High-Throughput Stability Screening of Neoantigen/HLA … · stability could be important, as only a small fraction of peptides in most current vaccines are capable of eliciting CD8þ

$Page 1: High-Throughput Stability Screening of Neoantigen/HLA … · stability could be important, as only a small fraction of peptides in most current vaccines are capable of eliciting CD8þ$
Research Article

High-Throughput Stability Screening ofNeoantigen/HLA Complexes ImprovesImmunogenicity PredictionsDylan T. Blaha1, Scott D. Anderson1, Daniel M. Yoakum1, Marlies V. Hager1,Yuanyuan Zha2, Thomas F. Gajewski2, and David M. Kranz1

Abstract

Mutated peptides (neoantigens) from a patient's cancergenome can serve as targets for T-cell immunity, butidentifying which peptides can be presented by an MHCmolecule and elicit T cells has been difficult. Althoughalgorithms that predict MHC binding exist, they are notyet able to distinguish experimental differences in half-lives of the complexes (an immunologically relevantparameter, referred to here as kinetic stability). Improve-ment in determining actual neoantigen peptide/MHCstability could be important, as only a small fraction ofpeptides in most current vaccines are capable of elicitingCD8þ T-cell responses. Here, we used a rapid, high-throughput method to experimentally determine pep-tide/HLA thermal stability on a scale that will be necessary

for analysis of neoantigens from thousands of patients.The method combined the use of UV-cleavable peptide/HLA class I complexes and differential scanning fluorim-etry to determine the Tm values of neoantigen complexes.Measured Tm values were accurate and reproducible andwere directly proportional to the half-lives of the com-plexes. Analysis of known HLA-A2–restricted immunogen-ic peptides showed that Tm values better correlated withimmunogenicity than algorithm-predicted binding affini-ties. We propose that temperature stability informationcan be used as a guide for the selection of neoantigens incancer vaccines in order to focus attention on those mutat-ed peptides with the highest probability of being expressedon the cell surface.

IntroductionSuccesses in cancer immunotherapy have revealed that cancer

neoantigens are the primary target for tumor-reactive lympho-cytes isolated from patients receiving checkpoint blockade ther-apy, such as anti–CTLA-4 and anti–PD-1 (1, 2). This finding,coupled with breakthroughs in sequencing technology, hasspurred interest in neoantigens as a source of therapeutic targets(3–7). Because each cancer evolves along unique pathways,principles of individualized medicine must be applied to thepursuit of neoantigens as therapeutic targets (8). This includeswhole-exome sequencing (WES) and RNA-seq to determine a listof mutated peptides that could potentially serve as epitopes forimmune recognition. From here, candidate neoantigen peptidesare assessed using algorithms that predict their likelihood ofproteosomal processing and their binding affinity for HLA mole-cules.Manyprotocols end the preliminary analysis here andmove

straight into screening for T-cell immunogenicity or vaccinetesting.

Studies from Harndahl and colleagues (9) and Stronen andcolleagues (10) suggest that the half-life of the ternary peptide–MHC(pepMHC) complex is a better indicator of immunogenicitythan binding affinity. The half-life is determined from the disso-ciation kinetics of the complex and can, thus, be considered ameasure of the kinetic stability of the complex. Most kineticstability and binding affinity studies have been done with thecommon HLA allele, HLA-A2. Predictions for this allele are likelyto be the most accurate when it comes to class I MHC bindingalgorithms such as NetMHC4.0. In the study by Stronen andcolleagues (10), 19% of the top 57 neoantigen peptides chosenbased on NetMHC4.0 binding predictions were immunogenic.Based on experimental measurement of dissociation rates (usinganti-b2m as a probe), the percentage of immunogenic peptidesincreased to 50% of peptides with half-lives greater than 5 hours.

The ability to improve selection of appropriate neoantigencandidates is important, given the increasing interest in the useof peptides in vaccines (11). Two studies in melanoma patients(12, 13), one using a peptide formulation and the other a mini-gene RNA vaccine, showed a relatively low frequency of CD8þ

T-cell responsiveness (16% and 29% of peptides in these studies,respectively). Because only a small fraction of a patient's neoanti-gen repertoire (approximately 10 to 20 neoantigen peptides) istypically selected for use in these vaccine trials, even a 2-foldimprovement in peptide candidate selection could be significantin terms of eliciting tumor immunity in patients.

Although the measurement of pepMHC dissociation rates(half-lives) provides an approach to determining the kineticstability of the complexes, it is tedious, time-consuming, and not

1Department of Biochemistry, University of Illinois, Urbana, Illinois. 2Departmentof Pathology, Department of Medicine, and the Ben May Department of Cancer,University of Chicago, Chicago, Illinois.

Note: Supplementary data for this article are available at Cancer ImmunologyResearch Online (http://cancerimmunolres.aacrjournals.org/).

D.T. Blaha and S.D. Anderson contributed equally to this article.

Corresponding Author: David M. Kranz, University of Illinois, 600 S. Mathews,Urbana, IL 61801. Phone: 217-244-2821; Fax: 217-244-5858;E-mail: [email protected]

doi: 10.1158/2326-6066.CIR-18-0395

�2018 American Association for Cancer Research.

CancerImmunologyResearch

Cancer Immunol Res; 7(1) January 201950

on March 7, 2020. © 2019 American Association for Cancer Research. cancerimmunolres.aacrjournals.org Downloaded from

Published OnlineFirst November 13, 2018; DOI: 10.1158/2326-6066.CIR-18-0395

http://crossmark.crossref.org/dialog/?doi=10.1158/2326-6066.CIR-18-0395&domain=pdf&date_stamp=2018-12-15

http://cancerimmunolres.aacrjournals.org/

amenable to high throughput. To develop a rapid system withhigh-throughput capabilities, we reasoned that temperature sta-bility should correlate with dissociation rates of the complexes. Ifso, amethod exists thatwouldbe available tomany labs because ituses a standard real-time (RT)-PCR instrument to determinetemperature denaturation curves and stability, characterized byTm, the temperature yielding half-maximal denaturation. Themethod, called differential scanning fluorimetry (DSF), relies onthe binding of a commercially availablefluorescent probe, SYPROOrange, to denatured proteins (14). Fluorescence is monitored inan RT-PCR instrument, where use of 384-well plates allowshundreds of complexes to be analyzed in a 1-hour, hands-freeexperiment.

Here, we developed a strategy that allows crude peptide pre-parations to be used together with HLA-A2 and UV-peptidecleavage reactions, followed by DSF. We measured the thermalstability of many previously published HLA-A2 binding peptidesfor which half-life measurements, and in some cases, bindingaffinity measurements, were reported. We also compared Tmvalues and half-lives for a collection of self-peptides and singleamino acid variants forwhichwedetermined half-lives using anti-b2mor soluble high-affinity T-cell receptor as probes. Tm values ofthe self-peptides and neoantigen peptides were then comparedwith binding affinities predicted using current algorithms. Wefound that thermal stabilities (Tm values) were correlated withhalf-lifemeasurements, but less correlatedwith predicted bindingaffinities, and in some cases, less correlated with measured bind-ing affinities. DSF provides a faster, simpler, and lower-costalternative to half-life measurements and can be used to screenthousands of potential class I–restricted neoantigens rapidly.

Materials and MethodsPeptides

Peptides were synthesized by GenScript or Cellmano Biotech.Sequences of GenScript peptides are shown in Table 1 andSupplementary Tables S1–S6, and sequences ofCellmanoBiotechpeptides are shown in Supplementary Tables S7 and S8. Peptideswere provided either as "crude" preparations for 96-well libraries(�30% purity) or as preparations that were purified by HPLC(�90%purity). Lyophilized peptideswere suspended indimethylsulfoxide (DMSO) at a working concentration of 20 mmol/L andstored at �20�C. Prior to use in peptide-exchange reactions,peptides were incubated at 37�C for 1 hour.

