Kaposi sarcoma-associated herpesvirus (KSHV), also called human herpesvirus-8 (HHV-8), is the causative agent of Kaposi sarcoma (KS), a vascular tumor of endothelial origin that predominantly affects immunocompromised individuals. KSHV also drives multicentric Castleman's disease (MCD), primary effusion lymphoma (PEL), and KSHV inflammatory cytokine syndrome (KICS). The virus is unevenly distributed globally: sub-Saharan Africa (SSA) carries the highest combined burden of HIV and KSHV co-infection, which directly explains why SSA accounts for the highest incidence of KS worldwide. Within SSA, KS falls into two clinical presentations: endemic KS (HIV-negative, historically affecting older males) and epidemic KS (HIV-positive, driving the dramatic rise in incidence tied to the AIDS epidemic).
A central diagnostic challenge in KS management is the inability to reliably distinguish active KS from asymptomatic KSHV infection using serological testing. Both populations mount immune responses against the virus, yet not all KSHV-seropositive individuals develop KS. Identifying specific antibody (Ab) signatures that mark disease progression would be transformative for early detection and risk stratification of people living with HIV (PLWH) who are at elevated risk of developing KS. This study undertook a proteome-wide approach, profiling Ab responses against a library of 1,988 peptides derived from the entire KSHV proteome to find molecular features that discriminate KS patients from asymptomatic infected individuals.
Prior serological approaches: Previous KSHV serology tools relied on immunofluorescence assays (IFAs) against infected cell lines and ELISA-based assays using recombinant proteins, particularly the latency-associated nuclear antigen (LANA/ORF73) and the glycoprotein K8.1. While these confirmed KSHV infection, they could not reliably distinguish KS disease from asymptomatic infection, nor could they pinpoint the specific epitopes driving Ab responses at high resolution. Protein-level immunodominance had been established (K8.1, ORF65, ORF38, ORF73 are the top antigenic proteins), but the peptide-level epitopes within those proteins were largely undefined before this study.
Study context: The cohort comprised 59 KS patients and 22 asymptomatic KSHV-infected individuals from Zambia and Tanzania, with both HIV-negative and HIV-positive individuals represented in each group. This design allowed simultaneous assessment of the independent effects of KS disease status and HIV co-infection on the anti-KSHV Ab repertoire, a critical distinction because HIV suppresses immune function in ways that could confound serological markers.
The study employed VirScan, a phage immunoprecipitation and sequencing (PhIP-Seq) platform developed for proteome-wide Ab profiling. The T7-Vir3 phage library was constructed to display 56-amino acid peptides with 28-amino acid overlaps tiled across the complete KSHV proteome, producing 1,988 unique peptides. This 56-mer design is long enough for peptides to adopt partial secondary structure while remaining amenable to the bacterial phage display system. Importantly, each plasma sample was run in duplicate to ensure reproducibility, and eight mock precipitations (PBS without plasma) were run on each plate to identify non-specific phage binding.
PhIP-Seq workflow: Normalized quantities of total IgG (2 micrograms per sample) were incubated with the phage library. Ab-phage complexes were then immunoprecipitated using a 1:1 mixture of Protein A and Protein G Dynabeads (Invitrogen), which capture all IgG subclasses. After immunoprecipitation, phage DNA was released, barcoded with donor-specific identifiers, and amplified through three sequential PCR rounds incorporating Illumina sequencing adapters. Libraries were pooled at equimolar concentrations, gel-purified, and sequenced on an Illumina NextSeq 550 at the University of Nebraska Medical Center Genomics Core.
Bioinformatics pipeline: Raw FASTQ reads were aligned to the library oligonucleotide reference using bowtie indices via the PhIP-stat repository. Normalized read counts were fit to a Gamma-Poisson distribution, generating a peptide-by-sample matrix of residual p-values. Peptides were called reactive only if they showed statistically significant enrichment in both duplicates (p < 0.05). A quality control filter removed peptides reactive in more than 25% of mock IPs, eliminating non-specific binders. Only replicates with Pearson r > 0.7 were included in downstream analyses, ensuring data quality across the 81-sample cohort.
Quantification metrics: Four distinct measures were used to characterize the Ab response: (1) percent reactivity, the proportion of individuals reactive to a given peptide or protein; (2) breadth, the total number of unique reactive peptides per individual; (3) relative breadth, breadth normalized by the number of library peptides per protein (capturing coverage extent); and (4) magnitude, quantified as MLXP (the -log10 of the p-value), representing signal intensity of Ab binding. These four metrics together enabled a multi-dimensional portrait of each individual's anti-KSHV Ab repertoire.
