovalent eOD-GT8 ranged from57 mM to 125 nM, with a geometric mean KD

of 3.4 mM (Fig. 2L and table S4), weaker thangermline-reverted VRC01-class Abs by a factor of

1460 25 MARCH 2016 • VOL 351 ISSUE 6280 sciencemag.org SCIENCE

Fig. 1. Development ofeOD-GT8. (A) Model of germline-reverted VRC01 (gray surface)interacting with eOD-GT7(cartoon) with the 58 positionssubjected to deep mutationalscanning shown as magenta,green, and orange spheres rep-resenting the three mutagenizedlinear segments. Binding enrich-ments, the ratio of the frequencyof a mutation in the top 10%binding population to the fre-quency of the same mutation inall cells displaying eOD-GT7,were computed for each muta-tion on eOD-GT7 for germline-reverted VRC01 and are shownas a heat map on the right, inwhich blue denotes unfavorablemutations, red denotes favor-able mutations, and whitedenotes the amino acid residuein eOD-GT7. (B) The combinedbinding enrichments fromindependent yeast displayscreens for 18 germline-revertedVRC01-class bnAbs are shownas a multidimensional heat mapin which the color scale fromyellow to red indicatesincreasing favorable averageenrichment. Symbol sizes reflectthe breadth of enrichment (thenumber of germline-revertedAbs with enriched binding foreach point mutation). Ifenriched, the eOD-GT7 aminoacid residue is indicated by across. (C) Sequence logosdepicting amino acids at each of16 positions in the combinatoriallibrary (top), the sequencesselected from the combinatoriallibrary for improved binding to germline-reverted VRC01-class bnAbs(middle), and the final sequence of eOD-GT8 (bottom). Abbreviations: A, Ala;C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn;P, Pro; Q, Gln; R, Arg; T, Thr; V, Val; W, Trp; Y, Tyr. (D) Surface plasmon res-onance (SPR) dissociation constants measured for both germline-revertedand mature VRC01-class bnAbs against eOD-GT6 and eOD-GT8. Solid blue

lines show geometric mean measured over all the data, using the value KD =100 mM for samples with KD > 100 mM; dashed blue lines show geometricmeans computed for the eight germline-reverted Abs or 12 bnAbs forwhich KDs < 100 mM could be measured for both eOD-GT6 and eOD-GT8.The lowermost dotted line signifies the limit of detection for our SPR instru-ment (16 pM); KDs below this value were measured by KinExa.

RESEARCH | REPORTSon July 23, 2020

http://science.sciencemag.org/

Dow

nloaded from

590 (geometric mean KD = 5.8 nM for the panel),most likely due to the naïve CDR3 loops in theformer as opposed to the affinity-matured CDR3loops in the latter. The VRC01-class naïve B cellaffinities are in the range expected to allow amultivalent eOD-GT8 immunogen, such as eOD-GT8 60mer (2, 9), to activate B cells and initiategerminal centers (21, 22). Our data also suggestthat eOD-GT8 has promise to produce VRC01-class memory even given competition from non–VRC01-class B cells, as eOD-GT8 exhibited a highdegree of CD4bs immunofocusing (Fig. 2C) and

VRC01-class precursors had an affinity advan-tage (factor of ≥3) over non–VRC01-class CD4bsepitope-binding precursors (Fig. 2L). The frequen-cies and eOD-GT8 affinities of bona fide VRC01-class precursors isolated here warrant humanimmunization studies with eOD-GT8 60mernanoparticles.Only 2 of 20 tested VRC01-class precursors

had detectable affinity for eOD-GT6 (Fig. 2L).Equilibrium binding KD values were 36 mM and69 mM, and these Abs had two of the highest af-finities for eOD-GT8 at 506 nM and 258 nM, re-

spectively (table S4). These data, combined withthe failure of eOD-GT6 probe B cell screens toisolate VRC01-class precursors, suggest that theengineered breadth and affinity improvementsin eOD-GT8 represent a major advance towardpractical utility in human vaccination.We sought to confirm that the isolated VRC01-

class precursors engage the CD4bs in the samestructural binding mode as VRC01-class bnAbs(6, 17, 23–25) and germline-reverted VRC01 (9).We solved the crystal structure of isolated pre-cursor VRC01c-HuGL2 (eOD-GT8 KD = 368 nM)

SCIENCE sciencemag.org 25 MARCH 2016 • VOL 351 ISSUE 6280 1461

Fig. 2. eOD-GT8-binding VRC01-class naïve B cells exist in healthy hu-man donors. (A) eOD-GT8+ naïve CD19+IgG– B cells. (B and C) eOD-GT8+ Bcell frequency (B) and eOD-GT8 KO– cells (C) among eOD-GT8+ B cells inindividual donors. (D) VH1-2 usage among eOD-GT8+/eOD-GT8 KO– sorted Bcells (n = 173) versus control B cells.VH1-2 (red) allele frequencies are indicated.(E) B cells expressing a 5–amino acid L-CDR3 among VH1-2+ B cells isolated byeOD-GT8 versus control B cells. (F) L-CDR3 sequence logos of VRC01-classbnAbs (top),VRC01-classnaïveprecursors (middle), andcontrolB cells (bottom).