Predicted peptide-binding affinities for HLA-A2Peptide-binding analyses were done using NetMHC 4.0

[an Artificial Neural Network (ANN), produced by the TechnicalUniversity ofDenmark, inwhich results are expressed as predictedequilibrium binding constants (in nmol/L; ref. 15)].

Generation of peptide/HLA-A2 complexes by UV-peptideexchange

HLA-A�02:01 heavy chain and b2-microglobulin (b2m) wereeach expressed in E. coli strain BL21 and refolded from inclusionbodies, together with UV-photocleavable peptide [KILGFVFJV,where J represents the cleavable 3-amino-3-(2-nitro) phenyl-propionic acid] as described previously (16, 17). Briefly,3 mmol/L of heavy chain and 2 mmol/L b2m inclusion bodieswere solubilized in 8 M urea (Fisher Scientific, BP-169-212) andrefolded in the dark at 4�C together with 30mg of UV peptide in a

1 L folding reaction. UV-peptide HLA-A2 monomers were con-centrated using an Amicon Stirred Ultrafiltration Cell with a10,000 Da MWCO ultracentrifugation disk and dialyzed over-night using a 10,000 Da MWCO Slide-A-Lyzer Dialysis Cassetteagainst 20 mmol/L Tris. The refolded complex was purified usinganion exchange and size exclusion HPLC and stored at �80�C inphosphate-buffered saline (PBS). For some studies, biotinylationwas carried out using an Avidity Biotinylation Kit (Avidity LLC),followed by HPLC purification.

UV-mediated peptide-exchange reactionsUV-mediated peptide-exchange reactions were performed in

either 0.6mL conical tubes (Denville Scientific, C2170)or 96-wellplates (CellTreat Scientific Products, 229190) to obtain pep/HLAcomplexes of interest. Each 115 mL reaction included 400 mmol/Lof exchange peptide and 4 mmol/L of UV-peptide HLA-A2 mono-mer. Tubes or 96-well plates were placed in a CL-1000 UltravioletCrosslinker (AnalytikJena US LLC) on ice for three 15-minuteincrements. To maximize exchange efficiency, the samples wereplaced 5 cm from the UV lamp andmixed with a pipette betweenexposure intervals. After the reaction, each 115 mL sample wasstored at 4�C in a 0.6 mL conical tube to avoid sample evapora-tion, which could potentially occur if peptides are stored in aplate. Reactions were stored for at least 48 hours prior to use in

Table 1. Tm analysis of HLA-A2 complexes containing immunogenic neoantigenpeptides

Gene SequenceNetMHC4.0 (nmol/L) Tm (�C)

Fritsch et al. (32)ME-1 FLDEFMEGV 3 63.2 � 0.4FNDC3B VVMSWAPPV 6 61.6 � 0.6PRDX5 LLLDDLLVSI 15 55.3 � 0.7GAS7 SLADEAEVYL 43 59.2 � 0.7KIAA0223 VLHDDLLEA 29 59.7 � 0.5GAPDH GIVEGLITTV 123 58.5 � 0.6HSP70 SLFEGIDIYT 18 59.8 � 0.6ACTININ FIASNGVKLV 281 56.9 � 0.5HAUS3 ILNAMIAKI 48 56.9 � 0.2CSNK1A1 GLFGDIYLAI 21 52.6 � 0.6CLPP ILDKVLVHL 56 57.1 � 0.4CDK4 ACDPHSGHFV 14410 59.9 � 0.3

Robbins et al. (33), Cohen et al. (34), and Prickett et al. (35)AHNAK FMPDFDLHL 6 60.1 � 0.2SRPX TLWCSPIKV 10 63.7 � 0.1COL18A1 VLLGVKLFGV 9 60.9 � 0.3ERBB2 ALIHHNTYL 18 59.4 � 0.5TEAD1 VLENFTIFLV 51 48.2 � 0.2TEAD1 SVLENFTIFL 85 55.8 � 0.2NSDHL ILTGLNYEV 8 61.4 � 0.1GANAB ALYGFVPVL 8 61.7 � 0.2

Durgeau et al. (36)CDC37L1 FLSDHLYLV 2 61.8 � 0.1FLNA HIAKSLFEV 22 56.4 � 0.4SPOP FLLDEAIGL 3 60.7 � 0.1

Ott et al. (12)ACPP VLAKKLKFV 59 56.8 � 0.7DCAKD LLHTELERFL 624 42.9 � 0.2CIT TLLSQVNKV 33 53.2 � 0.3

NOTE: HLA-A2–restricted neoantigen peptides, previously reported asimmunogenic by various methods, were analyzed using DSF. The gene name,neoantigen sequence, predicted binding affinity using NetMHC4.0, and Tmvalue with corresponding standard deviation for each complex tested areshown. Tm values represent the average and standard deviation of threereplicates in a single experiment.

Neoantigen/HLA Complexes

www.aacrjournals.org Cancer Immunol Res; 7(1) January 2019 51




stability assays. For several complexes that have been tested afterstorage, we found that they were stable at 4�C for 10 months.

Flow cytometry–based half-life measurementsA bead-based flow cytometry assay was used to measure pep/

HLA-A2 dissociation rates with a soluble high-affinity T-cellreceptor (TCR) or with an antibody to b2m, as described previ-ously (10). Biotinylated peptide/HLA-A2 complexes were cou-pled to streptavidin-coated microspheres (5-mm streptavidin-coated microspheres; Bangs Laboratories, Inc.) at a stoichiometryof about 2� 107 complexes per microsphere. After incubation onice for 30minutes, the beadmixture was washed three times withPBS containing 1% bovine serum albumin (Gold Biotechnology,A-421-250), to remove unbound peptide/HLA-A2. Samples werethen incubated at 37�Cand aliquotswere removed at various timepoints. Remaining (bound) peptide/HLA-A2 was determined byflow cytometry with an antibody to b2m (beta2-microglobulinmouse anti-human, clone B2M-01, PE conjugate; Invitrogen). Insome experiments, a soluble, biotinylated high-affinity TCR to theWT-1/HLA-A2 complex, refolded and purified as previouslydescribed (18, 19), was used to compare the dissociation detectedwith anti-b2m. Mean fluorescence intensity was plotted againsttime for each point, and curves were fit with a first-orderexponential decay curve to determine half-life value (OriginPro2017, OriginLab). Each half-life value measured was the mean oftwo or more replicate experiments. Flow cytometry data obtainedfrom an Accuri C6 (BD Biosciences) was analyzed with FCSExpress 6 software to determine mean fluorescence values foreach time point.

Differential scanning fluorimetryDSF was performed using a QuantStudio 7 Flex System real-

time PCR machine (Applied Biosystems), 384-well Microampoptical plates (Applied Biosystems, 4309849), and Microampadhesive film (Applied Biosystems, 4311971) for all peptide/HLA-A2 complexes. Each 20mL reaction consisted of 2.5mLof 40XSYPRO Orange (Invitrogen) dye, 8.5 mL of molecular biologygrade water, 5 mL of assay buffer (defined below), and the 4 mL ofUV-exchanged complexes. SYPRO Orange dye binds to hydro-phobic regions of proteins that are exposed upon denaturation,resulting influorescence enhancement.DSF assay buffer consistedof 40 mmol/L HEPES, 600 mmol/L NaCl, and 12 mmol/L EDTAfor most experiments at a pH of 7.4. We omitted the surfactantP20, used in a previous study (20). The real-time PCR machinewas programmed for a temperature ramp rate of 1�C per minutefrom 25�C to 99�C.

Determining Tm from nonlinear fitting of thermaldenaturation data

Triplicate data points for each pep/HLA-A2 complex wereanalyzed using OriginPro 2017 software by plotting thefluorescence at each temperature measured (820 data pointsper reaction). Fluorescence data were processed and fit with abi-Gaussian function to generate a curve with an R2 value of0.98 or greater. In some cases, it was necessary to truncatefluorescence values (by omitting early time points thatexhibited no increase in fluorescence at low temperatures) inorder to obtain a fit that accurately represented the data set.Complexes for which the bi-Gaussian fit resulted in an R2 valuelower than 0.98 were reevaluated with an additional peptide-exchange and DSF experiment. The first derivative of the curve

obtained from bi-Gaussian fitting was used to calculate theprecise midpoint (Tm).

