HIV and prior exposure to antiretroviral therapy alter tumor composition and tumor:T-cell associations in diffuse large B-cell lymphoma

Abstract

Diffuse large B-cell lymphoma (DLBCL) is the most common subtype of lymphoma worldwide, accounting for up to 40% of new non-Hodgkin Lymphoma (NHL) globally. People living with HIV are up to 17 times more likely to develop NHL, and as such, DLBCL is the leading cause of cancer death in this high-risk population. While histologically indistinguishable, HIV-associated (HIV+) and HIV-negative (HIV−) DLBCL are molecularly distinct, and biologic differences may have implications for development of future therapeutic interventions. Further, the impact of immunologic differences in people with HIV, including preceding ART, remains largely unknown. Here, we investigate the impact of HIV infection and ART exposure on the clinical features of DLBCL and T-cell immune response by performing imaging mass cytometry on our unique patient cohort in Malawi. In this cohort, HIV infection is positively prognostic, and HIV+/ART-naïve patients have the best outcomes. No established biomarkers other than Ki67 are associated with HIV or ART status, and the only tumor-intrinsic biomarkers that remain prognostic are MYC and MYC/BCL2 protein co-expression. Finally, TCR clonality is associated with distinct tumor-T cell interactions by HIV/ART status, indicating differential anti-tumor immune responses. We demonstrate previously undescribed HIV and ART-related differences in the DLBCL tumor microenvironment.

Graphical Abstract

Diffuse large B cell lymphoma (DLBCL) is the most common subtype of lymphoma in both HIV+ and HIV− patients. Data from a cohort of patients in Malawi, where HIV is endemic, revealed that both HIV and exposure to antiretroviral therapy (ART) prior to DLBCL diagnosis significantly alter the tumor microenvironment and T-cell:tumor associations. While HIV− tumors in our cohort are characterized by Th2-driven T cell expansions, HIV+/ART-experienced tumors also reflect a Th1 response. Further, HIV+/ART-naïve patients had the highest overall survival, and while these tumors exhibit increased CD8⁺ T-cell presence, they lack tumor-driven clonal expansions. Finally, cell of origin and morphological classifications failed to associate with or predict outcome, while HIV and ART exposure, MYC expression, and MYC/BCL2 co-expression by IHC remained negatively prognostic.

INTRODUCTION

Human immunodeficiency virus (HIV) affected nearly 40 million people globally in 2022(1), and prevalence is predicted to increase, with the greatest burden occurring in Africa and the US.(2) While the implementation of antiretroviral therapy (ART) has allowed a greater life expectancy for people with HIV (PWH), cancer has become a leading cause of death in PWH.(3–5) HIV infection leads to impaired systemic immunity, genetic alterations, susceptibility to oncogenic viral infections, and chronic B-cell activation, all of which contribute to increased risk of malignancy.(6) PWH have an 11–17 fold increased risk of developing non-Hodgkin lymphoma compared to HIV-uninfected (HIV−) individuals of which diffuse large B-cell lymphoma (DLBCL) is the most common lymphoma subtype.(6–8)

Over the past two decades, extensive molecular characterization of HIV− DLBCL has identified biologically and therapeutically relevant genomic subtypes of HIV− DLBCL with clinically meaningful biomarkers.(9–11) Yet, while histologically indistinguishable and treated similarly, DLBCL arising in PWH (HIV+ DLBCL) exhibits distinct genetic and transcriptomic profiles compared to HIV− DLBCL, suggesting an impact of HIV on tumorigenesis and immune response.(12,13) Additionally, increased HIV viremia load at 6 months post DLBCL-diagnosis has been linked to higher mortality rates at 5 years, implicating a role for ART in shaping tumor response.(14) Though the high incidence of DLBCL in HIV+ populations is well established, the specific impact of HIV and ART on the tumor itself, as well as biomarker applicability, remains poorly understood.

Despite significant progress in understanding tumor-host interactions in HIV− DLBCL, the impacts of HIV and ART, on the DLBCL tumor microenvironment (TME) remain understudied.(15) Previous work has identified differences in HIV+ compare to HIV− DLBCLs, with the former being more proliferative, demonstrating increased angiogenic markers, decreased CD4⁺ T cells, and an enrichment of CD8⁺ T cells.(16–18) Yet, the functionality and clinical impact of these cells and tumor phenotypes remain unknown. Both HIV and cancer have been shown to independently induce a chronic immune response, resulting in a dysfunctional T cell state known as T cell exhaustion.(19) Modern immunotherapies utilize this response by targeting co-inhibitory “checkpoint” receptors, such as PD-1, TIM3, and LAG3, thereby preventing or reversing T-cell exhaustion.(20) Further, ART has been shown to partially reduce T-cell exhaustion, though this has yet to be studied in the concomitant context of HIV and cancer.(21) Despite the high incidence of DLBCL in PWH, and the intrinsic link between Tex, HIV, and anti-tumor immune response, the impact of HIV and ART on the TME and tumor immune response are poorly understood.

