Key differences exist in the TME of EBV+ vs. EBV- pediatric cHL, including increased M1 macrophage polarization and a cytotoxic/Th1 viral response in EBV+ cases (35–37). In addition there are contrasts between the TME in pediatric and adult EBV+ cHL, suggesting that adults may have reduced anti-cancer immunity and an “aged” TME, which could explain the inferior clinical outcome (31, 35, 38). EBV+ cHL in adults is enriched for FOXP3+ Tregs (37, 39) and immunosuppressive cytokines such as IL-4, IL-5, IL-6, IL-9, IL-10, IL-13, and TGF-β. In contrast, in EBV+ cHL in children has increased cytotoxic/T-helper cell 1(Th1) CD8+ T-cell infiltration characterized by TIA-1 and T-bet expression (36) and M1-polarized TAMs which may potentially contribute to the favorable outcome through effective immune surveillance (31, 34). The biology underlying differences in EBV+ TME across ages including the potential role of senescence in older adults remain unclear, requiring further study across all age groups.

Gene expression profiling (GEP) of cHL biopsies has expanded our understanding of the TME and allowed for the development TME-based biomarkers. Scott et al. developed a 23-gene expression prognostic model, HL27 (40), which predicted outcome among adults with advanced-stage cHL. To evaluate the applicability of this model in pediatric HL, Johnston et al. performed GEP on cHL tumors from pediatric patients treated on the Children’s Oncology Group (COG) trial, AHOD0031 (5) which evaluated a risk adapted-approach to the treatment of intermediate-risk cHL. HL27 (40) did not predict outcome in the pediatric cohort, suggesting potential age-related biological differences in the TME. Indeed, Spearman correlation analysis of GEP-based TME component scores with age revealed that eosinophil, B-cell, and mast cell signatures were more prevalent in younger patients, while macrophage and stromal signatures were more pronounced in older patients (41). Based on these observations, Johnston et al. developed a distinct GEP-based predictive model for pediatric cHL, PHL-9C, which is predictive of 5-yr event-free survival (EFS) among children treated on COG AHOD0031 (41). This model is composed of: Tregs, mast cells, T helper 2 (Th2) cells, myeloid-derived suppressor cells, and HRS cells (41). We are currently investigating the utility of PHL-9C in brentuximab vedotin-containing treatment (5).

Recent technical advances in multiparameter imaging (MPI) have allowed for a more comprehensive characterization of the TME at single cell resolution (11). These approaches provide details on co-expression patterns, cellular composition, and cell-to-cell spatial interactions, which have allowed for detailed descriptions of the architecture of the TME in cHL. A series of studies in adult cHL have utilized MPI to better understand the mechanisms of response to therapy and identify biomarkers. MPI of cHL tumors has identified a spatial relationship between TAMs that express PD-1 ligand (PD-L1) and HRS cells (42). Unexpectedly, the PD-L1+ macrophages outnumbered PD-L1+ HRS cells in the TME of cHL and co-localized with T-cells, suggesting that this interaction may be a key target of PD-1 blockade. This may explain the efficacy of PD-1 therapy in cHL despite frequent loss of B2M leading to decreased expression of MHC-I (7, 14, 43–46). Further work by Aoki et al. evaluating the TME in paired diagnostic and relapsed cHL samples identified a spatial interaction between CXCR5+ HRS cells and CXCL13+ TAMs (47). Although previous biomarker studies in cHL had revealed an association between the abundance of macrophages and clinical outcome in adult cHL (48–50), the specific cellular interactions and their relevance in refractory disease were not known. Using spatial TME information, Aoki et al. developed a model predictive of outcome in the relapsed setting (51). The model (RHL4S) is composed of spatial scores from 4 variables (51, 52), each of which is independently associated with failure free survival after autologous stem cell transplant: CXCR5+ HRS cells, PD1+ CD4+ T-cells, CD68+ macrophages, and CXCR5+ non-malignant B-cells.

Single cell suspensions of lymphoma tissue represent another resource to investigate complex TME ecosystems with high resolution (11). Utilizing time-of-flight cytometry (CyTOF), Cedar et al. found that cHL has a TME that is enriched for CD4+ T-cells. Within this population there is an expansion of Th1 polarized CD4+ T effector cells that express PD-1, and may represent another relevant target of PD-1 blockade. This work also identified Th1 polarized Tregs which did not express PD-1 (53). More recently, single cell RNA sequencing (scRNA-seq) has provided an opportunity to define the phenotypes of individual HRS cells and immune cells in the cHL TME (51, 54, 55). Aoki et al. integrated both scRNA-seq and MPI to study cHL in adults. This work has defined an interaction between HRS cells and type 1 regulatory (Tr1) T-cells with enhanced expression of the inhibitory receptor LAG3. LAG3+ T-cells, which are functionally immunosuppressive, co-localize with HRS cells that do not express MHC class II (54), providing a potential mechanism for immune evasion in this cellular subset.

To date, single cell level studies in cHL have largely been restricted to adult cohorts. There is evidence, however, that the TME at the single cell level may differ across age groups. In a recent study Stewart et al. performed scRNA-seq and MPI on cHL cases and integrated their data with an earlier report (56) to include cases across the age spectrum (6-80y). This work identified an abundance of mononuclear phagocytes including dendritic cells and monocytes in the vicinity of HRS cell and found that these cells express immune checkpoints including PD-L1 and TIM3. The expression of immune checkpoints in mononuclear phagocyte populations increased with age (56), indicating the potential for distinct myeloid cell biology across age groups. Indeed, as opposed to the strong evidence in adult cHL (50) for a prognostic role of TAMs, this relationship is not observed in pediatric cHL (31, 35, 57).