(G) L-CDR1 lengths of 27 VRC01-class naïve B cells. (H) Light chain Vgene usageof 27 VRC01-class naïve B cells. Known VRC01-class bnAb Vkgenes are in red.(I) H-CDR3 lengths of VRC01-class naïve B cells versus control B cells. (J) TotalB cells screened andVRC01-class naïve B cells found in 15 individuals. (K) Poissondistribution modeling of the number of VRC01-class naïve B cells. Vertical linesshow the 2.5% and 97.5% quantiles. (L) SPR dissociation constants for eOD-GT6 or eOD-GT8 binding to VRC01-class or non–VRC01-class Abs derived fromeOD-GT8–sorted human naïve B cells. Solid red lines indicate geometric mean.

RESEARCH | REPORTSon July 23, 2020

http://science.sciencemag.org/

Dow

nloaded from

in complex with eOD-GT8 in two crystal forms(I222, 2.16 Å, and C2, 2.44 Å; table S5). Com-parison of this structure with the complex ofcore-gp120 bound to VRC01 [PDB ID: 3NGB (6)]shows the same binding mode (Fig. 3A), includ-ing specific H-CDR2 and L-CDR3 conformations(Fig. 3B) (26) that together account for morethan 67.2% of the Fv domain buried surface area(Fig. 3C and table S6).When interface residues ofeOD-GT8 and core-gp120 are aligned, VH and VL

of VRC01c-HuGL2 and VRC01 have high similar-ity (Ca RMSD 0.7 Å; Fig. 3A and fig. S9). Thesestructural observations confirm VRC01c-HuGL2as a bona fide VRC01-class precursor and support

the conclusion that all of the eOD-GT8–specificnaïve B cells using VH1-2 and a 5–amino acidL-CDR3 are bona fide VRC01-class precursors.Comparison of the eOD-GT8–VRC01c-HuGL2structurewith a 1.82 Å unliganded VRC01c-HuGL2structure shows that the important H-CDR2 andL-CDR3 loops are preconfigured in the unboundstate and do not require any conformationalchanges for engagement with gp120 CD4bs (Fig.3D), heightening the appeal of VRC01-class germ-line targeting. A 2.9 Å unliganded structure ofeOD-GT8 (Fig. 3E and fig. S10) demonstratesfaithful mimicry of the VRC01-class antibody-bound conformation (27), thus helping to explain

the increased affinity of eOD-GT8 for true VRC01-class bnAb precursors.The interaction of the naïve human B cell

repertoire with vaccine antigens has not beencharacterized previously. Given the vast immuno-globulin sequence space, direct probing of thehuman naïve B cell repertoire was a critical testof the physiologically relevant binding potentialof the germline-targeting immunogen. The anti-body sequence features, binding affinities, andhigh structural similarity of the eOD-GT8–specificnaïve B cell–derived antibodies to VRC01 all demon-strate the power of germline-targeting designwhencombined with human B cell probing. Similar

1462 25 MARCH 2016 • VOL 351 ISSUE 6280 sciencemag.org SCIENCE

F F

Fig. 3. Structural analysis of eOD-GT8 and human germline antibodyVRC01c-HuGL2 complex. (A) Crystal structures of VRC01c-HuGL2 + eOD-GT8(blue, LC; salmon, HC; orange, eOD-GT8) and ofmatureVRC01 + gp120 (PDB ID:3NGB, in white) shown in the same orientation, showing eOD-GT8 superimposedon gp120, and showing only the antibody Fv regions for clarity. (B) Comparison ofthe H-CDR2 and L-CDR3 conformations from the structures in (A). (C) Com-parison of buried surface areas for theVH andVL residues of VRC01c-HuGL2andmature VRC01 + gp120, in their bound forms. (D) Comparison of H-CDR2 and

L-CDR3 conformations of unliganded and eOD-GT8-liganded VRC01c-HuGL2Fab. All atoms of VH andVLwere aligned. In the left image, H-CDR2and L-CDR3are shown as sticks; in the right image theCDRs are shown according toB-factorsreporting local structural flexibility using a relative scale in which increasing wirethickness and warmness of color (blue to red) indicates increasing mobility.(E) Crystal structure of unliganded eOD-GT8 shown in cartoon representation(left) and a superposition of unliganded and VRC01c-HuGL2–bound formsof eOD-GT8 (right; Ca RMSD = 0.4 Å).