Finally, recent studies have established clinical relevance of spatially-defined tumor architecture, providing new evidence that the potentially targetable heterogeneity of DLBCL is driven by the absence or presence of immune cells within the TME.(22,23) Given the known impacts of HIV on systemic immunity and, thus, the unique pressures under which tumors develop in PWH, it is critically important to characterize tumor:host interactions to improve outcomes for this vulnerable population worldwide.

In this study, we aimed to characterize the TME of HIV−, HIV+/ART-experienced, and HIV+/ART-naïve DLBCL. This work is enabled by longstanding clinical research collaboration in Malawi, with extremely limited public sector health care resources and high HIV prevalence, providing insights into lymphomagenesis and tumor-host interactions under varying degrees of immune pressure.(13,14,24–29) This unique patient cohort allows us to study DLBCL arising in HIV−, HIV+/ART-naïve and HIV+/ART-experienced patients as a comparative model system to investigate the tumor microenvironment and T-cell responses in DLBCL to address critical gaps in tumor immunology.

METHODS

Kamuzu Central Hospital Lymphoma Study

The KCH Lymphoma Study has prospectively enrolled all patients with newly diagnosed pathologically confirmed lymphoproliferative disorders at a national teaching hospital in Lilongwe, Malawi since 2013 (NCT02835911).(14,24) All patients provided written informed consent prior to enrollment. All patients were treated with CHOP (cyclophosphamide, doxorubicin, vincristine, and prednisone), the current standard of care in Africa, during the study period, and a subset of patients (n=34) also received rituximab as part of a previously reported clinical trial (NCT02660710).(24,29) Treatment toxicity for patients included in this cohort has been previously reported.(28,29) Consolidative radiotherapy was not available in Malawi during the study period for persistent or bulky disease. Patients with relapsed or refractory disease were treated with salvage chemotherapy regimens including ifosfamide, etoposide, and cisplatin or gemcitabine plus oxaliplatin.(30) Patients who were able to achieve remission with salvage chemotherapy were referred to the Malawi Ministry of Health for autologous stem cell transplant or alternative therapies outside the country but this was infrequently completed. HIV+ patients continued or initiated ART at the time of lymphoma diagnosis. Patients who had been prescribed ART for at least 6 months prior to lymphoma diagnosis and study enrollment were categorized as “ART-experienced” (ART-exp.), while those who had been prescribed ART for less than 6 months prior to enrollment, or who were newly diagnosed with HIV, with categorized as “ART-naïve”. All patients met the following inclusion criteria: enrolled between 2013 and 2021, over 18 years of age at time of enrollment, histologically confirmed de novo DLBCL, and known ART duration. Patients were followed every three months for the first two years and then either every six months in person or every two months by phone until five years follow up, as previously described, with censoring occurring on March 20^th, 2023.(28,29) No patients were lost to follow-up. Standardized clinical and laboratory assessment at enrollment included Ann Arbor Stage, lactate dehydrogenase (LDH) levels, Eastern Cooperative Oncology Group (ECOG) performance status, as well as HIV RNA and CD4⁺ T-cell counts for HIV+ patients. LDH levels were categorized as greater than twice the upper limit of normal (LDH>2) and ECOG scores were categorized as greater than or equal to 2 (ECOG≥2). Pre-treatment excisional tissue biopsies were collected at diagnosis and stored as formalin-fixed paraffin-embedded (FFPE) tissue blocks. FFPE and frozen blood pellets were shipped to UNC. Overall survival (OS) was calculated as time between date of study enrollment and date of death or censoring, and progression-free survival (PFS) was calculated as time between date of study enrollment and date of progression, death, or censoring, whichever occurred first. Both OS and PFS were calculated for cell-of-origin (COO), morphology, MYC, MYC/BCL2 protein co-expression, Ki67, LDH>2, and ECOG≥2, and adjusted for HIV/ART status via multivariate Cox regression. The Kamuzu Central Hospital (KCH) Lymphoma Study was approved by the University of North Carolina Institutional Review Board and the Malawi National Health Sciences Research Committee.

Immnunohistochemistry

Primary diagnoses were determined by conventional histology and immunohistochemistry (IHC) using antibodies as previously described: CD3 (clone PS1), CD20 (clone L26), CD30 (clone 15B3), CD45 (code NCL-L-LCA-RP), CD138 (clone MI15), BCL2 (clone bcl2/100/D5), Ki67 (Clone MM1), TdT (Clone TdT-338), and HHV8 (NCL-HHV8-LNA), all from Leica Biosystems (Buffalo Grove, IL, US).(13,14,25) BCL2 (clone 124) and MYC (clone y69) expression were assessed by IHC using antibodies from Ventana Medical Systems (Tuscon, AZ, USA) on the Ventana Discovery Ultra.(13) MYC positivity was defined as staining of >40% of neoplastic cells. While a cutoff of >50% positivity Is commonly used for BCL2, we defined it as ≥70%, based on the work of Green et al., due to the frequent strong positive staining in this cohort.(13,31,32) None of our cases fell between 50% and 70% positivity for BCL2. Ki67 was quantified by light microscopy. In situ hybridization (ISH) for Epstein-Barr Virus small RNAs (EBER) was performed on a Leica Bond platform (Leica Biosystems) according to the manufacturer’s instructions. COO was assigned per the algorithm by Hans et al. using CD10 (clone NCL-CD10–270) and BCL6 (PA0204) from Leica Biosystems, and MUM1 (M7259m) from Dako (Carpinteria CA, USA).(33)