Although the bi-Gaussian approach described above was usedfor all of the experiments reported here, we also compared twoalternative approaches used previously to determine Tm valuesfrom fluorescence data. The first uses a Microsoft Excel macrocalled "deleteaftermax" to automatically truncate data sets andremove post-peak quenching data (21). The sigmoidal denatur-ation curve was fit nonlinearly to a Boltzmann equation, and theTm was obtained by finding the midpoint of the unfoldingtransition, or the peak of the first derivative curve. Comparisonof DSF data analyzed with this method and the bi-Gaussianfit for various representative complexes showed similar results(Supplementary Table S1). A second approach determined the Tmby plotting the first derivative of the fluorescence emission as afunction of temperature (qF/qT), processing the curve with thepeak fitting algorithm in OriginPro 2017, applying a sigmoidalbaseline, and fitting the peak of the derivative curve with a bi-Gaussian function (20). In our experience, this fitting techniquerequires high concentrations (10 mmol/L and above) of pep/HLAcomplexes to achieve a high signal-to-noise ratio. The concentra-tions of complexes used were obtained by refolding each peptidewith heavy chain and b2m, whereas the UV-exchange methodusedhere yields lower concentrations of complexes but is requiredfor high throughput.

The Cancer Genome Atlas (TCGA) dataset analysisGene-expression and somatic mutation data of 266 skin

cutaneous melanoma samples from TCGA, which were pro-cessed by Broad Institute's TCGA workgroup (release dateOctober 10, 2013), were downloaded and analyzed further(22, 23). Briefly, 15,924 genes expressed in more than 80% oftumor samples were clustered using K-means clustering algo-rithm (k ¼ 12). Clusters containing 13 established T-cellsignatures transcripts (CD8A, CCL2, CCL3, CCL4, CXCL9,CXCL10, ICOS, GZMK, IRF1, HLA-DMA, HLA-DMB, HLA-DOA,and HLA-DOB) were selected to categorize 266 metastaticmelanoma into T-cell gene signature high (called T-cellinflamed or inflamed), T-cell gene signature intermediate, andT-cell gene signature low (called non–T-cell inflamed, or non-inflamed) tumor groups using ConsensusClusterPlus v.1.16.0(24). Genes differentially expressed between T-cell inflamedand non–T-cell-inflamed tumor groups were detected usingANOVA and filtered by false discovery rate. Synonymous sin-gle-nucleotide variants were excluded from consideration. Datafrom patients predicted to be HLA-A�0201-positive by twoindependent pipelines, ATHLATES (25) and PHLAT (26), wereused to predict neoantigen peptide sequences with the VariantEffect Predictor v78 (27). Wild-type and altered 17-merpairs were used to obtain 9-mer binding scores from theSYFPEITHI web server (28). In the present study, we selected86 (Supplementary Table S7) and 91 (Supplementary Table S8)neoantigen peptides from the inflamed and noninflamedtumor groups, respectively, for DSF analysis. Peptides werespecifically chosen because they spanned a comparably largerange (2–7,000 nmol/L) of predicted binding affinities usingNetMHC 4.0.

Statistical analysisFor comparisons between the inflamed and noninflamed

TCGA cohort data, a two-tailed Student t test was used to

Blaha et al.

Cancer Immunol Res; 7(1) January 2019 Cancer Immunology Research52




determine significance, with significance accepted at P < 0.05.To compare Tm with half-life and binding affinity parameters, thex-axis was set to a logarithmic scale. R2 values, based on the line ofbest fit, were determined using OriginPro 2017 software. Tmvalues and standard deviations reflect the average of a minimumof three replicates. Experimentally measured half-lives andstandard deviations are the average of two independent timecourses. Error bars for all figures indicate the standard deviationof replicates.

ResultsValidation of DSF as a high-throughput approach

A study has shown that DSF could be used with refolded andpurified peptide/class I and /class II MHC complexes to measurethermal stability (20). In order for this approach to be highthroughput, it is necessary to develop a process that does notrequire individual refolding andpurification of each complex. It isalso important to demonstrate that thermal stability is directlycorrelatedwith half-life, as the latter appears to be correlatedmostclosely with immunogenicity (10). To achieve these goals, weadapted the method of UV cleavage of a modified peptide/HLA-A2 complex, in thepresence of thepeptide of interest,which yieldspeptide/HLA-A2 complexes (16), scaling down the typical reac-tion volume to increase the number of complexes that could begeneratedby a single preparation ofUVpeptide/HLA-A2. In initialstudies,we examined thermal denaturationof complexes contain-ing a peptide from Wilms tumor antigen (WT-1, RMFPNAPYL).Comparison of complexes using the UV-exchange procedure(Fig. 1A) or refolded and purified WT-1/HLA-A2 complexes(Fig. 1B) showed similar denaturation profiles and Tm values(see Materials and Methods for details; Supplementary Fig. S1).The UV peptide/HLA-A2 complex itself yielded a Tm value of53.5�C (Fig. 1C). The replicates for these three complexes, con-ducted both in the same or different DSF experiments, werereproducible, with standard deviations of less than one degree(Supplementary Table S2).

To assess a range of Tm values, we next examined severaladditional peptides that are known to bind to HLA-A2 withdifferent affinities and that are well studied. These complexesincluded the MART-1 nonomer (AAGIGILTV), MART-1 modifieddecamer (ELAGIGILTV), and tyrosinase (YMDGTMSQV; Fig. 1D–

F). We also included a number of other commonly studied self-peptides (Fig. 1G). Tm values of the complexes ranged from 43�Cto 65�C, indicating that the method has a wide working range tocompare stabilities. Reproducibility of the method was alsoevident by results with different UV-exchanged preparations foreach of the three peptides, MART-1, WT-1, and tyrosinase(Fig. 1H). We have stored exchanged peptide complexes at 4�Cfor up to 10 months with identical DSF results to those from theinitially tested complex.

Effects of DMSO, peptide purity, and pH on thermaldenaturation

We examined the influence of three additional experimentalparameters. First, peptides are often dissolved in DMSO in orderto achieve greater solubility. To analyze the impact of DMSO, theUV-peptide/HLA-A2 complex was examined at various DMSOconcentrations (Fig. 2A). Increasing concentrations of DMSOyielded a reduction in observed Tm values, such that there wasapproximately a 2�Cdifference between1%and10%DMSO.Our

standard peptide exchange used a final DMSO concentration of2% v/v and our DSF assay used a final DMSO concentrationof 0.5% v/v.

In order to reduce the cost of commercial peptide synthesis, it isadvantageous to use minimal purification (often called "crude"peptide scale from many sources). We compared a number ofpeptides that were ordered at the small-scale crude level versus thesame peptides orderedwith greater than 90%purity (Fig. 2B). Theresults showed nearly identical Tm values with each of the twopreparations, suggesting that peptide purity levels for these 9-meror 10-mer peptides did not significantly impact Tm values.

Finally, we noticed about a 3�C lower Tm value for theUV-peptide complex when DSF was run at a pH of 8.4 ratherthan 7.4. Analysis at a range of pH values showed that thetemperature stability of complexes was reduced at an elevatedpH (Fig. 2C). This same trend was observed for a larger collectionof peptide complexes (Fig. 2D), indicating the need to analyzecomplexes at a consistent pH (from here on, we used pH7.4).

