Kaposi’s sarcoma-associated herpesvirus T cell responses in HIV seronegative individuals from rural Uganda

T cell responses to Kaposi’s sarcoma-associated herpesvirus (KSHV) are likely essential in the control of KSHV infection and protection from associated disease, but remain poorly characterised. KSHV prevalence in rural Uganda is high at >90%. Here we investigate IFN- γ T cell responses to the KSHV proteome in HIV-negative individuals from a rural Ugandan population. We use an ex-vivo IFN- γ ELISpot assay with overlapping peptide pools spanning 83 KSHV open reading frames (ORF) on peripheral blood mononuclear cells (PBMC) from 116 individuals. KSHV-specific T cell IFN- γ responses are of low intensity and heterogeneous, with no evidence of immune dominance; by contrast, IFN- γ responses to Epstein–Barr virus, Cytomegalovirus and influenza peptides are frequent and intense. Individuals with KSHV DNA in PBMC have higher IFN- γ responses to ORF73 (p = 0.02) and lower responses to K8.1 (p = 0.004) when compared with those without KSHV DNA. In summary, we demonstrate low intensity, heterogeneous T cell responses to KSHV in immune-competent individuals.

T he adaptive response to Kaposi sarcoma-associated herpesvirus (KSHV) is important in protecting from KSHVassociated malignancies, i.e., Kaposi's sarcoma (KS), multicentric Castleman's disease and primary effusion lymphoma 1,2 . This is shown by the substantial increase in the risk of developing KS-and other KSHV-associated malignancies in immunocompromised individuals 3,4 . A higher frequency of KSHV-specific T cells has been reported in KS patients who achieve complete remission compared to cases with unresolved KS [5][6][7] . In comparison to KS patients, more frequent and higher KSHV-specific T cell responses have been reported in asymptomatic individuals 8 . Treatment with highly active antiretroviral therapy of AIDS-KS patients has been associated with an improved KSHV-specific T cell response, correlating with KS regression 9,10 . Furthermore, the pivotal role of T cells in controlling other herpesviruses, including Epstein-Barr virus (EBV), CMV and HSV, has been well investigated 11,12 , although similar evidence (specific requirements and immunodominant epitopes) for KSHV is less well understood. Understanding T cell responses to KSHV is important for the development of KSHV vaccines and immunotherapy for KS.
During the adaptive immune response to viral infections, CD8+ T cells play a vital role in killing virally-infected cells and producing effector cytokines, while CD4+ T cells produce cytokines to enhance CD8+ T cells cytotoxic killing of infected cells. Several studies have reported a predominant role of cytotoxic CD8+ T cells in the control of classic, iatrogenic and AIDSrelated KS 1,2,13,14 . However, these studies have focused on HLA-A2 epitopes, a frequent haplotype in Caucasians 15 , and were performed in mostly, non-KS endemic areas, with only two studies reporting cellular responses to KSHV in sub-Saharan Africa and only then in a limited number of study participants 16,17 .
The~165 kB double-stranded DNA KSHV genome encodes about 90 proteins 18 with a very high number of potential T cell epitopes in each protein. Previous studies have identified epitopes in individual participants, but these have not been confirmed in other studies. We developed a systematic approach using overlapping peptides to the entire KSHV proteome to investigate KSHV cellular immunity in US healthy donors and patients with KSHV-related disease 19,20 . We made peptide pools corresponding to each KSHV open reading frame (ORF) and measured responses by IFN-γ ELISpot. We showed that participants responded typically to between one and five peptide pools but, that there was no overlap between participants in the peptide pools recognised. Responses were heterogeneous and generally low compared to responses to EBV and CMV and suggested a lack of immunodominance.
Detection of KSHV DNA as well as its quantity in peripheral blood mononuclear cells (PBMC) has been associated with KS disease progression [21][22][23][24] and increased HIV viremia in HIVinfected individuals 25 . In individuals without KS-or any other KSHV-associated disease, we have previously shown that KSHV DNA in PBMC was associated with increased KSHV antibody levels to lytic KSHV protein K8.1 and to malaria parasitaemia 26 . An association between increased lytic KSHV antibody levels and KS disease risk has also been reported 27 . Similarly in KS patients, correlations between KSHV in PBMC with increasing KSHV lytic antibody titres have been shown 25 . In combination, these data suggest that KSHV in PBMC corresponds to viral reactivation and disease progression. Therefore we hypothesise that loss of KSHV T cell control leads to increased KSHV DNA in PBMC.