Imaging mass cytometry

A total of n=66 EBV-negative DLBCL FFPE tumors were submitted for tissue microarray (TMA) generation and IMC. For each sample, 2–3 regions of interest (ROIs) representative of the tumor were selected by a pathologist (n=183). Ablation and imaging mass cytometry were performed at the Spatial Molecular Profiling Shared Resource at Cedars Sinai as previously described.(23) Using PhenoGraph, expression and spatial features of 45 markers were hierarchically clustered to identify cell lineages as previously described (Supplemental File 1).(23) A representative ROI and cellular identification by protein expression data is provided in Supplemental Figure 1. ROIs with poor tissue quality or those that were damaged or shifted during ablation were excluded from analysis (n=47), resulting in a final n of 56 tumors available for study. Density of cell type X was calculated as 1 - average nearest neighbor distance of each tumor cell to the nearest 5 cells of type X, truncated at 50μm. Cellular composition of tumors was calculated as either a percentage of total cells in the ROI (cellular percentage) or as a proportion of immune cells in the ROI (immune percentage). Spatial clusters for each cell lineage were generated using k-means clustering (k=15) to identify the patterns of cellular interactions. Heatmaps were generated by defining these clusters using the aforementioned spatial density metric. To account for the uneven distribution of ROIs among samples, ROI cellular expression and spatial data were averaged by patient.

TCR sequencing

DNA was extracted from n=59 pre-treatment FFPE DLBCLs (QIAmp DNA FFPE Advanced and Blood Mini kits). Samples below 10 ng/μL were excluded. We performed TCR sequencing on DNA from n=53 pre-treatment DLBCL FFPE tumors, using the immunoSEQ Human TCRB assay (Adaptive Biotechnologies) and running the pooled libraries on a NextSeq 500. Raw sequencing data were processed using Adaptive Biotechnologies’ pipeline and clonality metrics were calculated after random downsampling to 108 productive templates, averaged over 100 iterations (exclusion criteria: failed quality control, <100 productive templates). n=22 of these samples had matched IMC data and were utilized in this study. Productive Simpson clonality is the square root of Simpson’s diversity index for the productive rearrangements of a sample. Maximum (max) productive frequency is the frequency of the most expanded clone in a sample. Unique productive rearrangements is the number of unique productive TCR rearrangements in a sample.

Statistical tests

To determine correlations between HIV/ART status and COO, morphology, MYC, MYC/BCL2 co-expression, Ki67, LDH>2ULN and ECOG ≥2 chi-square analyses were performed. A pairwise Wilcoxon rank-sum test was used to compare clinical/demographic variables, including CD4, HIV RNA, and duration on ART, by HIV/ART status. To test the association of clinical/demographic values with outcomes, median cut-off was used to categorize the data as necessary, and hazard ratios were estimated by Cox regression. To compare cellular and immune percentages by HIV/ART status, pairwise Wilcoxon rank-sum tests were used. To test associations between bulk clonality and immune percentages or B-cell neighborhood analyses, Spearman’s correlation coefficient was used. Analyses were performed in R 4.2.0 using the dplyr, tidyverse, survival, Rphenograph, and mclust packages. Figures were generated using the ggplot2, ggpubr, survminer, uwot, and corrplot packages.

Data availability

IMC data available upon request from the corresponding author, YF. TCR sequencing data available on the ImmuneAccess database (Adaptive Biotechnologies, Seattle, WA, USA).

RESULTS

Patient characteristics

n=152 adult DLBCL patients with known HIV and ART status were enrolled in the KCH study between 2013 and 2019, of which n=57 were HIV−. N=61 were HIV+/ART-Exp., and n=34 were HIV+/ART-naïve (Table 1). Only n=15 (10%) tumors were EBV-positive (EBV+) by EBER in situ hybridization. Only four patients had central nervous system involvement, (n=2 HIV−, n=2 HIV+/ART-naive), two of which were EBV+ (n=1 HIV−, n=1 HIV+/ART-naïve). HIV+ patients were significantly younger than HIV− patients (p=0.046), and though not significant, HIV+/ART-naive tumors trended towards increased GC cell-of-origin subtyping. The majority of patients were strongly BCL2-positive, regardless of HIV/ART status, warranting use of a 70%-positivity cutoff. N=56 of these tumors were utilized for IMC, and of these, n=22 had available TCRseq data. Additional patient characteristics of each molecular study cohort are presented in Supplemental Table 1.

Table 1.

Clinical and demographic features of KCH DLBCL patient cohort, separated by HIV/ART status.