Tm and half-life comparisons of self-peptide variantsTo determine if the temperature stability of UV-exchanged

peptide/HLA-A2 complexes serves as a surrogate for half-lives,we used WT-1, MART-1, and their single-residue alanine variants.Because we have produced a high-affinity TCR against WT-1/HLA-A2 (18), we used the soluble TCR as a probe for dissociationof the peptide from the ternary complex. This approach wascompared with a similar experiment using anti-b2m as a probefor dissociation of b2m from the ternary complex, as has beenused to determine half-lives of neoantigen HLA-A2 complexes(10). Biotinylated peptide/HLA-A2 complexes were immobilizedon streptavidin-coated 5-mm microspheres, and the amount ofcomplex that remained after dissociation at 37�C, at various timepoints, was determined by flow cytometry with anti-b2m orsoluble TCR (Fig. 3A and B). Half-lives were determined byfirst-order kinetics of the dissociation curves, yielding values of15.1 hours and 14.7 hours, respectively. In order to examine acomplex with a predicted shorter half-life, we also assessed thehalf-life of the anchor-modified variant WT-1-M2A, whichyielded half-lives of 1.2 hours and 0.9 hours with anti-b2m andTCR probes (Fig. 3C and D). These results support the view thatthe entire ternary complex dissociated with approximately thesame kinetics.

We used DSF to determine the Tm values of WT-1 and MART-1alanine variants (Fig. 4A and B; Supplementary Table S3).Reduced Tm values, compared with the wild-type peptide, wouldbe expected for substitutions at anchor residues, such as the M2Asubstitution of WT-1. With M2A substitution, a 10�C lower Tmthan the wild-type was observed, and other variants predicted tobe at anchor residues also showed significantly lower Tm values(Fig. 4A). The structure of the WT-1 peptide within the WT-1/HLA-A2 complex (29) was shown to compare Tm values with theposition of the side chains within the pocket of the class Imolecule. Likewise, peptide variants of MART-1/HLA-A2 com-plexes were also analyzed for thermal stability, and Tm values areshown below the structure (30) of the MART-1 peptide (Fig. 4B).As expected, variants at anchor positions showed the largestreductions in Tm values. Changes from (or to) glycine oftenresulted in losses in stability, potentially due to conformationalchanges in backbone flexibility.

To directly compare the Tm values and half-lives of variouscomplexes, we measured half-lives of complexes that spanned






a roughly 20�C range of Tm values, using microspheres andanti-b2m (Supplementary Table S4). We found that the half-livesalso spanned a wide range (0.6–82 hours) and that these twoparameters were correlated (R2 ¼ 0.99; Fig. 4C).

Tm values correlate with reported half-lives of other HLA-A2complexes

Harndahl and colleagues have measured the binding affinitiesand half-lives of various pep/HLA-A2 complexes and have found

that for some peptides, the affinities for HLA-A2 do not correlatewith half-life of the complex (9). To determine if Tm valuescorrelated better with half-life or binding affinity, we synthesizeda collection of peptides reported by Harndahl and colleagues (9)and used DSF to examine thermal stability of the UV-peptide–exchanged complexes. The Tm values for these complexescorrelated with their reported half-lives (R2 ¼ 0.94; Fig. 5A;Supplementary Table S5). In contrast, a plot of Tm valuesagainst measured binding affinity yielded a lower degree of

Figure 1.

Thermal denaturation curves generatedfrom DSF of self-peptide/HLA-A2complexes. A, DSF denaturation profileof the WT-1 (RMFPNAPYL)/HLA-A2complex prepared throughUV-mediatedligand exchange at 4 mmol/L. For allmeasurements, the temperature ramprate was 1�C/minute from 25�C to 99�C.The curve was created using OriginPro2017 software to plot all 820 time pointsgenerated from QuantStudio real-timePCR. B, Thermal denaturation curve ofrefolded WT-1/HLA-A2 monomer at4 mmol/L, used for comparison withUV-generated complex. C,DSF profile ofrefolded UV peptide/HLA-A2. The UVpeptide (KILGFVFJV) contains3-amino-3-(2-nitro)phenyl-propionicacid, denoted with a J in the peptidesequence. D–F, DSF profiles forUV-exchanged self-peptide/HLA-A2complexes with MART-1 9-mer(AAGIGILTV), MART-1 anchor-modified10-mer (ELAGIGILTV), and tyrosinase(YMDGTMSQV). For A–F, curves arerepresentative of 3 independentexperiments, with three replicates perexperiment. G, Comparison of Tm valuesfrom several well-characterized self-peptides, including NYESO-1(SLLMWITNC) and HER2 (KIFGSLAFL).H, Comparison of Tm values from threeseparate DSF experiments with UV-exchanged self-peptide/HLA-A2complexes. UV-exchanged complexeswith tyrosinase (YMDGTMSQV),WT-1 (RMFPNAPYL), and MART-1(AAGIGILTV) were analyzed inthree separate experiments ofthree replicates each. DifferentUV-exchanged preparations were usedfor each trial. Error bars represent thestandard deviation between replicates ofindividual DSF runs.

Blaha et al.





correlation, whether using binding data from the approach ofAssarson and colleagues (R2 ¼ 0.26; ref. 31) or Harndahl andcolleagues (R2 ¼ 0.60; Fig. 5B and C; ref. 9). Two key peptidesfrom the study in reference (9) that illustrated the distinctionbetween affinity and half-life were the immunogenic peptide

FLTSVINRV and a nonimmunogenic peptide, NQNDNEETV.These two peptide complexes had identical measured bindingaffinities, yet had half-lives of 22.3 hours and 1.3 hours, respec-tively. These half-life values correlated with the Tm values of58.0�C and 46.4�C for these two complexes, respectively. Thus,

Figure 2.

Effects of DMSO, peptide purity, andpH on thermal denaturation of pep/HLA-A2 complexes. A, Tm values ofUV-peptide/HLA-A2 in PBS andvarious concentrations of dimethylsulfoxide (DMSO). The UV-exchangemethod uses peptides dissolved inDMSO, yielding final concentrations ofDMSO present at 0.5% in a typical DSFexperiment. B, Comparison of Tmvalues fromDSF of HLA-A2 complexesgenerated with either pure (90%) orcrude (as low as 30%) peptidepreparations. Correlation coefficient(R2 ¼ 0.99) was calculated usingOriginPro 2017. C, Tm values weredetermined by DSF at various pHvalues, using assay buffers containingHEPES, boric acid, or acetic acid.Unless otherwisenoted, all subsequentDSF experiments were performed atpH7.4usingHEPESas theassaybuffer.D, Comparison of Tm values (R2 ¼0.95) from various self-peptide/HLA-A2 and neoantigen/HLA-A2complexes at pH 7.4 and pH 8.4.Melting temperatures were increasedat 7.4 (compared with pH 8.4) by anaverage of 2.5�C. Error bars representthe average and standard deviation ofthree replicates in a single experiment.

Figure 3.

Dissociation measurements ofWT-1/HLA-A2 complexes using twodifferent probes. A, Dissociation curveof WT-1/HLA-A2 complexes usinganti-b2m as a probe (t1/2 ¼ 15.7).B, Dissociation curve of WT-1/HLA-A2complexes using soluble high-affinityTCR as a probe (t1/2 ¼ 14.4).C, Dissociation curve of WT-1–M2Acomplexes, a reduced stability variant(RAFPNAPYL), using anti-b2m(t1/2 ¼ 1.3). D, Dissociation curve ofWT-1–M2A/HLA-A2 using soluble TCR(t1/2 ¼ 1.0). In these experiments,refolded WT-1/HLA-A2 or WT-1–M2A/HLA-A2 were biotinylated with BirAligase, purified by HPLC ion exchangeandgelfiltration chromatography, andimmobilized on SA-coated 5-mmmicrospheres. Curves show data froma single experiment, representative oftwo independent experiments, withone replicate each. Half-life values arethe averages of two independentexperiments, with one replicate each.






stability asmeasuredby thedissociation of the complex correlatedwith thermal stability.