RESEARCH | REPORTSon July 23, 2020

http://science.sciencemag.org/

Dow

nloaded from

methods, including both protein design and hu-man B cell probing methods, could be used toimprove and evaluate germline-targeting immu-nogens for other classes of HIV bnAbs and forAbs against other pathogens. Thesemethodsmaybe particularly important to develop and testgermline-targeting approaches for bnAbs thatrely heavily onHCDR3 and hencemay have lowerprecursor frequencies.

REFERENCES AND NOTES

1. D. R. Burton, J. R. Mascola, Nat. Immunol. 16, 571–576 (2015).2. J. G. Jardine et al., Science 349, 156–161 (2015).3. P. Dosenovic et al., Cell 161, 1505–1515 (2015).4. H. X. Liao et al., Nature 496, 469–476 (2013).5. N. A. Doria-Rose et al., Nature 509, 55–62 (2014).6. T. Zhou et al., Science 329, 811–817 (2010).7. J. F. Scheid et al., Science 333, 1633–1637 (2011).8. S. Hoot et al., PLOS Pathog. 9, e1003106 (2013).9. J. Jardine et al., Science 340, 711–716 (2013).10. A. T. McGuire et al., J. Exp. Med. 210, 655–663 (2013).11. See supplementary materials on Science Online.12. T. A. Whitehead et al., Nat. Biotechnol. 30, 543–548 (2012).13. Filtering was done to limit library size, to exclude mutations

detrimental to binding the majority of mature bnAbs, to reducehydrophobic exposure, to exclude unpaired cysteines, and tominimize nonconservative changes to epitope components.

14. A. P. West Jr., R. Diskin, M. C. Nussenzweig, P. J. Bjorkman,Proc. Natl. Acad. Sci. U.S.A. 109, E2083–E2090 (2012).

15. Several design features of eOD-GT8 are likely responsiblefor this immunofocusing, including the relatively small sizeof eOD-GT8 (175 amino acids) and its compact structure(no exposed loops except Loop D and V5 within the CD4bs),as well as the glycan shielding by 10 glycans covering much ofthe eOD-GT8 surface outside the CD4bs.

16. B. J. DeKosky et al., Nat. Med. 21, 86–91 (2015).17. T. Zhou et al., Immunity 39, 245–258 (2013).18. Two of the isolated VRC01-class precursors had incomplete

H-CDR3 sequences preventing determination of H-CDR3 length.19. H. Morbach, E. M. Eichhorn, J. G. Liese, H. J. Girschick,

Clin. Exp. Immunol. 162, 271–279 (2010).20. The frequency of 1 in 2.4 million is an underestimate of the

true frequency among naïve B cells, because not all B cellscounted by the sorter as eOD-GT8tri+/SA+/eOD-GT8-KO– weresorted into a well (cell sorter loss), paired heavy chain andlight chain (HC and LC) sequences were recovered from fewerthan half of eOD-GT8tri+/SA+/eOD-GT8-KO– B cells sortedinto wells [a result of the inherent limitations of single-cellpolymerase chain reaction (PCR)], and B cells bearing lambdalight chains were not analyzed. By correcting for cell sorterand PCR losses, the frequency of VRC01-class naïve B cellprecursors is calculated as 1 in 400,000 naïve B cells (11).VRC01-class precursors may also exist in the memory B cellpopulation in healthy humans, but their frequency remainsto be measured.

21. T. A. Shih, E. Meffre, M. Roederer, M. C. Nussenzweig,Nat. Immunol. 3, 570–575 (2002).

22. J. M. Dal Porto, A. M. Haberman, G. Kelsoe, M. J. Shlomchik,J. Exp. Med. 195, 1215–1221 (2002).

23. X. Wu et al., Science 333, 1593–1602 (2011).24. R. Diskin et al., Science 334, 1289–1293 (2011).25. R. Diskin et al., J. Exp. Med. 210, 1235–1249 (2013).26. T. Zhou et al., Cell 161, 1280–1292 (2015).27. We conclude that mutations that led to the design of

eOD-GT8 from eOD-GT6 further stabilize the antibody-boundstate, based on a higher similarity between the VRC01c-HuGL2–bound and unliganded eOD-GT8 (all-atom RMSD =0.98 Å, alignment of 1206 atoms) versus GL-VRC01–boundeOD-GT6 (PDBID: 4JPK) and unliganded eOD-GT6 (PDBID:4JPJ) (all-atom RMSD = 3.0 Å, alignment of 1343 atoms)(fig. S10).