	HIV− (N=57)	HIV+/ART-exp. (N=61)	HIV+/ART-naïve (N=34)	Total (N=152)
Age (years)
Median [Min, Max]	48.0 [19.0, 80.0]	46.0 [18.0, 65.0]	43.0 [22.0, 62.0]	45.0 [18.0, 80.0]
Sex
Female	23 (40.4%)	21 (34.4%)	14 (41.2%)	58 (38.2%)
Male	34 (59.6%)	40 (65.6%)	20 (58.8%)	94 (61.8%)
Prior ART (months)
Median [Min, Max]	NA	59.3 [6.18, 178]	0.838 [0, 4.57]	30.3 [0, 178]
CD4 Count
Median [Min, Max]	NA	147 [13.0, 844]	115 [9.00, 392]	139 [9.00, 844]
HIV RNA
Median [Min, Max]	NA	0 [0, 913000]	10500 [0, 1890000]	40.0 [0, 1890000]
Cell of Origin (COO)
GC	23 (56.1%)	23 (56.1%)	22 (73.3%)	68 (60.7%)
non-GC	18 (43.9%)	18 (43.9%)	8 (26.7%)	44 (39.3%)
EBER
Positive	7 (12.3%)	6 (10.0%)	2 (5.9%)	15 (9.9%)
Negative	50 (87.7%)	54 (90.0%)	32 (94.1%)	136 (90.1%)
Ki67
Median [Min, Max]	0.800 [0.300, 0.950]	0.800 [0.100, 0.950]	0.900 [0.500, 0.950]	0.800 [0.100, 0.950]
MYC Expression > 40%
Positive	10 (24.4%)	18 (39.1%)	11 (42.3%)	39 (34.5%)
Negative	31 (75.6%)	28 (60.9%)	15 (57.7%)	74 (65.5%)
BCL2 Expression ≥ 70%
Positive	50 (98.0%)	51 (89.5%)	25 (86.2%)	126 (92.0%)
Negative	1 (2.0%)	6 (10.5%)	4 (13.8%)	11 (8.0%)
MYC/BCL2 Co-expression
Positive	9 (22.0%)	14 (30.4%)	9 (36.0%)	32 (28.6%)
Negative	32 (78.0%)	32 (69.6%)	16 (64.0%)	80 (71.4%)
Stage
1–2	33 (57.9%)	24 (39.3%)	16 (47.1%)	73 (48.0%)
3–4	24 (42.1%)	37 (60.7%)	18 (52.9%)	79 (52.0%)
ECOG Score
0–1	27 (47.4%)	37 (60.7%)	23 (67.6%)	87 (57.2%)
2–4	30 (52.6%)	24 (39.3%)	11 (32.4%)	65 (42.8%)
LDH
Median [Min, Max]	443 [168, 1890]	519 [27.0, 4380]	537 [144, 4800]	479 [27.0, 4800]
Treatment
CHOP	48 (84.2%)	44 (72.1%)	26 (76.5%)	118 (77.6%)
R-CHOP	9 (15.8%)	17 (27.9%)	8 (23.5%)	34 (22.4%)
Survival (months)
Median [Min, Max]	14.9 [0.0657, 60.0]	12.6 [0, 60.0]	35.5 [0.329, 60.0]	14.7 [0, 60.0]

ECOG=Eastern Cooperative Oncology Group Score; LDH=Lactate Dehydrogenase

Exclusion of EBV+ Tumors

Notably, there was a low frequency of EBV+ tumors in our cohort, only comprising 10% of cases (n=15). Of these, only 4 had available IMC data, and 1 had additional TCR data. While another study has shown that EBV+ HIV+ tumors have altered mutational profiles and clinical features, in our cohort EBV positivity was not associated with HIV/ART status, expression of Ki67, MYC, BCL2, MYC/BCL2 co-expression, or overall survival.(34) As such, given the impacts of EBV on lymphomagenesis and the low frequency of EBV positivity in our cohort, EBV+ tumors were excluded for downstream analysis, resulting in a final n of 137 patients (n=50 HIV−, n=55 HIV+/ART-exp., and n=32 HIV+/ART-naïve).

ART stratification results in biologically distinct HIV+ cohorts

As expected, the HIV+/ART-naïve group had significantly higher HIV RNA (p<0.001), lower CD4⁺ T-cell counts (p=0.042) and shorter duration on ART (p<0.001) at DLBCL diagnosis and study enrollment compared with HIV+/ART-exp. (Figure 1A–C). Notably, median time on ART in HIV+/ART-naïve patients was 0.9 months and HIV+/ART-exp. patients was 55 months. Additionally, each group experienced differing rates of CD4 reconstitution as measured by peripheral blood CD4 counts. After 5 years on ART, HIV+/ART-naïve patients that survived DLBCL regained peripheral CD4 T cells at an increased rate compared to HIV+/ART-experienced patients (p=0.005, ANOVA) (Figure S2).

Figure 1. Biological stratification by ART exposure and differential outcomes by HIV/ART status.

Comparison of A) HIV viral load (log), B) CD4⁺ T cell count by blood draw, and C) ART duration between HIV+/ART-exp. and HIV+/ART-naïve patients. HIV+/ART-exp. patients had higher viral loads (p=<0.001), CD4⁺ T cell counts (p=0.042), and ART duration prior to diagnosis (p=<0.001) compared to HIV+/ART-naïve patients. D) Kaplan-Meier curve depicting OS by HIV status. HIV+ patients have increased OS compared to HIV− (p=0.041). E) Kaplan-Meier curve depicting OS by HIV/ART status. HIV+/ART-naive patients have improved OS compared to HIV− (p=0.026)

HIV+/−naïve patients have improved overall survival in our cohort.