A study by Stronen and colleagues examined the half-lives of acollection of 57 neoantigen peptides derived from three melano-ma patients, concluding that stability (half-lives) correlated betterwith immunogenicity (12 of the 57 peptides testedwere judged tobe immunogenic) than affinities predicted by the NetMHC4.0algorithm (10). These 57 neoantigen peptides were synthesizedand UV-peptide–exchanged complexes were examined by DSF.A plot of Tm value versus half-life showed correlation (R2 ¼0.73; Fig. 5D; Supplementary Table S6). A lower correlationbetween Tm and half-life for this data set, relative to the correla-tions between Tm and half-life of the Harndahl and colleaguespeptides (Fig. 5A; ref. 9), may be attributed to those neoantigencomplexes at the low end of the half-life range in the Stronen andcolleagues study (ref. 10; i.e., it appears that the assay did notresolve half-lives below values of about 3 hours). If these valueswere eliminated, an R2 value of 0.81 was observed (Fig. 5D). Theprevious study (10) suggested that neoantigen peptides with half-lives greater than 5 hours represent a reasonable "cutoff" forchoice of peptides with a higher probability of being immuno-genic. The corresponding Tm value is about 53�C.

Correlation of Tm values with predicted binding affinitiesBecause prediction algorithms are often used as the sole

approach to identify potential neoantigen peptides that bind toMHC, we compared the Tm data described above to predictedbinding affinities using NetMHC4.0 (15). The plots for Harndahland colleagues' peptides (9), Stronen and colleagues' peptides(10), and the WT-1/MART-1 variants and other self-peptides areshown in Fig. 5E–G. The R2 values ranged from0.38 to 0.62, lowerthan correlations withmeasured half-lives. The lower correlationsobservedbetween thermal stability andpredicted affinity couldbedue to an actual difference in the two parameters (Tm and bindingaffinity, as suggested from data in Fig. 5B and C), the precision ofthe binding algorithm or a combination of both.

Tm values of HLA-A2 complexes containing neoantigenpeptides identified in TCGA

A study (22) analyzed whether there was a difference in thenumbers and HLA-A2–binding quality of mutated peptides iden-tified fromWES analysis of 266melanoma samples that exhibitedeither an inflamedor noninflamed tumormicroenvironment. Theresults suggested that no significant difference between the twosets of samples existed. In order to further examine these predictedneoantigens, we used DSF to analyze approximately 100 com-plexes from each inflamed or noninflamed data set, which hadsimilar ranges of predicted HLA-A2 binding affinities (Fig. 6A andB). Themeasured Tm values of these two sets ranged from 35�C to62�C,withdistributions thatwere similar between the two sampletypes (Fig. 6C andD). The correlations of Tm values and predictedbinding affinities (Fig. 6E and F) were similar to the correlationsfor peptides described above, providing further evidence that aDSF-based scan would allow improved identification of the moststable complexes. Statistical analysis using a two-tailed Studentt test (significance accepted at P < 0.05) showed no significantdifferences between Tm measurements of inflamed and nonin-flamed tumor neoantigens (P ¼ 0.39). This observation furthersupports the view that neoantigen quantity or quality alone doesnot account for ability or inability to generate an inflamed tumormicroenvironment.

Figure 4.

DSF analysis of self-peptides and single peptide variants. A, Comparison ofTm values for self-peptide WT-1 (RMFPNAPYL) and its alanine variants. Inpositions where an alanine was present in the wild-type sequence, it wasreplaced with a glycine (WT1-A6G). The structure of the WT-1 peptidefrom theWT-1/HLA-A2 structure is shown (PDB: 3HPJ). B, Tm values for MART-1anchor-modified 10-mer (ELAGIGILTV) and its alanine variants. In positionswhere an alanine was present in the wild-type sequence, it was replacedwith a glycine (MART1-A3G). The structure of the MART-1 10-mer peptide fromthe MART-1/HLA-A2 structure is shown (PDB: 1JF1). C, Correlation (R2 ¼ 0.99)between half-lives measured experimentally using the anti-b2m probe andTm values obtained from DSF. Error bars represent the average andstandard deviation of two independent experiments, with three replicatesper experiment.

Blaha et al.





Tm values of HLA-A2 complexes containing immunogenicneoantigen peptides

Several studies (12, 32–36) have identified sequences ofhuman neoantigen peptides that were found to elicit potent T-cellresponses. From these studies, we had 26 peptides synthesized

and examined them as HLA-A2 complexes using DSF. Althoughthe Tm values ranged from 43�C to 64�C (Table 1), 23 of the 26complexes (88%) exhibited Tm values above 53�C, the temper-ature stability that we determined corresponded to a half-lifeof 5 hours, and 22/26 of the peptide complexes (85%) exhibited

Figure 5.

Correlationof Tmvalueswith half-lives, predictedbinding affinities, or measured binding affinities.A, Plot of Tm values, determined by DSF, versushalf-lives for 23 T-cell epitopes restricted byHLA-A2 (9). Epitopes reported as immunogenicare shown in open squares (R2 ¼ 0.94). B andC, Peptides with known binding affinitymeasurements analyzed by Assarson et al. (31)and Harndahl et al. (9) were examined by DSFand Tm values were plotted versus IC50

measurements from Assarson et al. (R2 ¼ 0.26)or Harndahl et al. (R2¼ 0.60). Epitopes that theydefined as immunogenic are shown in opensquares.D,Plot of Tmvalues, determinedbyDSF,versus half-lives of 53 HLA-A2–restrictedneoantigens derived from three melanomapatients (10). The correlation (R2 ¼ 0.73)between published half-lives and measured Tmvalueswas increased (R2¼ 0.81) if data for thosepeptides that reached the shortest measurablehalf-life (10). Immunogenic neoantigens (12 of53) from the latter study are shown in opensquares. E, Correlation (R2 ¼ 0.62) betweenbinding affinity, predicted in silico withNetMHC 4.0, and experimentally measured Tmvalues for the 23 unique T-cell epitopes (9).Immunogenic epitopes are shown in opensquares. F, Correlation (R2 ¼ 0.39) betweenbinding affinity, predicted in silico with NetMHC4.0, and experimentally measured Tm values for53 melanoma-specific, HLA-A2–restrictedneoantigens (10). The 12 immunogenicneoantigens are shown in open squares.G, Correlation (R2 ¼ 0.40) between bindingaffinity, predicted in silico with NetMHC 4.0,and experimentally measured Tm values forHLA-A2–restricted self-peptides or their singlepeptide variants (Supplementary Table S2).






Tm values that were above 55�C. Thus, the majority of immuno-genic peptides, derived from multiple published studies, exhib-ited high thermal stability as HLA-A2 complexes.

DiscussionPeptide-binding algorithms have been developed for many of

the common HLA class I alleles (15, 37, 38). This has allowed forpredictions of whichmutated peptides identified in cancers couldpotentially bind to theHLA alleles expressed by a patient, with thegoal of developing vaccine strategies to elicit T-cell responsesagainst neoantigens expressed by cancer cells. In two vaccine trialsin melanoma patients (12, 13), the frequency of CD8þ T-cellresponses against these neoantigens was only 16% and 29%,respectively, indicating that improved methods for identifyingimmunogenic neoantigens are needed to improve patientresponses. Because vaccine strategies require limits on the numberof potential antigens (to date, between 10 and 20 peptides havebeen included), it is critical that those neoantigens with thehighest probability of success be identified.

A number of studies have suggested that dissociation ratesof peptide/HLA complexes are more accurate predictors ofimmunogenicity than binding affinity (e.g., refs. 9, 10). Although

some investigators have used the term "stability" to refer topepMHC dissociation rates, it should be noted that historically,stability studies of proteins use denaturants to quantitate thefolded to unfolded transition. Accordingly, temperature or chao-tropic agents (e.g., urea, guanidine) have been used to denatureproteins or complexes, and stability is reflected in the temperatureor the concentration of agent that yields 50% denaturation. Theseapproaches also facilitate direct comparisons among differentprotein systems. Studies of pepMHC that examine dissociationrates are more accurately defined as a measure of "kineticstability," as has been done by some investigators (39).