A key finding was that KS patients showed significantly higher magnitude of anti-KSHV Ab responses compared to asymptomatic individuals, but not significantly higher breadth. While there was a trend toward more reactive peptides in KS, the difference in total number of reactive peptides did not reach statistical significance. This finding implies that upon KS development, the humoral response focuses on existing epitopes (increasing intensity against already-recognized targets) rather than broadening to include new ones. This is consistent with an anamnestic response driven by increased KSHV lytic replication and antigen exposure during tumorigenesis.
Protein-level hierarchy: Proteins in the latency, structural, and glycoprotein functional groups were significantly more targeted than expression/replication and immunomodulation proteins. The most immunodominant proteins identified, in decreasing order of percent reactivity and magnitude, were ORF73 (LANA), K8.1/K8.1B, ORF65 (small capsid protein), and ORF38 (tegument protein). This ordering recapitulated findings from previous protein bead array and microarray studies, validating the VirScan platform for KSHV serology. Notably, K15, a multiply-spanning transmembrane protein involved in signaling, also emerged as a high-reactivity target in this cohort.
Relationship to Ab titer: Prior IFA measurements of total anti-KSHV Ab end-point titer were available for the cohort. Breadth and magnitude both correlated significantly with Ab titer, but with only moderate Spearman correlation coefficients (rs = 0.40 for breadth and rs = 0.54 for magnitude). This modest correlation indicates that total Ab titer does not fully capture the peptide-level breadth or intensity of the anti-KSHV response, supporting the value of high-resolution PhIP-Seq over conventional titer assays for interrogating individual responses.
Age effects: Since endemic KS patients were significantly older than epidemic KS patients (reflecting the underlying epidemiology of HIV-unassociated KS), a potential confound existed. However, analysis revealed no significant correlation between age and breadth or magnitude of anti-KSHV response (Spearman rho = 0.24 and 0.18, respectively), indicating that age differences did not explain the immune response differences observed between EnKS and EpKS groups.
A major contribution of this study was defining the specific peptide-level epitopes within immunodominant KSHV proteins. For ORF65 (the small capsid protein), peptide-level Ab responses were mapped at amino acid resolution across the full protein. The dominant epitope was localized to residues 140-168 (ORF65140-168), which was highly recognized in both KS and asymptomatic individuals. Sequences of ORF65 peptides in the library were first aligned to a Zambian consensus sequence to confirm the library represented circulating KSHV strains in the study cohort, validating the geographic relevance of epitope data.
LANA (ORF73): LANA is the most immunodominant KSHV protein and the gold standard for KS tissue diagnosis (detected by immunofluorescence in tumor biopsies). In prior work by this group, the dominant LANA epitope was localized to a region bridging the central repeat region into the conserved C-terminal domain. In this study, seven discriminative peptides in the LANA C-terminus were identified as KS-predictive in the machine learning models, with the region spanning ORF73997-1053 and ORF731016-1072 forming part of the immunodominant C-terminal epitope cluster.
K8.1 glycoprotein: K8.1 is widely used in KSHV seroprevalence studies, but its epitopes had not been previously mapped at peptide resolution. This study found that K8.1 was second only to LANA in percent reactivity. Three splice variants are present in the library: unspliced K8.1, K8.1A, and K8.1B. The dominant immunogenic region mapped to residues 29-56 (K8.129-56), with KS patients showing significantly higher magnitude of response to K8.1B compared to asymptomatic individuals. Three K8.1-derived peptides spanning this immunodominant epitope were identified as KS-predictive discriminative features.
K15 transmembrane protein: K15 (Type P allele) is a multiply-spanning membrane protein with signaling functions and has been proposed as a therapeutic target. The study found high breadth and magnitude of Ab responses to K15 overall, with a focal point of antigenicity at residues 421-448 (K15421-448, sequence: VNRDPPNVFGYASILVSGAEESREPSPQ) in KS patients, but minimal reactivity in asymptomatic individuals. Structural prediction placed this immunodominant region in the cytoplasmic tail of K15, meaning it is not surface-accessible on intact virions or infected cells, limiting its therapeutic utility as a neutralizing target despite its discriminative serological value.
Disentangling the effects of KS disease from HIV co-infection required stratified analysis, since approximately half of KS patients in the cohort were HIV-positive (epidemic KS, EpKS) and half were HIV-negative (endemic KS, EnKS). The stratification revealed that overall breadth and magnitude of anti-KSHV responses were lower in EpKS compared to EnKS, while there was no significant difference between HIV-negative asymptomatic and HIV-positive asymptomatic individuals. This asymmetry suggests that HIV co-infection specifically diminishes the KSHV Ab response in the context of active KS, rather than broadly suppressing anti-KSHV immunity in asymptomatic infection.