HIV status was prognostic for OS (HIV+ HR=0.65, p=0.041) and marginally prognostic for PFS (HIV+ HR=0.70, p=0.088) (Figure 1D,S3A). Further stratification by ART status revealed prognostically significant differences in outcomes, in which HIV+/ART-naïve patients had increased OS (HR=0.51, p=0.026) and PFS (HR=0.55, p=0.039) compared to HIV− patients. There was no significant difference in outcome between HIV+/ART-exp. and HIV− cases (Figure 1E,S3B).

COO and morphology are not prognostic in an HIV+ DLBCL inclusive cohort.

Of the tumor-intrinsic biomarkers and stratification systems we tested, only Ki67 associated with HIV or HIV/ART status, with HIV+/ART-naïve tumors having increased Ki67 (p=<0.001).(7,13) When accounting for HIV/ART status, COO, morphology, and Ki67 did not significantly impact outcomes in this cohort (Supplemental Table 2). Co-expression of MYC/BCL2 was significant for OS (HR=2.51, p<0.001) and PFS (HR=2.56, p<0.001). Notably, when accounting for HIV/ART status expression of MYC alone was prognostic for both OS (HR=2.07, p=0.006) and PFS (HR=2.19, p=0.002). Further, when separated, expression of MYC was marginally prognostic for OS and PFS in HIV− patients (HR=2.27,p=0.06 and HR=2.22, p=0.07), and prognostic for PFS in HIV+ patients (HR=2.14, p=0.02). LDH > 2 and ECOG ≥ 2 remained negatively prognostic for both OS and PFS after adjusting for HIV/ART status (Table 2).

Table 2.

Prognostic ability of traditional DLBCL biomarkers and classification systems when accounting for HIV/ART determined by cox-hazard regression.

	OS			PFS
	HR	CI	p-value	HR	CI	p-value
Morphology
Centroblastic	0.97	0.5, 1.89	>0.9	1.01	0.52, 1.97	>0.9
Immunoblastic	1.03	0.53, 2.00	>0.9	0.99	0.51, 1.92	>0.9
COO
GC	0.91	0.57, 1.46	0.7	0.87	0.55, 1.39	0.6
non-GC	1.10	0.69, 1.76	0.7	1.15	0.72, 1.83	0.6
MYC > 40%	2.07	1.23,3.48	0.006	2.19	1.32, 3.62	0.002
BCL2 ≥ 70%	1.19	0.47, 3.02	0.7	1.12	0.44, 2.85	0.8
MYC/BCL-2 Co-expression	2.51	1.47,4.28	<0.001	2.56	1.52, 4.32	<0.001
Ki67 ≥ 80%	1.12	0.73. 1.74	0.6	1.14	0.74, 1.75	0.5
LDH > 2ULN*	2.10	1.33, 3.32	0.001	2.08	1.33, 3.26	0.001
ECOG > 2†	2.78	1.79, 4.31	<0.001	2.81	0.39, 4.31	<0.001

^*LDH > 2 ULN = lactate dehydrogenase > twice the upper limit of normal

^†ECOG > 2 = Eastern Cooperative Oncology Group score > 2

HIV/ART status impacts immune cell composition and distribution of DLBCL.

As the differences in outcomes were not captured by traditional tumor-intrinsic biomarkers, we next investigated the TME by IMC of a representative subset of cases. Nine major cell lineages were identified: B cells (considered to overwhelmingly represent tumor in this context), CD4⁺ T cells, regulatory T cells (Tregs), CD8⁺ T cells, dendritic cells (cDCs), endothelial cells, macrophages, myeloid cells, and stroma. Further subtyping enabled the identification of immune cell subtypes, including M1 macrophages, Th1 CD4⁺ T cells, Th2 CD4⁺ T cells, active effector CD8⁺ T cells, and terminally exhausted CD8⁺ T cells. Uniform manifold approximation and projection (UMAP)s depicting the major cell populations, subpopulations, and marker distribution were generated, reflecting the loss of CD4⁺ T cells and myeloid cell lineages from HIV− to HIV+/ART-exp. and HIV+/ART-naïve cases (Figure 2A,S4). Immune cells made up a greater percentage of the TME in HIV− tumors (median=33%) compared to HIV+/ART-exp. and HIV+/ART-naïve ones (median=25% and 27%, respectively) (Figure 2B). Of the immune component, HIV+ tumors had decreased CD4⁺ T cell proportions, including Tregs, Th1, and Th2 cells (HIV− median=45%, HIV+ mean=20%) (Figure 2C). Further, HIV+/ART-naïve patients had increased proportions of CD8⁺ T cells (HIV+/ART-naive median=9%, HIV− median=5%) and terminally exhausted CD8⁺ T cells, defined by co-expression of PD-1, TIM-3, and LAG-3, compared to HIV− tumors (HIV+/ART-naive median=3%, HIV− median=1%) (Figure 2C).

Figure 2. HIV and ART alter immune cell composition and T cell subtypes in DLBCL TME.