Here, we took the more conventional biochemical approachusing temperature denaturation to measure thermal stability ofthe pepMHC complexes. We demonstrated that 50% denatur-ation temperature (Tm) serves as a surrogate for dissociation rate.Our approach usingUV-exchangeable reactions at small scale alsofacilitates high throughput because thermal denaturation can bemonitored rapidly usingDSF in a standardRT-PCR thermal cycler.The DSF assay was validated using the HLA-A2 system, and acomprehensive collection of peptides from many previous stud-ies. We also conducted our own analysis of dissociation ratesusing a panel of self-peptides that we have studied in the contextof high-affinity TCR engineering (17, 40).

Figure 6.

Analysis of neoantigen/HLA-A2complexes identified from TCGA.A and B, Distribution of NetMHC 4.0values for HLA-A2–restrictedneoantigens from TCGA-indicatedinflamed and noninflamed tumormicroenvironments, respectively.C and D, Distribution of Tm values,obtained from DSF, for HLA-A2–restricted neoantigens frominflamed and noninflamed tumormicroenvironments, respectively.E, Relationship (R2 ¼ 0.53) betweenpredicted binding affinity and Tm for 85neoantigens found in inflamed tumormicroenvironments. F, Correlation(R2 ¼ 0.38) of predicted bindingaffinity and Tm for 91 neoantigensfound in noninflamed tumormicroenvironments. Analysis revealedno statistically significant differencebetween Tm values measured forpeptides from the two sources(inflamed and noninflamed). P ¼ 0.4assessed froma single experimentwiththree replicates of each neoantigenin the inflamed and noninflamedcohorts, using a two-tailed Studentt test (t ¼ 0.85, two degrees offreedom) with significance acceptedat P < 0.05).

Blaha et al.





HLA-A2 complexes exhibited Tm valueswith aworking range ofover 20�C and standard deviations that were generally less than0.5�C, even when complexes were prepared from different UV-exchange reactions. The Tm values using five complexes derivedfrom the UV-exchange procedure (MART-1-9-mer, WT1, MART1-10-mer, TAX, tyrosinase: 43.7�C, 57.8�C, 59.3�C, 61.9�C, 65.5�C)were very similar to the Tm values determined by another labwhere each complex was refolded with HLA-A2 individually(40.8�C, 56.3�C, 59.4�C, 66.3�C, 67.3�C, respectively; R2 value0.97; ref. 20). This indicated that the DSF approach is not onlyamenable to high-throughput UV exchange but that results arecomparable among different labs.

Tm values correlatedwith half-livesmeasured for a panel of self-peptides and single amino acid variants (R2¼ 0.99). As expected,lower Tm values were observed for peptides with substitutions atanchor residues.We also noted that peptides with substitutions toor from glycine, even in nonanchor residues, often yielded alteredTm values. The binding affinities of glycine-substituted peptideswere predicted less accurately than most other substitutions byNetMHC4.0. The ability of glycine residues to influence backboneconformations, and the impact on adjacent residues, likelyaccounts for these observations. This result also further illustratedthe value of conducting experimental stability assessments, ratherthan relying only on predictive algorithms.

Analysis of a collection of peptides, comprising other self-peptides and neoantigens, for which published half-lives weredetermined by independent methods also showed correlationwith Tm values. As experimentally measured binding affinitieswere available for some of these peptides, we were able to showthat thermal stability correlated more closely with dissociationrate than with affinity. A study analyzing peptides from thehuman tumor antigen NY-ESO-1 also showed weak correlationbetween half-life and either measured or predicted bindingaffinity (41).

These findings raise the mechanistic question of why thermalstability is correlated with kinetic stability, whereas eitherstability parameter has a lower correlation with measured orpredicted binding affinity. The likely explanation for the highcorrelation between thermal and kinetic stability is that eachmeasurement is performed with the ternary pepMHC complexand the same atomic interactions that control dissociation ofthe complex control its temperature denaturation. In contrast,binding affinities involve both association and dissociationrates of the complexes. Measurements of peptide affinities forMHC heavy chain/b2m are not trivial (42) and involve com-petition assays with labeled indicator peptides as pioneered bySette and colleagues (43), refolding reactions (44), or surfaceplasmon resonance (45). In any case, association rates ofpeptides are expected to impact the final measurement, andthese association rates likely vary among peptides. This possi-bility is supported by evidence that dynamics of the MHCmolecule or the peptide influence the interaction (46). Theconformational effects of such dynamics could impact theassociation rates. Hence, it is not surprising that the stabilityof the ternary complex (which is not influenced by associationrate) does not always correlate with affinity. Peptide-bindingaffinity methods are inherently complicated (42), as the MHCheavy chain/b2m dimer is not stable without a peptide, pre-venting direct binding studies. These assays are also sensitive toother conditions of the assay, as we have shown here with theimpact of pH on thermal stability.

The lower correlation between stability and predicted affinitycould be influenced by the same issues as measured affinity (e.g.,association rate contributes to affinity). However, an algorithmbased on stability data has also been developed (NetMHCStab,ref. 47), and this also showeda low correlationwhen the Tmvaluesmeasured here were plotted versus the predicted stability (R2 ¼0.49). The most likely explanation for this is that predictionalgorithms (affinity or stability) have insufficient training sets toencompass all sequence space. The thermal stability results withthe glycine-substituted peptides in our study illustrates an exam-ple of this problem. Although future algorithms might improvethese predictions, experimental measures of stability, such as theDSF method described here, are needed to analyze peptide/MHCcomplexes.

The residence time (48) of the ternary pepMHC complexon the cell surface is important physiologically because spe-cific T cells must have sufficient time to survey and interactwith the target cells. Such T cells are sensitive to the density ofthe antigenic complex (49). Antigen processing steps preced-ing surface expression of the complex are also critical (50).The ability to elicit tumor-reactive cytotoxic T lymphocytesrequires not only that the peptide be bound by a class Imolecule but that the protein is translated and processed insufficient quantities. Potential neoantigen peptides with highstability as a complex might not be expressed on the surfacebecause the peptide processing machinery might be unable tocleave or deliver the peptide efficiently or assemble the finalcomplex (50). Strategies to quantify these steps for eachpeptide would be valuable but are difficult to examine in ahigh-throughput format. The rules that govern whether apeptide is processed and presented are topics of interest, butimprovements in mass spectrometry methods to detect andquantitate peptides eluted from MHC should provide addi-tional insight (51–53). Finally, for a peptide to be immuno-genic, antigen-reactive T cells must exist in the repertoire torespond to the complex.

It has been suggested that immunogenicity often involvesTCR-mediated interactions with central residues of a peptide(e.g., positions 4 and 5), and that hydrophobic side chainsincrease the chance of immunogenicity (54, 55). Studies havedeveloped an immunogenicity "fitness" algorithm based onthe similarity of peptides to a library of known immunogenicforeign peptides (56). Immunogenicity can also be influencedby the degree of tolerance against self or mutated peptides,and we have argued that tolerance may be controlled by thesize of the repertoire of structurally related self-peptides (57).Once these factors are more fully understood, it should bepossible to develop algorithms that improve the success rate ofneoantigen-based vaccines.

In summary, the DSF assay will allow high-throughputexperimental identification of the most stable neoantigenpeptides, derived from WES/RNA-sequence data, at a scalerequired for treatment of larger patient populations. Forexample, on a research-level scale, the amount of refoldedUV-peptide/HLA-A2 protein is sufficient for performing DSFwith thousands of peptides, and the 384-well format used inmost RT-PCR machines will allow robotic assay development.Production of many different class I alleles will enable directcomparisons of Tm values across alleles, a feature that canaccommodate variation in predictive accuracy of bindingalgorithms among alleles.