Other studies evaluating the TME in pediatric cHL on the single cell level have been limited to flow cytometry and traditional imaging techniques. In a study evaluating the TME in pediatric cHL by flow cytometry, cHL cases were found to have high CD7 expression and an expansion of CD45RO+ T-cells. Of note, elevated CD7 has also been described in adult cHL (58, 59). An investigation of LAG3+ T-cells in pediatric cHL using IHC on patients treated on the COG AHOD0031 revealed that 73/115 (63%) of the baseline pediatric cHL tumors demonstrated LAG3+ staining in the TME (60).

HRS cells harbor several genomic alterations that enable them to modulate the TME resulting in immune privilege ( Figure 2 ). Upregulation of programmed death-ligand 1 (CD274/PD-L1) and 2 (PDCD1LG2/PD-L2) is a hallmark of HRS cells and a key mechanism of immune evasion (67). Overexpression of PD-L1 and PD-L2 can occur via polyploidy/genome duplication, arm-level gain of 9p, or focal amplification of 9p24.1, leading to amplification of both the PD-L1 and PD-L2 gene loci (18, 68). 9p24.1 amplification is detected in approximately 75% of cases (18, 69). Immune evasion by HRS cells is further mediated by alterations in the MHC class I and II, which decrease the recognition of malignant cells by cytotoxic CD8+ and CD4+ T cells, respectively.

Genomic sequencing of HRS cells and ctDNA from patients with cHL have identified loss-of-function alterations in B2M, a subunit on MHC class I, in 33-70% of cHL cases (15, 18, 19). In most cases, B2M alterations are biallelic. Ectopic expression of B2M in a HL cell line with B2M mutation restored MHC I expression, highlighting the key role of these alterations in loss of MHC I in cHL (15). In an examination of B2M protein expression in a large cohort of cHL tumors by immunohistochemistry, it was observed that B2M loss was more common in younger patients (median age 30 vs. 47, p<0.0001), suggesting that this mutation may be a feature of cHL in children and AYAs (15). Alterations in other components of MHC-I have also been described in HRS cells. Truncating and missense mutations in HLA-B have been identified in 17-20% of cHL cases (18, 19). Additional structural variants and complex events involving HLA-B have also been reported (19).

Alterations impacting MHC class II also occur recurrently in cHL. Structural variants in Class II Major Histocompatibility Complex Transactivator (CIITA), which regulates transcription of MHC class II genes, are observed in 9-16% of cases (19, 70, 71). Structural variants, including balanced translocations, and complex rearrangements of CIITA downregulate CIITA and are associated with reduced expression of MHC class II (18, 70, 72).

Gain-of-function mutations in interleukin-4 receptor (IL-4R), which mediates IL-4 and IL-13 signaling, are present in 5-10% of cHL cases (19, 63). The majority of these mutations are truncating and clustered at the cytosolic C terminus of the IL-4R protein, leading to disruption of the immunoreceptor tyrosine-based inhibitory motif (ITIM). In vitro studies have demonstrated that IL-4R mutations induce downstream expression of STAT6, leading to increased synthesis of CCL17, a regulatory T cell attractant, providing another possible mechanism of immune tolerance (63).

As information emerges regarding the genomic landscape of cHL, key questions that can now be addressed include: 1) can cHL be divided into clinically relevant molecular subtypes?; 2) is cHL in pediatrics distinct from that in adults? In a recent report of whole genome sequencing of HRS cells from pediatric and adult cases, a higher mutational burden was observed in pediatric and AYA patients compared to older adults >50y. In addition, an accelerated rate of the “aging” mutational signatures SBS1 and SBS5 was observed among pediatric and AYA patients (19). The increased mutational burden was independent of sequencing coverage and EBV status. This suggests that pediatric and AYA cases may have a distinctively high mutational burden which is not observed in older adults.

Molecular subtypes of cHL have been defined from a large cohort of cases (n=366) that were profiled using ctDNA in a recent report by Alig et al (58). In this work, two clusters emerged: the H1 cluster is characterized by younger age (median age = 30y vs. 42y in the H2 cluster, p=0.02), higher mutational burden, and mutations in NF-kB, JAK/STAT and PI3K signaling. The H2 cluster had a more even age distribution and was defined by lower mutational burden, more frequent somatic copy number alterations, and mutations in TP53 and KMT2D. Patients in the H2 cluster had an inferior progression free survival. In addition, an analysis of the transcriptional signature of patients >65y revealed a distinct signature defined by lower rates of cytokine response over T-cell activation signatures. Heger et al. also recently studied a cohort of 243 patients with cHL by both ctDNA to profile HRS cells and RNAseq to characterize the TME (73). This work identified three clusters of HL: “oncogene driven HL”, which resembles the H1 cluster identified by Alig et al. with high mutational burden, alterations in HL driver genes, and a cold TME; “inflammatory immune escape HL” characterized by copy number changes resulting in immune escape; and “virally-driven HL” which was enriched in cases with EBV and/or human herpes virus 6. Aoki et al. independently identified 4 molecular subgroups using DNA of enriched HRS cells, which revealed cluster correlations with clinical parameters (e.g. EBV status and age) similar to those described in Alig et al. (H1/H2 clusters) (74, 75). Additionally, each of the four molecular subgroups demonstrated distinctive TME correlates and unique gene expression signatures in HRS cells.

Despite differences in methodology used to define molecular subtypes—including sequencing methods, clustering algorithms, and the number of genes analyzed—all molecular classifications have identified certain common subtypes associated with mutational burden, age, and EBV status. However, the underlying biology of each molecular subtype and their targetable vulnerabilities remain unknown. Moving forward, further biological investigations and methodological optimizations, including the development of publicly available tools applicable to individual cases such as the LymphGen system, will be essential to implement the classification system in future clinical trials and/or routine clinical practice (76).