A) Uniform manifold approximation and projection (UMAP)s depicting tumor, CD4⁺ T cell, CD8⁺ T cell, dendritic cell (cDC), macrophage (Mac), myeloid (Myel), stroma, and regulatory T cell (Treg) components of the DLBCL TME determined by PhenoGraph clustering by HIV/ART status. B) Stacked bar plot showing percentage of cell lineages measured as a proportion of total cells in TME per sample of n=20 HIV−, n=22 HIV+/ART-exp., and n=13 HIV+/ART-naïve tumors. C) Stacked bar plot showing percentage of immune cell lineages in TME measured as a proportion of immune cells in the TME per sample by HIV/ART status. D) Jitter plots representing CD4⁺ and CD8⁺ T cell percentage of total cells in TME by HIV/ART status. HIV− tumors have increased proportions of CD4⁺ T cells compared to HIV+/ART-exp. (p=0.0016) and HIV+/ART-naïve p<0.001) and HIV+ ART-naïve tumors have increased CD8⁺ T cells (p=0.032) compared to HIV− ones. E) Jitter plots representing CD4⁺, CD8⁺, Th1, Th2, Treg, Terminally Exhausted CD8⁺, and Active Effector CD8⁺ T-cell composition as a percentage of immune cells in the TME. HIV+/ART-naïve tumor immune microenvironments are enriched for terminally exhausted CD8⁺ T cells (p<0.001) and HIV− tumors are enriched for Th1 cells (p=0.002, p=0.005). P-values determined by pairwise Wilcoxon rank sum test (D,E).

HIV/ART status shifts T cell subtype populations in DLBCL.

HIV− tumors had increased CD4⁺ T-cell presence, both by cellular percentage (p<0.001 vs. HIV+/ART-exp., p<0.001 vs. HIV+/ART-naive) and immune percentage (p<0.001 vs. HIV+/ART-exp., p<0.001 vs. HIV+/ART-naïve) (Figure 2D). HIV+/ART-naïve tumors, however, had increased CD8⁺ T-cell presence compared to HIV− cases by cellular percentage (p=0.032) and compared to both HIV− and HIV+/ART-naïve cases by immune percentage (p<0.001 and p=0.038, respectively) (Figure 2E). HIV+/ART-naïve tumors were also enriched for terminally exhausted CD8⁺ T cells compared to HIV− and HIV+/ART-exp. tumors (p<0.001 and p=0.012, respectively) (Figure 2E). HIV− tumors had an increased proportion of Th1 T cells compared to HIV+/ART-exp. and HIV+/ART-naïve patients (p<0.001 and p<0.001 respectively), with no difference in Th2 or Treg cells between the three groups (Figure 2E).

Differential T-cell and B cell spatial association by HIV/ART status.

Fifteen tumor cell spatial clusters were identified by the density of nearest neighbors using k-means clustering separately by HIV/ART status. It is important to note that these neighborhoods are not mutually exclusive and contain overlap. For example, tumor-rich neighborhoods may also contain a subset of CD4 neighborhoods. In all three groups, tumor cells were most densely surrounded by other tumor cells. In HIV− tumors, 31% of tumor cells were found in CD4⁺ T-cell rich neighborhoods (clusters 7–9, 11–12, 14), 15% CD8⁺ T-cell rich neighborhoods (clusters 2, 7, 13–15), and to a lesser extent, Th1 (6%, clusters 7, 14), Treg (3%, cluster 14), and terminally exhausted CD8⁺ T-cell rich neighborhoods (6%, clusters 14–15) (Figure 3A). In HIV+/ART-exp. tumors, tumor cells formed some CD4⁺ T-cell rich neighborhoods (9%, clusters 1, 13, 14), but lost Th1 neighborhoods and had an increased percentage of CD8⁺ T-cell rich neighborhoods (24%, 8–11, 13, 15) (Figure 3B). While fewer than HIV− tumors, HIV+/ART-naïve tumors also maintained CD4⁺ T-cell rich neighborhoods (6%, clusters 1–2). Additionally, HIV+/ART-naïve tumors reflected a drastically increased number CD8⁺ T-cell rich neighborhoods (48%, clusters 1, 4, 8–14) and terminally exhausted CD8⁺ T-cell (13%, clusters 8, 10) neighborhoods, indicating increased tumor cell:CD8⁺ T-cell interactions (Figure 3C). Most notably, HIV− tumors exhibited an increased association between tumor and CD4⁺ rather than CD8⁺ T cells, whereas HIV+/ART-naïve tumors exhibited increased CD8⁺ T cell:tumor clustering.

Figure 3. Spatial interactions of B cells and CD8 T cells by HIV/ART status.

Heatmaps depicting the B cell spatial phenotypes determined by density of nearest neighbors to tumor cells in A) HIV−, B) HIV+/ART-exp., and C) HIV+/ART-naïve tumors. Heatmaps depicting CD8⁺ T cell spatial phenotypes determined by density of nearest neighbors to CD8⁺ T cells in D) HIV−, E) HIV+/ART-exp., and F) HIV+/ART-naïve tumors. Clusters were numbered 1 through 15, and generated independently per HIV/ART group. Cell participation in a cluster is not mutually exclusive, and therefore a tumor-rich cluster may also be a CD4⁺ T cell rich cluster, for example. Clusters were identified as cell-type rich neighborhoods when that neighbor’s density was > 0.5.