KS-specific effects by HIV status: When KS and asymptomatic groups were compared after stratification, KS-specific elevations in ORF73 and K15 reactive peptide counts were significant in endemic (HIV-negative) KS versus HIV-negative asymptomatic individuals. For epidemic (HIV-positive) KS, the K8.1 breadth and magnitude were significantly elevated compared to HIV-positive asymptomatic individuals. This protein-specific differential suggests that the immunodominance hierarchy in the context of KS is modulated by HIV status, and that diagnostic markers may need to be adapted for HIV-positive versus HIV-negative KS populations.
Epitope-level resilience: Critically, HIV co-infection had little to no effect on percent reactivity and average magnitude against the dominant epitopes of K8.1 and ORF65. The core epitope responses to these two proteins were therefore robust regardless of HIV status. Variability was observed in K15 responses among KS patients and ORF38 responses among asymptomatic individuals when HIV status was considered, suggesting these proteins' Ab patterns are more sensitive to immune status. Approximately 50% of HIV-positive participants were aviremic with normal CD4 counts at time of sampling, allowing partial assessment of antiretroviral therapy-mediated immune reconstitution effects.
Principal component analysis: To assess whether protein-level responses could discriminate KS from ASY, PCA was applied to protein-level breadth and magnitude matrices. Protein-level breadth explained only 14.5% of total variance in the first two principal components. Protein-level magnitude performed better at 29.7% variance explained, and showed better visual separation between KS and ASY clusters in PCA space. However, neither protein-level metric achieved clean separation, motivating the shift to peptide-level features and machine learning for classification.
The identification of 25 discriminative KSHV peptides capable of classifying KS patients from asymptomatic individuals was the central translational finding of this study. Because protein-level PCA showed insufficient discrimination, the authors moved to peptide-level binary classification. The full 81-sample dataset was partitioned into training (75%, n = 60: 12 EnKS, 32 EpKS, 8 HIV- ASY, 8 HIV+ ASY), validation (hold-out within training), and test (25%, n = 21: 4 EnKS, 11 EpKS, 3 HIV- ASY, 3 HIV+ ASY) sets, stratified by both disease status and HIV co-infection status to prevent enrichment of any subgroup in the unseen test set.
Feature selection and model training: Five independent resampled datasets were created from the training set for building predictive models. In each, univariate feature selection used a two-proportions z-test with continuity correction (prop.test in R) to identify peptides differentially reactive between KS and ASY (p < 0.05). The resulting feature sets were used to train five candidate classifiers. Performance was evaluated using the area under the receiver operating characteristic curve (AUC), positive predictive value (PPV), and negative predictive value (NPV). The test set was held out completely until final model selection to ensure unbiased performance estimates.
Model 5 performance: Among the five candidate models, Model 5 achieved the strongest and most consistent performance on the held-out test set, with a median AUC of 0.967. This near-unity AUC indicates near-complete discrimination between KS and ASY based on peptide-level Ab reactivity alone. Sixteen of the 25 features in Model 5 were shared across all five models, representing a highly stable core discriminative feature set. The remaining nine peptides unique to Model 5 included: K11-56, ORF44477-532, ORF19281-336, ORF25449-504, ORF63281-336, ORF64337-392, ORF731016-1072, ORF73997-1053, and ORF65113-168, covering structural, tegument, and regulatory KSHV proteins.
Biological composition of the 25 peptides: The 25 discriminative peptides were not random across the proteome; they encoded biologically coherent signatures. KS-predictive peptides included three spanning the K8.1 immunodominant epitope, seven spanning the LANA C-terminus immunodominant cluster, one immediately following the ORF65 immunodominant epitope, and two encompassing the K15 immunodominant region. In contrast, ASY-defining peptides were derived from non-immunodominant proteins across multiple functional groups (immunomodulation, replication, and structural), suggesting that asymptomatic infection maintains a broader, more distributed Ab response while KS focuses the response on a subset of dominant targets.