In the current study, we use our systematic approach to investigate T cell responses to nearly the entire KSHV proteome in 116 KSHV+HIV-negative individuals with a wide age range from a rural cohort in Uganda, where both KSHV and KS are endemic. In addition, we compare T cell responses among KSHV seropositive individuals with and without KSHV DNA in PBMC. The pattern of IFN-γ responses to KSHV antigens compared to controls (EBV and CEF). Most of these individuals respond to EBV (76%) and CEF (69%) antigens (Table 1). Eighty-nine individuals out of 116 have a positive IFN−γ response to at least one KSHV ORF (77%); 44% respond to one to two KSHV ORFs and 33% respond to three or more KSHV ORFs. One individual responds to 37 KSHV ORFs (Table 1).

Results
Next, we compare the magnitude of the IFN-γ responses. The magnitude of IFN-γ responses to CEF (Figs. 1 and 2) and EBV ( Fig. 2) range from 40 spot forming cell (SFC)/million PBMC to 3713 and 3891-SFC/million PBMC, respectively (Supplementary Fig. 1). There is no consistent pattern of response to KSHV ORFs (Fig. 3), and when an IFN-γ response is observed to individual KSHV ORFs, the magnitude of the response is generally low (Figs. 1 and 2 and Supplementary Data 1 and 2). The raw data used to draw Fig. 1 are shown in Supplementary Data 1, while the Table 1 Proportions of individuals responding with IFN-γ production and their overall characteristics.  Furthermore, there is no pattern observed either by age group or sex. However, we observe that the proportion of individuals responding to at least one KSHV ORF increased with age ( Fig. 4a) and is similar in males and females (Fig. 4b). Furthermore, the number of reactive KSHV ORFs per individual increases with the increasing age of the study participant ( Supplementary Fig. 2). Among KSHV peptide pools, K8.1 is the most commonly recognised ORF (38%) followed by ORF73 (21%) (Fig. 3). Surprisingly, 23% of individuals respond to a Simian immunodeficiency virus (SIV) peptide (Fig. 3). Number of reactive antigens per individual in relation to KSHV DNA in PBMC. As mentioned above, 23%, 44% and 33% of individuals respond to no KSHV antigen, 1-2 antigens and 3-37 antigens, respectively. To explore whether there is a correlation between detectable levels of KSHV in PBMC, which could be indicative of a loss of T-cell control, we then compare the three groups (no response, response to 1-2 antigens or 3-37 antigens) focusing on those with KSHV DNA in PBMC. Individuals who respond to 1-2 KSHV antigens have the lowest levels of KSHV DNA in PBMC, while those without any IFN-γ response to KSHV and those with IFN-γ responses to three or more KSHV antigens have similar levels of KSHV DNA in PBMC (Fig. 6a). When we broke the groups into five smaller groups, we observe a decline in KSHV viral load from the non-responders to those responding to two peptide pools, then an increase in KSHV viral load thereafter (Fig. 6b). The low levels of KSHV viral load among those responding to 1-2 peptide pools and the higher levels of KSHV viral load among those responding to 3 or more peptide pools could be explained by the fact that very few individuals reacting to ORF73 were in the 1-2 peptide pool group.

Discussion
Previous studies showed that individuals without KS disease have higher T cell responses compared to those with KS 8,28-30   from the current study are similar to those seen in healthy US participants 20 . The low level of T cell responses to KSHV differs from the robust T cell responses to other human herpesviruses and requires further investigations. We think, however, that the level of responses is not due to a technical artefact such as the concentration of peptides as we stringently evaluated this possibility in our assay development and validation 20 . Although responses to KSHV are low on average, some individuals had very high responses to certain peptide pools. For example, we observed 3243 SFU per million cells as the maximum response for the K8.1 peptide pool that is very comparable to 3891-SFU per million cells that is the maximum response to the EBV peptide pool. In addition, responses to individual peptides shown in Table 2 are comparable to responses to peptide pools. Furthermore, the number of peptides per KSHV ORF is comparable to the number of peptides in the EBV peptide pool. KSHV has been shown to encode for numerous proteins including host protein homologues that aid host immune evasion 35,36 . For instance, MHC class I, MHC class II, CD1d, CD31, CD54, B7-2, MICA, MICB and IFN-γR1 degradation by KSHV RTA, K5 and K3 proteins had been shown to promote CD4+ T cells, CD8+ T cells and NK cells dysfunction 37 . However, the individuals we studied do not have any form of KSHV-associated disease and are of a wide age range, including older individuals who had probably been infected with KSHV for decades. Therefore, the responses in these individuals, while of low intensity and heterogenous, were sufficient to control KSHV in terms of disease risk. Myoung et al. showed suppression of spontaneous KSHV reactivation in tonsillar B cells by activated tonsillar CD4 T cells in an MHC-independent manner 38 . However, these results have not been replicated and the mechanism of this unspecific KSHV control was not investigated. Future studies investigating these findings further are warranted.