Disclosure of Potential Conflicts of InterestT.F. Gajewski reports receiving commercial research funding from Merck,

Bristol-Myers Squibb, Seattle Genetics, Genentech, Ono, Aduro, Evelo, Bayer,and Incyte; has ownership interest in Jounce, Aduro, and Evelo; and is aconsultant/advisory board member for Aduro, FogPharma, Jounce, Adaptim-mune, FivePrime, and Sanofi. D. Kranz has ownership interest in BellicumPharmaceuticals, Agenus Inc., and Jounce Therapeutics and is a consultant/advisory board member for AbbVie. No potential conflicts of interest weredisclosed by the other authors.

Authors' ContributionsConception and design: D.T. Blaha, S.D. Anderson, D.M. KranzDevelopment of methodology: D.T. Blaha, S.D. AndersonAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): D.T. Blaha, S.D. Anderson, D.M. Yoakum, Y. Zha,T.F. GajewskiAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis):D.T. Blaha, S.D. Anderson,D.M. Yoakum,M.V.Hager,Y. Zha, T.F. Gajewski, D.M. KranzWriting, review, and/or revision of themanuscript:D.T. Blaha, S.D. Anderson,D.M. Yoakum, M.V. Hager, Y. Zha, T.F. Gajewski, D.M. Kranz

Administrative, technical, or material support (i.e., reporting or organizingdata, constructing databases): D.T. Blaha, S.D. Anderson, D.M. YoakumStudy supervision: D.M. Kranz

AcknowledgmentsThis work was supported by NIH grants CA178844 and CA187592

(to D.M. Kranz). We thank Abby Brown for experimental assistanceand Brian Baker and Lance Hellman (University of Notre Dame) forhelpful discussions about DSF analysis. We thank Barbara Pilas and theRoy J. Carver Biotechnology Center Flow Cytometry Facility and MarkBand of the Functional Genomics Unit for the use of their real-timePCR equipment.

The costs of publication of this article were defrayed in part by thepayment of page charges. This article must therefore be hereby markedadvertisement in accordance with 18 U.S.C. Section 1734 solely to indicatethis fact.

Received June14, 2018; revisedAugust 31, 2018; acceptedNovember 9, 2018;published first November 13, 2018.

References1. Gubin MM, Zhang X, Schuster H, Caron E, Ward JP, Noguchi T, et al.

Checkpoint blockade cancer immunotherapy targets tumour-specificmutant antigens. Nature 2014;515:577–81.

2. Ribas A, Wolchok JD. Cancer immunotherapy using checkpoint blockade.Science 2018;359:1350–5.

3. Schumacher TN, Schreiber RD. Neoantigens in cancer immunotherapy.Science 2015;348:69–74.

4. Blank CU, Haanen JB, Ribas A, Schumacher TN. The "cancer immuno-gram". Science 2016;352:658–60.

5. Schumacher TN, Hacohen N. Neoantigens encoded in the cancer genome.Curr Opin Immunol 2016;41:98–103.

6. Tran E, Robbins PF, Rosenberg SA. 'Final common pathway' of humancancer immunotherapy: targeting random somatic mutations.Nat Immunol 2017;18:255–62.

7. Hu Z, Ott PA, Wu CJ. Towards personalized, tumour-specific, therapeuticvaccines for cancer. Nat Rev Immunol 2018;18:168–82.

8. Sahin U, Tureci O. Personalized vaccines for cancer immunotherapy.Science 2018;359:1355–60.

9. Harndahl M, Rasmussen M, Roder G, Dalgaard Pedersen I, Sorensen M,Nielsen M, et al. Peptide-MHC class I stability is a better predictor thanpeptide affinity of CTL immunogenicity. Eur J Immunol 2012;42:1405–16.

10. Stronen E, Toebes M, Kelderman S, van BuurenMM, Yang W, van Rooij N,et al. Targeting of cancer neoantigens with donor-derived T cell receptorrepertoires. Science 2016;352:1337–41.

11. Editorial. The problem with neoantigen prediction. Nat Biotechnol2017;35:97.

12. Ott PA, Hu Z, Keskin DB, Shukla SA, Sun J, Bozym DJ, et al. An immu-nogenic personal neoantigen vaccine for patients with melanoma. Nature2017;547:217–21.

13. Sahin U, Derhovanessian E, Miller M, Kloke BP, Simon P, Lower M, et al.Personalized RNA mutanome vaccines mobilize poly-specific therapeuticimmunity against cancer. Nature 2017;547:222–6.

14. Niesen FH, Berglund H, Vedadi M. The use of differential scanningfluorimetry to detect ligand interactions that promote protein stability.Nat Protoc 2007;2:2212–21.

15. Andreatta M, Nielsen M. Gapped sequence alignment using artificialneural networks: application to the MHC class I system. Bioinformatics2016;32:511–7.

16. RodenkoB, ToebesM,Hadrup SR, van EschWJ,Molenaar AM, SchumacherTN, et al. Generation of peptide-MHC class I complexes through UV-mediated ligand exchange. Nat Protoc 2006;1:1120–32.

17. Sharma P, Kranz DM. Subtle changes at the variable domain interface ofthe T-cell receptor can strongly increase affinity. J Biol Chem 2018;293:1820–34.

18. Harris DT, Hager MV, Smith SN, Cai Q, Stone JD, Kruger P, et al.Comparison of T cell activities mediated by human TCRs and CARs thatuse the same recognition domains. J Immunol 2018;200:1088–100.

19. Zhang B, BowermanNA, Salama JK, Schmidt H, SpiottoMT, Schietinger A,et al. Induced sensitization of tumor stroma leads to eradication ofestablished cancer by T cells. J Exp Med 2007;204:49–55.

20. Hellman LM, Yin L, Wang Y, Blevins SJ, Riley TP, Belden OS, et al.Differential scanning fluorimetry based assessments of the thermal andkinetic stability of peptide-MHC complexes. J Immunol Methods2016;432:95–101.

21. HuynhK, Partch CL. Analysis of protein stability and ligand interactions bythermal shift assay. Curr Protoc Protein Sci 2015;79:28.9.1–14.

22. Spranger S, Luke JJ, Bao R, Zha Y, Hernandez KM, Li Y, et al. Density ofimmunogenic antigens does not explain the presence or absence of theT-cell-inflamed tumor microenvironment in melanoma. Proc Natl AcadSci USA 2016;113:E7759–e68.

23. Spranger S, BaoR,Gajewski TF.Melanoma-intrinsic beta-catenin signallingprevents anti-tumour immunity. Nature 2015;523:231–5.

24. Wilkerson MD, Hayes DN. ConsensusClusterPlus: a class discovery toolwith confidence assessments and item tracking. Bioinformatics 2010;26:1572–3.

25. Liu C, Yang X, Duffy B, Mohanakumar T, Mitra RD, Zody MC, et al.ATHLATES: accurate typing of human leukocyte antigen through exomesequencing. Nucleic Acids Res 2013;41:e142.

26. Bai Y, Ni M, Cooper B, Wei Y, Fury W. Inference of high resolution HLAtypes using genome-wide RNA or DNA sequencing reads. BMC Genomics2014;15:325.

27. McLarenW, Pritchard B, Rios D, Chen Y, Flicek P, Cunningham F. Derivingthe consequences of genomic variantswith the Ensembl API and SNPEffectPredictor. Bioinformatics 2010;26:2069–70.

28. Rammensee H, Bachmann J, Emmerich NP, Bachor OA, Stevanovic S.SYFPEITHI: database for MHC ligands and peptide motifs. Immunoge-netics 1999;50:213–9.

29. Borbulevych OY, Do P, Baker BM. Structures of native and affinity-enhanced WT1 epitopes bound to HLA-A�0201: implications for WT1-based cancer therapeutics. Mol Immunol 2010;47:2519–24.

30. Sliz P, Michielin O, Cerottini JC, Luescher I, Romero P, Karplus M, et al.Crystal structures of two closely related but antigenically distinct HLA-A2/melanocyte-melanoma tumor-antigen peptide complexes. J Immunol2001;167:3276–84.