ACKNOWLEDGMENTS

We thank L. Stamatatos, T. Whitehead, and M. Nussenzweig fordiscussions; the Flow Cytometry Core at the La Jolla Institute forAllergy and Immunology and L. Nosworthy for expert cell-sortingassistance; H. Tien for technical support with crystallization robots;and A. Irimia for discussions and technical help. This work wassupported by the International AIDS Vaccine Initiative NeutralizingAntibody Consortium and Center (W.R.S., I.A.W., D.R.B.); CAVDfunding for the IAVI NAC Center (W.R.S., I.A.W., D.R.B.); CAVD

Vaccine Immunology Statistical Center (VISC) (A.C.D.); theRagon Institute of MGH, MIT, and Harvard (D.R.B. and W.R.S.);the Bayer Science and Education Foundation (F.S.); the HelenHay Whitney Foundation and Howard Hughes Medical Institute(J.G.J.); and National Institute of Allergy and Infectious Diseasesgrants P01 AI094419 (W.R.S.), CHAVI-ID UM1 AI100663 (W.R.S.,S.C., I.A.W., D.R.B.), P01 AI110657 (I.A.W.), and R01 AI084817(I.A.W.). Portions of this research were carried out at the StanfordSynchrotron Radiation Lightsource (SSRL), a Directorate of SLACNational Accelerator Laboratory and an Office of Science UserFacility operated for the U.S. Department of Energy (DOE) Office ofScience by Stanford University. The SSRL Structural MolecularBiology Program is supported by the DOE Office of Biological andEnvironmental Research, and by the National Institute of GeneralMedical Sciences (including P41GM103393). The data presentedin this manuscript are tabulated in the main paper and in thesupplementary materials. Coordinates and structure factors forVRC01c-HuGL2 Fab, VRC01c-HuGL2+eOD-GT8 (2.44 Å), andVRC01c-HuGL2+eOD-GT8 (2.16 Å) have been deposited with theProtein Data Bank with accession codes 5IFA, 5IF0, and 5IES,

respectively. Sequences for heavy and light chains of HuGL1through HuGL27 have been deposited at NCBI with GenBankaccession codes KU760929 to KU760982. Materials andinformation are available by MTA from the Scripps ResearchInstitute. IAVI and the Scripps Research Institute have filed apatent relating to the eOD-GT8 immunogens in this manuscript,which included inventors J.G.J., D.W.K., and W.R.S. W.R.S. isa co-founder and stockholder in Compuvax Inc., which hasprograms in non-HIV vaccine design that might benefit indirectlyfrom this research.

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/351/6280/1458/suppl/DC1Materials and MethodsFigs. S1 to S14Tables S1 to S7References (28–41)

21 November 2015; accepted 24 February 201610.1126/science.aad9195

TUMOR IMMUNOLOGY

Clonal neoantigens elicit T cellimmunoreactivity and sensitivity toimmune checkpoint blockadeNicholas McGranahan,1,2,3* Andrew J. S. Furness,3,4* Rachel Rosenthal,3*Sofie Ramskov,5 Rikke Lyngaa,5 Sunil Kumar Saini,5 Mariam Jamal-Hanjani,3

Gareth A. Wilson,1,3 Nicolai J. Birkbak,1,3 Crispin T. Hiley,1,3 Thomas B. K. Watkins,1,3

Seema Shafi,3 Nirupa Murugaesu,3 Richard Mitter,1 Ayse U. Akarca,4,6

Joseph Linares,4,6 Teresa Marafioti,4,6 Jake Y. Henry,3,4 Eliezer M. Van Allen,7,8,9

Diana Miao,7,8 Bastian Schilling,10,11 Dirk Schadendorf,10,11 Levi A. Garraway,7,8,9

Vladimir Makarov,12 Naiyer A. Rizvi,13 Alexandra Snyder,14,15

Matthew D. Hellmann,14,15 Taha Merghoub,14,16 Jedd D. Wolchok,14,15,16

Sachet A. Shukla,7,8 Catherine J. Wu,7,8,17,18 Karl S. Peggs,3,4 Timothy A. Chan,12