In addition to the T cell responses being low in intensity and frequency, they were very diverse in targeting KSHV antigens across sexes and age groups. Nevertheless, in combination, we have shown that a large number of individuals responded to at least one KSHV antigen (77%). Response to at least one KSHV antigen increases with age and is similar in males and females. This finding is consistent with the early age of infection with KSHV in endemic areas. This implies that the longer someone is infected with the virus, the higher the probability of developing detectable T cells specific for at least one KSHV antigen.
Individuals who responded to one to two KSHV antigens have a lower KSHV viral load compared to those without any detectable KSHV-specific T cell response. Surprisingly, also individuals responding to three or more antigens have a higher KSHV viral load similar to those without any KSHV T cell response. This may imply that T cell response to an individual viral protein such as K8.1 could suppress viral load.
Most studies have failed to show any association between KSHV viral load in PBMC and T cell responses 5,16,30 . This might have been attributed to the small sample sizes of these previous studies. Nevertheless, one study showed increased T cell responses to ORF73 among individuals with detectable KSHV in PBMC 39 , consistent with our observations in this study. Contrary to T cell responses to ORF73, T cell responses to K8.1 are higher in those without KSHV DNA in PBMC, supporting the hypothesis that the presence of KSHV-specific T cell responses inhibits viral replication, hence reducing the risk of detecting the virus in the blood. However, the frequency of this T cell response in KSHV seropositive individuals without detectable viruses is low, implying that there are other mechanisms through which these individuals control the virus.
We also observe reactivity to SIV, a non-human primate virus. This is very surprising because the SIV-CM9 peptide used in the assay is restricted by a common class I allele (Mamu A*01) in macaques. The equivalent epitope in HIV is also restricted by a single HLA allele, yet all the study participants are HIV uninfected. Due to the limitations of sample availability, we could not investigate this observation further, but this is most likely an example of heterologous T cell reactivity.
The strength of this study is the large number of participants compared to most past studies on KSHV T cell responses. In addition, by focusing on HIV uninfected individuals from a KSHV endemic area, as well as including a very wide age range of individuals of both sexes, it allows us to provide a comprehensive evaluation of the T-cell response across the age spectrum in healthy individuals. An additional strength of this study is the use of peptide pools that spanned the KSHV proteome.
In conclusion, IFN-γ T cell responses to KSHV are diverse and when they do occur, they are not as intense as those against EBV, CMV and influenza. However, the responses to particular antigens (for example K8.1) may be more protective from viral reactivation than others (for example ORF73/LANA). These results provide insights into the potential success of immune therapeutic vaccines in preventing or treating KS disease among KSHV-infected individuals at high risk of developing KS.

Methods
Ethical approvals. This study is approved by the Uganda Virus Research Institute-Research and Ethics Committee (reference number: GC/127/16/09/566), the Uganda National Council for Science and Technology (reference number: HS2123) and the London School of Hygiene and Tropical Medicine Ethics Committee (reference number: 11881). Written informed consent was obtained from all adults aged 18 years and above. Children below 18 years were consented to the study by their parents or guardian; we also sought, in addition to parental consent, written assent from children aged between 8 and 17 years. Transport refund is provided to all invited study participants following Institution Review Board regulations and approval.
KSHV ELISPOT assay. Overlapping 15mer peptides are synthesised to span nearly the entire KSHV proteome, and peptides are divided into pools representing each ORF 20   K8.1 peptide deconvolution. The K8.1 peptide pool (of 44 peptides) is deconvoluted in four study participants who responded to K8.1 peptide pool and had enough PBMC for the assay. Using a grid-type method 20 , 14 smaller overlapping peptide pools of six to seven peptides are generated. IFN-γ responses to the smaller peptide pools are determined using the ELISPOT assay described above.