To further characterize CD8⁺ T-cell interactions in the TME, we considered CD8⁺ T-cell spatial clusters by the density of nearest neighbors. By HIV/ART status, 68% of HIV− CD8⁺ T cells were found in tumor cell rich neighborhoods (clusters 3–11), 64% in CD4⁺ T-cell rich neighborhoods (clusters 4–5, 8, 10–15), 50% in Th1 rich neighborhoods (clusters 8, 10–15), and 23% in terminally exhausted CD8⁺ T-cell rich neighborhoods (clusters 9–10) (Figure 3D). HIV+/ART-exp. cases had a similar proportion of CD8⁺ T-cells in tumor cell rich neighborhoods (70%, clusters 1–10, 13, 15), fewer clusters in CD4⁺ T-cell rich neighborhoods (17%, clusters 1, 8–10), and similar clusters in terminally exhausted CD8⁺ T-cell rich neighbors (25%, clusters 8–9, 11,14–15) (Figure 3E). Finally, HIV+/ART-naïve CD8+ T cells had the most tumor cell rich neighborhoods (86%, clusters 1–8, 12–15), the fewest CD4⁺ T-cell rich neighborhoods (12%, clusters 8–9), and the most terminally exhausted CD8⁺ T-cell rich neighborhoods (50%, clusters 6,8,11–15) (Figure 3F). As such, HIV+/ART-naïve tumors reflected an increased spatial association between tumor cells and CD8⁺ T cells, and a greater presence of terminally exhausted CD8⁺ T cells, with a loss of CD4⁺ spatial association with tumor cells. In contrast, HIV− tumors had increased CD4⁺, Th1, and Treg spatial associations with tumor cells, and HIV+/ART-exp. tumors reflected an intermediary phenotype.

Tumor TCR clonality is associated with terminally exhausted T-cell presence in the tumor.

For the n=22 tumors with IMC data with corresponding TCRseq clonality data, we tested for associations between clonality metrics, immune cell composition (cellular percentage), and spatial association with B cells (neighbor density). In the combined cohort, productive Simpson clonality positively correlated with CD8⁺ T cell percentage (p=0.039, R=0.47) and terminally exhausted CD8⁺ T cell percentage (p=0.004, R=0.57), which also positively correlated with max productive frequency (p=0.006, R=0.55) (Figure 4A). Productive Simpson clonality and max productive frequency positively correlated with terminally exhausted CD8⁺ T-cell neighbor density (p=0.005, R=0.57 and p=0.018, R=0.5), indicating that the clonally expanded TCRs are derived from CD8⁺ T cells that interact with the tumor and enter a terminally exhausted state (Figure 4A). When separated into the three HIV/ART groups, clonality was differentially associated with immune cell and tumor cell neighbor density. In HIV− tumors, productive Simpson clonality and max productive frequency positively correlated with both Th2 % (p=0.003, R=0.93 and p<0.001, R=0.96) and Th2 neighbor density (p<0.001, R=0.96 and 0.003, R=0.93) (Figure 4B,C). In HIV+/ART-exp. tumors, productive Simpson clonality also positively correlated with Th2% (p<0.001, R=0.87) and Th2 neighbor density (p<0.001, R=0.93), as well as Th1 neighbor density (p=0.021, R=0.79), CD4⁺ T cell neighbor density (p=0.002, R=0.9), and terminally exhausted CD8⁺ T-cell density (p=0.021, R=0.79) (Figure 4B,C). Finally, in HIV+/ART-naïve tumors, productive Simpson clonality and max productive frequency only correlated with CD8⁺ T cell % (p=0.003, R=0.93 and 0.007, R=0.96), but not CD8⁺ T-cell neighbor density, whereas total productive templates positively correlated with active effector CD8⁺ T-cell neighbor density (p=0.023, R=0.82) (Figure 4B,C). This implies that T-cell response in HIV− DLBCL is driven largely by Th2 cells. Meanwhile, response in HIV+/ART-exp. is driven by CD4⁺ T cells (Th1 and Th2), and HIV+/ART-naïve clonal expansions may not be a result of tumor-targeting T cells, despite the increased CD8⁺ T cell:tumor association. To investigate this, we further investigated TCR epitopes to identify different T-cell targets between the three groups.

Figure 4. Tumor bulk clonality metrics are associated with HIV/ART status and terminally exhausted T-cell presence in tumor.

A) Correlation matrix of clonality metrics with cellular composition and tumor:cell type spatial association determined by nearest neighbor density across entire cohort (n=22). Clonality associates positively with CD8+ T-cell percentage and terminally exhausted CD8⁺ T-cell density. B) Correlation matrix of clonality metrics with tumor: cell type spatial association separated by HIV/ART status for n=7 HIV−, n=8 HIV+/ART-exp., and n=7 HIV+/ART-naïve tumors. D) Correlation matrix of clonality metrics with cellular composition separated by HIV/ART status for n=7 HIV−, n=8 HIV+/ART-exp., and n=7 HIV+/ART-naïve tumors.

DISCUSSION

Herein, we begin to understand the impact of HIV infection and ART exposure on immune composition and T-cell response to DLBCL. To our knowledge, this is the first description of an integrated analysis of tumor immune landscape and TCR repertoires of an HIV-inclusive DLBCL cohort stratified by ART exposure.