31. Assarsson E, Sidney J, Oseroff C, Pasquetto V, Bui HH, Frahm N, et al.A quantitative analysis of the variables affecting the repertoire of T cellspecificities recognized after vaccinia virus infection. J Immunol 2007;178:7890–901.

Blaha et al.





32. Fritsch EF, Rajasagi M, Ott PA, Brusic V, Hacohen N, Wu CJ. HLA-bindingproperties of tumor neoepitopes in humans. Cancer Immunol Res2014;2:522–9.

33. Robbins PF, Lu YC, El-Gamil M, Li YF, Gross C, Gartner J, et al. Miningexomic sequencing data to identify mutated antigens recognized by adop-tively transferred tumor-reactive T cells. Nat Med 2013;19:747–52.

34. CohenCJ, Gartner JJ, Horovitz-FriedM, Shamalov K, Trebska-McGowanK,Bliskovsky VV, et al. Isolation of neoantigen-specific T cells from tumor andperipheral lymphocytes. J Clin Invest 2015;125:3981–91.

35. Prickett TD, Crystal JS, Cohen CJ, Pasetto A, Parkhurst MR, Gartner JJ, et al.Durable complete response from metastatic melanoma after transfer ofautologous T cells recognizing 10 mutated tumor antigens. Cancer Immu-nol Res 2016;4:669–78.

36. Durgeau A, Virk Y, Corgnac S, Mami-Chouaib F. Recent advances intargeting CD8 T-cell immunity for more effective cancer immunotherapy.Front Immunol 2018;9:14.

37. Jurtz V, Paul S, Andreatta M,Marcatili P, Peters B, NielsenM. NetMHCpan-4.0: improved peptide-MHC class I interaction predictions integratingeluted ligand and peptide binding affinity data. J Immunol 2017;199:3360–8.

38. Rasmussen M, Fenoy E, Harndahl M, Kristensen AB, Nielsen IK,Nielsen M, et al. Pan-specific prediction of peptide-MHC class Icomplex stability, a correlate of T cell immunogenicity. J Immunol2016;197:1517–24.

39. Lazarski CA, Chaves FA, Jenks SA, Wu S, Richards KA, Weaver JM, et al. Thekinetic stability of MHC class II: peptide complexes is a key parameter thatdictates immunodominance. Immunity 2005;23:29–40.

40. Smith SN, Sommermeyer D, Piepenbrink KH, Blevins SJ, Bernhard H,Uckert W, et al. Plasticity in the contribution of T cell receptor variableregion residues to binding of peptide-HLA-A2 complexes. J Mol Biol2013;425:4496–507.

41. Schmidt J, Guillaume P, Dojcinovic D, Karbach J, Coukos G, Luescher I. Insilico and cell-based analyses reveal strong divergence between predictionand observation of T-cell-recognized tumor antigen T-cell epitopes.J Biol Chem 2017;292:11840–9.

42. Margulies DH, Corr M, Boyd LF, Khilko SN. MHC class I/peptide interac-tions: binding specificity and kinetics. J Mol Recognit 1993;6:59–69.

43. Sidney J, Southwood S, Moore C, Oseroff C, Pinilla C, Grey HM, et al.Measurement ofMHC/peptide interactions by gel filtration ormonoclonalantibody capture. Curr Protoc Immunol 2013;Chapter 18:Unit 18.3.

44. Harndahl M, Justesen S, Lamberth K, Roder G, Nielsen M, Buus S. Peptidebinding to HLA class I molecules: homogenous, high-throughput screen-ing, and affinity assays. J Biomol Screen 2009;14:173–80.

45. Khilko SN, Corr M, Boyd LF, Lees A, Inman JK, Margulies DH. Directdetection of major histocompatibility complex class I binding to antigenicpeptides using surface plasmon resonance. Peptide immobilization andcharacterization of binding specificity. J Biol Chem 1993;268:15425–34.

46. Ayres CM, Corcelli SA, Baker BM. Peptide and peptide-dependent motionsin MHC proteins: immunological implications and biophysical under-pinnings. Front Immunol 2017;8:935.

47. Jorgensen KW, RasmussenM, Buus S, NielsenM. NetMHCstab - predictingstability of peptide-MHC-I complexes; impacts for cytotoxic T lymphocyteepitope discovery. Immunology 2014;141:18–26.

48. Tummino PJ, Copeland RA. Residence time of receptor-ligand complexesand its effect on biological function. Biochemistry 2008;47:5481–92.

49. Chervin AS, Stone JD, Holler PD, Bai A, Chen J, Eisen HN, et al. The impactof TCR-binding properties and antigen presentation format on T cellresponsiveness. J Immunol 2009;183:1166–78.

50. Blum JS, Wearsch PA, Cresswell P. Pathways of antigen processing. AnnuRev Immunol 2013;31:443–73.

51. Bassani-Sternberg M, Coukos G. Mass spectrometry-based antigen discov-ery for cancer immunotherapy. Curr Opin Immunol 2016;41:9–17.

52. Muller M, Gfeller D, Coukos G, Bassani-Sternberg M. 'Hotspots' of antigenpresentation revealed by human leukocyte antigen ligandomics for neoan-tigen prioritization. Front Immunol 2017;8:1367.

53. CreechAL, Ting YS,Goulding SP, Sauld JFK, BarthelmeD, RooneyMS, et al.The role of mass spectrometry and proteogenomics in the advancement ofHLA epitope prediction. Proteomics 2018;18:e1700259.

54. Calis JJ,MaybenoM,Greenbaum JA,WeiskopfD,De SilvaAD, Sette A, et al.Properties of MHC class I presented peptides that enhance immunogenic-ity. PLoS Comput Biol 2013;9:e1003266.

55. Chowell D, Krishna S, Becker PD, Cocita C, Shu J, Tan X, et al. TCR contactresidue hydrophobicity is a hallmark of immunogenic CD8þ T cellepitopes. Proc Natl Acad Sci USA 2015;112:E1754–62.

56. Luksza M, Riaz N, Makarov V, Balachandran VP, Hellmann MD, SolovyovA, et al. A neoantigenfitnessmodel predicts tumour response to checkpointblockade immunotherapy. Nature 2017;551:517–20.

57. Stone JD, Harris DT, Kranz DM. TCR affinity for p/MHC formed by tumorantigens that are self-proteins: impact on efficacy and toxicity. Curr OpinImmunol 2015;33:16–22.






2019;7:50-61. Published OnlineFirst November 13, 2018.Cancer Immunol Res Dylan T. Blaha, Scott D. Anderson, Daniel M. Yoakum, et al. Complexes Improves Immunogenicity PredictionsHigh-Throughput Stability Screening of Neoantigen/HLA

Updated version

10.1158/2326-6066.CIR-18-0395doi:

Access the most recent version of this article at:

Material

Supplementary

http://cancerimmunolres.aacrjournals.org/content/suppl/2018/11/13/2326-6066.CIR-18-0395.DC1

Access the most recent supplemental material at:

Cited articles

http://cancerimmunolres.aacrjournals.org/content/7/1/50.full#ref-list-1

This article cites 56 articles, 19 of which you can access for free at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected]

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://cancerimmunolres.aacrjournals.org/content/7/1/50To request permission to re-use all or part of this article, use this link



http://cancerimmunolres.aacrjournals.org/lookup/doi/10.1158/2326-6066.CIR-18-0395

http://cancerimmunolres.aacrjournals.org/content/suppl/2018/11/13/2326-6066.CIR-18-0395.DC1

http://cancerimmunolres.aacrjournals.org/content/7/1/50.full#ref-list-1

http://cancerimmunolres.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://cancerimmunolres.aacrjournals.org/content/7/1/50


High-Throughput Stability Screening of Neoantigen/HLA … · stability could be important, as only a small fraction of peptides in most current vaccines are capable of eliciting CD8þ

Documents