Responding individual peptides identified from the peptide matrix are confirmed by stimulating with these individual responding peptides to run a third IFN-γ ELISPOT assay. KSHV PCR. KSHV viral load is quantified from PBMC KSHV seropositive study participants. After PBMC isolation using ficoll density gradient method, two million  cells are removed, spun at 13,000 rcf for 10 min to form cell pellets. The supernatants are poured off and the cell pellets are stored at −80°C. The PBMC pellets are retrieved from the −80°C freezer, thawed and genomic DNA extracted using a QIAamp blood kit (Qiagen, Valencia, CA), following the manufacturer's instructions. KSHV viral load is quantified using real-time PCR. KSHV DNA is detected using primers (K6 forward  primer K6-10F 5′-CGCCTAATAGCTGCTGCTACGG-3′, K6 reverse primer K6-10R  5′-TGCATCAGCTGCCTAACCCAG-3′) and a probe (K6 probe p-K6-10 5′-R-CACCCACCGCCCGTCCAAATTC-Q-3′) previously reported to be specific to the K6 gene region 40 . In addition, the number of cellular equivalents are determined using a quantitative assay specific to human endogenous retrovirus 3 (ERV-3), which is present in two copies per genomic cell, using these primers (ERV-3 Forward primer PHP10-F 5′-CATGGGAAGCAAGGGAACTAATG′ ERV-3 Reverse primer PHP10-R 5′-CCCAGCGAGCAATACAGAATTT-3′) and a probe (ERV-3 Probe PHP-P505 5′-R-TCTTCCCTCGAACCTGCACCATCAAGTCA-Q-3′). To quantify both ERV-3 and KSHV DNA, seven two-fold serial dilutions of K6 and ERV-3 are made from clone stocks (starting with 1 × 10 6 dilutions to 1 × 10 0 ) to form a standard curve on every plate. ERV-3 was cloned into Bluescript II KS vector (Stratagene, La Jolla, CA, USA) KSHV K6 cloned using PCR Topo II vector, Topo TA Cloning kit, Invitrogen, K 4600-40. Primers (500 μl each), probes (500 μl), nuclease-free water (5.5 ml) and universal master mix (Applied Biosystems, Foster City CA) (12.5 ml) are mixed together to form the working solution master mix. Thereafter 10 μl of the standards, sample and controls each, are added to a 96-well PCR plate (ThermoFisher catalogue number 7306737) per well, in triplicate, and 40 μl of the working solution master mix added to each well. This was followed by plate sealing and DNA amplification using an ABI ViiA7 machine. All samples are tested in triplicate for both assays and the estimated copy number for each individual reaction is averaged. The KSHV viral load is determined in PBMC DNA by calculating the viral DNA copies per million cells. Any sample that is quantifiable by the ERV-3 assay but not positive in all three reactions in the KSHV K6 assay is designated as positive but not quantifiable. The lower limit of detection for the ERV-3 assay has been previously determined as 10 copies per million cells while the KSHV K6 assay has a lower limit of sensitivity of 1 copy per million cells. This procedure has been reported elsewhere [41][42][43] . In order to prevent contamination, qPCR reagents, sample processing, DNA extractions, and qPCR setup are conducted in dedicated laboratory areas. Internal no template controls (NTC) were included on each plate, in triplicate, to assess crosscontamination. Assay quality controls including NTC, assay controls and sample triplicate % coefficient of variation are used to validate data. Importantly, the individual dilution parameters of the assay standard curve are assessed after each run to ensure performance within expected ranges. Any assay that has failures of assay controls or standard curves is repeated.

Statistics and reproducibility
Study population and study design. The General Population Cohort (GPC) is a community-based cohort of about 22,000 people living in 25 adjacent villages 44,45 , in rural southwestern Uganda. Previously, we have documented a high seroprevalence (>90%) of KSHV in the GPC 46,47 . For this study, we recruited individuals from the GPC who were KSHV seropositive, but HIV seronegative in 2017, to investigate determinants of KSHV viral detection. Only one individual was recruited per household 26 . Blood samples were processed to isolate PBMC (stored in liquid nitrogen for later immunological assays) and plasma (stored at −80°C for later use). Cell pellets containing two million PBMC were stored at −80°C for later DNA extraction. Details of the main study design have been reported previously 26 . KSHVspecific T cell IFN-γ responses are determined in 116 participants aged 6-87 years.
No statistical method is used to predetermine sample size. Purposive sampling is used to ensure that both male and female participants with and without detectable viral DNA in PBMC from each age group are selected. Simple randomisation is used to select individuals in each age group. Investigators are blinded to allocation during experiments and recruitment. Details of the characteristics of the participants (including the sex and age of participants) selected are shown in Table 1. Graphical representation is done using GraphPad Prism 8.0.1 and STATA version 13 (StataCorp, College Station, Texas USA). SFC per million PBMC are calculated from the number of spots counted by the ELISpot reader and the number of PBMC plated using Microsoft Office Excel (version 15.23). A cut-off of 40 SFC is applied. Therefore, individuals with SFC below 40 are considered nonresponders. χ 2 test, Kruskal-Wallis test, Mann-Whitney (Wilcoxon rank-sum) test are used for statistical analysis in STATA version 13 (StataCorp, College Station, Texas USA). No data were excluded from the analysis.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The data shown in all figures and tables generated in this study are provided in the Supplementary information/Source Data file. Source Data are provided with this paper.