The VirScan PhIP-Seq platform is inherently restricted to linear or quasi-linear, continuous epitopes. Because the phage library is expressed in bacterial cells, peptides lack the mammalian post-translational modifications (glycosylation, phosphorylation, ubiquitination) present on native KSHV proteins. Neutralizing antibodies against KSHV glycoproteins such as gB (ORF8), gH (ORF22), gL (ORF47), gM (ORF39), and gN (ORF53) frequently target conformational epitopes or glycan-dependent structures that are invisible to a 56-mer linear peptide library. The study detected only private (individual-specific) Ab responses to these glycoproteins beyond K8.1, consistent with the known dependence of anti-gB and anti-gH/gL neutralizing Abs on quaternary protein structure. This means the discriminative 25-peptide panel almost certainly does not capture the full immunological picture of KS pathogenesis.
Cohort size and composition: The cohort of 81 individuals (59 KS, 22 ASY) is modest relative to what would be required for rigorous clinical validation. The 25% test set comprised only 21 individuals, meaning the AUC = 0.967 figure is derived from a small sample and may be subject to optimistic bias, particularly given class imbalance (15 KS vs. 6 ASY in the test set). The authors explicitly acknowledge that future studies with larger, longitudinal sample sets within and outside of SSA will be needed to validate these findings. The cohort also consisted exclusively of cutaneous KS cases, with no PEL, MCD, or IRIS cases, limiting generalizability to other KSHV-associated diseases.
Epitope resolution constraints: The 28-amino acid overlap between consecutive 56-mer peptides means that a reactive peptide may encompass more than one epitope. The study can localize immunodominance to a 28-amino acid window, but cannot resolve individual epitopes within that window. Fine-mapping with shorter overlapping peptides or alanine scanning would be required to define the minimal epitope precisely. This limitation has downstream consequences for vaccine design, where minimal-epitope precision is typically required.
Mucosal immunity not assessed: The study analyzed only plasma-derived IgG Abs. KSHV is transmitted primarily through saliva in SSA, and mucosal IgA responses in saliva or other mucosal secretions may play important roles in infection control and disease prevention. The contribution of mucosal immunity to the KS/ASY distinction therefore remains unexplored. Similarly, the functional properties of the identified Abs (IgG subclass distribution, affinity, avidity, ADCC activity, neutralization capacity) were not determined, limiting the ability to distinguish immunologically protective from non-protective responses.
The authors identify the 25 discriminative KSHV peptides as strong candidates for development as a prognostic screening tool targeting asymptomatic KSHV-infected individuals, particularly PLWH. The clinical imperative is clear: KS is far more treatable when detected early, and the high incidence of KSHV infection among PLWH in SSA means a reliable serological test that predicts KS development before overt tumor formation could substantially reduce KS-associated mortality. The PPV and NPV of the 25-peptide model were both noted as strong, meaning the panel correctly classified both KS-positive and KS-negative cases with high confidence. Longitudinal studies tracking asymptomatic KSHV-seropositive individuals over time will be required to determine whether the 25-peptide signature predicts conversion to KS before clinical presentation.
Vaccine implications: The immunodominant epitopes identified here, particularly K8.129-56 and ORF65140-168, are natural targets of strong Ab responses and are therefore immediate candidates for inclusion in KSHV vaccine constructs. While the cytoplasmic location of K15421-448 limits its use as an immunogen for neutralizing Ab induction, the LANA C-terminal epitopes and the K8.1 surface glycoprotein epitopes represent accessible and immunogenic targets. The study also draws attention to an unexplored area: since intrabodies (intracellularly expressed antibody fragments) targeting KSHV LANA and viral IL-6 (vIL-6) have shown pre-clinical therapeutic potential, the precise epitopes identified here could inform intrabody design against the LANA C-terminus.
Non-neutralizing Ab functions: Several non-neutralizing Ab effector mechanisms warrant investigation in follow-up work. Ab-dependent cell cytotoxicity (ADCC) has been detected in KSHV-seropositive individuals, but no difference in ADCC-mediating Ab prevalence was found between KS and ASY in prior work by this group. The 25-peptide panel now provides a substrate for testing whether Abs to specific epitopes correlate with ADCC or other Fc-mediated effector activities. Natural killer (NK) cell functional differences between KS and ASY populations could also influence disease progression and interact with specific anti-KSHV Ab populations.
Expanding the model: The authors suggest expanding the predictive model to include responses to other known human pathogens, since the VirScan platform includes libraries from multiple viral and bacterial pathogens and thus enables concurrent profiling of the broader pathogen-specific Ab repertoire. Incorporating other co-pathogen responses (EBV, CMV, HIV itself) could improve model performance by capturing immune context variables that modulate KSHV-specific responses. Geographic validation outside SSA (including classic KS among Mediterranean populations and HIV-negative MSM KS) will be essential before any diagnostic or prognostic tool could be generalized.