Sine R. Hadrup,5 Sergio A. Quezada,3,4† Charles Swanton1,3†

As tumors grow, they acquire mutations, some of which create neoantigens thatinfluence the response of patients to immune checkpoint inhibitors. We explored theimpact of neoantigen intratumor heterogeneity (ITH) on antitumor immunity. Throughintegrated analysis of ITH and neoantigen burden, we demonstrate a relationshipbetween clonal neoantigen burden and overall survival in primary lung adenocarcinomas.CD8+ tumor-infiltrating lymphocytes reactive to clonal neoantigens were identified inearly-stage non–small cell lung cancer and expressed high levels of PD-1. Sensitivityto PD-1 and CTLA-4 blockade in patients with advanced NSCLC and melanoma wasenhanced in tumors enriched for clonal neoantigens. T cells recognizing clonalneoantigens were detectable in patients with durable clinical benefit. Cytotoxicchemotherapy–induced subclonal neoantigens, contributing to an increased mutationalload, were enriched in certain poor responders. These data suggest that neoantigenheterogeneity may influence immune surveillance and support therapeuticdevelopments targeting clonal neoantigens.

Recent studies have highlighted the rele-vance of tumor neoantigens in the recog-nition of cancer cells by the immune system(1–4), prompting a renewed interested inpersonalized vaccines and cell therapies

that target cancer mutations (5, 6). However,although genomic data are revealing the extentof genetic heterogeneity within single tumors (7),the influence of intratumor heterogeneity (ITH)upon the neoantigen landscape and sensitivity toimmune modulation is unclear.

To explore neoantigen heterogeneity andits influence on antitumor immunity in early-stage non–small cell lung cancer (NSCLC), weapplied a bioinformatics pipeline to seven pri-mary NSCLCs subjected to multiregion se-quence analysis (table S1) (8, 9). In total, 2860putative neoantigens were predicted acrossthe cohort, with a median of 326 neoantigenspredicted per tumor (range of 80 to 741) (Fig. 1A).Neoantigen heterogeneity varied considerab-ly, with an average of 44% neoantigens found

SCIENCE sciencemag.org 25 MARCH 2016 • VOL 351 ISSUE 6280 1463

RESEARCH | REPORTSon July 23, 2020

http://science.sciencemag.org/

Dow

nloaded from

immunogenHIV-1 broadly neutralizing antibody precursor B cells revealed by germline-targeting

SchiefAdachi, Michael Kubitz, Allan C. deCamp, Jean-Philippe Julien, Ian A. Wilson, Dennis R. Burton, Shane Crotty and William R.Ereño-Orbea, Oleksandr Kalyuzhniy, Isaiah Deresa, Xiaozhen Hu, Skye Spencer, Meaghan Jones, Erik Georgeson, Yumiko Joseph G. Jardine, Daniel W. Kulp, Colin Havenar-Daughton, Anita Sarkar, Bryan Briney, Devin Sok, Fabian Sesterhenn, June

DOI: 10.1126/science.aad9195 (6280), 1458-1463.351Science 

, this issue p. 1458Scienceshaping of the B cell response with subsequent immunogens may eventually elicit bNAbs in people.their affinities for the immunogen, and structural analysis suggest that the immunogen is a promising candidate. Further germline versions of the immunoglobulin genes harbored by a particular class of bNAbs. The frequencies of these B cells,goal: They engineered an immunogen that could engage B cells from HIV-uninfected individuals that express the

report an important step toward thiset al.Scientists aim to design vaccines that would elicit such antibodies. Jardine Some HIV-infected individuals develop heavily mutated, broadly neutralizing antibodies (bNAbs) that target HIV.

Baby steps toward bNAbs

ARTICLE TOOLS http://science.sciencemag.org/content/351/6280/1458

MATERIALSSUPPLEMENTARY http://science.sciencemag.org/content/suppl/2016/03/23/351.6280.1458.DC1

CONTENTRELATED

http://stm.sciencemag.org/content/scitransmed/7/319/319ra206.fullhttp://stm.sciencemag.org/content/scitransmed/7/296/296ra112.fullhttp://stm.sciencemag.org/content/scitransmed/7/298/298ra120.fullhttp://stm.sciencemag.org/content/scitransmed/7/310/310rv7.full

REFERENCES

http://science.sciencemag.org/content/351/6280/1458#BIBLThis article cites 35 articles, 14 of which you can access for free

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Terms of ServiceUse of this article is subject to the

is a registered trademark of AAAS.ScienceScience, 1200 New York Avenue NW, Washington, DC 20005. The title (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement ofScience

Copyright © 2016, American Association for the Advancement of Science

on July 23, 2020

http://science.sciencemag.org/

Dow

nloaded from