There is limited literature on the biological differences between HIV+ and HIV− DLBCL, in part because in high-income counties, HIV+ DLBCL is relatively uncommon and mainly affects marginalized groups. While other studies have reported positive and negative prognostic implications of HIV status, in our cohort, we observed that HIV+ patients had improved five-year OS and PFS.(35,36) We further observed that HIV+/ART-naïve patients had improved outcomes compared to HIV− patients. A prior study from the Center for AIDS Research Network of Integrated Clinical Systems similarly identified lymphoma developing on ART versus not on ART as an independent risk factor for mortality among patients with HIV+ lymphoma in a large United States cohort.(37) Survival differences among HIV+ DLBCL patients depending on ART exposure may imply distinct tumorigenesis and/or immune response, or a therapeutic benefit of initiating ART alongside chemotherapy.

Additionally, in our cohort the majority of tumor-intrinsic prognostic classification systems and biomarkers failed to distinguish tumors by HIV/ART status, and were not independently prognostic. HIV/ART status alone was most valuable as a predictive biomarker in our cohort, suggesting that biological differences between these tumors remains uncaptured by traditional prognostic techniques, and prompting further investigation of the TME.

We find that HIV/ART status impacts the cellular composition and spatial organization, which may be due to differential anti-tumor responses and/or setting of lymphomagenesis. Our data suggests that HIV− T-cell clonal expansions are likely driven by Th2 cells, whereas HIV+/ART-exp. expansions may be driven by a more generic CD4⁺ T cell (Th1 and Th2) response to tumor. Finally, while HIV+/ART-naïve tumors have increased CD8⁺ T cell presence and tumor cell spatial association, TCR clonality did not associate with CD8⁺ T-cell (activated effector or terminally exhausted) presence or proximity to tumor. It is possible that the increased CD8⁺ T-cell presence, activation, and exhaustion are at least partly driven by a non-tumor targeting T-cell response. These CD8⁺ T-cell clonal expansions may instead result from responses to opportunistic infections or HIV itself.

This study may hold implications for immunotherapy application in HIV+ DLBCL. In our cohort, we observed differences in the T cell presence and spatial association with tumor to related to systemic immunity. A better understanding of the immune landscape in HIV+ DLBCL, and how it may differ depending on ART exposure, is necessary to developing equitable treatment paradigms for all DLBCL patients. The most promising immunotherapy to date in DLBCL is CAR-T therapy, which is often not feasible for many patients worldwide due to cost and implementation barriers, including manufacturing time and required hospital infrastructure.(38–40) The majority of immune checkpoint inhibition (ICI) clinical trials in DLBCL have focused on PD-1 blockade, which has shown suboptimal responses, likely due to the variability of PD-L1 expression in DLBCL.(41) We identify CD8⁺ T cells that express PD-1, LAG-3, and TIM-3 across all three HIV/ART groups. LAG-3 expression in DLBCL has been linked to poor overall survival, and current therapeutics targeting these three markers are in trial.(20,42) Though HIV+ DLBCL patients have historically been excluded from DLBCL clinical trials, small studies have suggested possible safety and efficacy of ICI in HIV+ non-Hodgkin lymphoma patients.(43,44)

While stratifying HIV+ patients using a six-month cut-off for ART duration does not capture the profound heterogeneity of HIV+ DLBCL, it does reveal biological differences related to DLBCL biomarkers and anti-tumor T-cell response. HIV+/ART-exp. DLBCL patients represent a highly heterogeneous group due to variations in HIV infectious course, as well as ART regimen and response.(45,46) Importantly, as the global population continues to grow and age, the incidence of non-Hodgkin lymphoma, including DLBCL, is increasing.(47) Further, the proportion of HIV+ DLBCL patients with prior ART exposure will continue to increase as accessibility to ART improves and life expectancy of PWH increases.(48,49) Despite relatively small sample sizes, we observe distinct patterns that may be valuable to prognostication and treatment decisions in this population.

Thus, in our cohort we establish an impact of not just HIV, but prior ART exposure on patient outcomes, TME composition, and T-cell response. Given the generalized epidemic of HIV in Malawi, paired with successful ART-scale up efforts, our unique cohort allows for investigation of HIV− and HIV+ patients with further stratification by ART exposure across various demographics, minimizing potential confounding variables. This clinical and immunological study conducted in one of the lowest-income countries in the world can help provide such insights to potentially benefit HIV+ and HIV− DLBCL patients worldwide. This work contributes to the groundwork for improving cancer care in a high-risk population, and further work elaborating these tumor:host interactions and functionalities will be necessary to elucidate potential therapeutic avenues.

STATEMENT OF SIGNIFICANCE.

HIV results in an increased risk of non-Hodgkin lymphoma, of which diffuse large B-cell lymphoma is the most frequent globally. However, how HIV and treatment with antiretroviral therapy (ART) impact DLBCL tumorigenesis and immune response remains largely unknown. Here, we establish associations between HIV/ART and outcome, tumor microenvironment, and tumor-T-cell interactions in a unique cohort of HIV−, HIV+/ART-naive, and HIV+/ART-experienced patients from Malawi.