We analyzed WES data to provide a genetic characterization of 58 curated immune surveillance genes in dDLBCL and rrDLBCL, post-treatment patients, which is essential for understanding the impact of the treatment in a genomic context. More than 70% of both dDLBCLs and rrDLBCLs patients harbor alterations in immune surveillance genes in our data, which is in concordance with external cohorts. No significant difference in individual gene mutation frequencies of immune surveillance genes between dDLBCLs and rrDLBCLs was observed in our data or in external cohorts, but in our rrDLBCL patients a smaller number of affected genes (
n = 11) were observed than in dDLBCL (
n = 35). A finding differing from external validation cohorts possibly introduced by diagnostic samples containing both cured and later relapsing patients, and in the rrDLBCL cohort, patients with both short and long time to new relapse could be included. However, more than half of genes affected in rrDLBCLs showed higher gene mutation frequency than in dDLBCLs, some near double (
HLA-A) in our data and external cohorts (Figs.
1 and
5). Hence, during tumor progression and development of resistance in rrDLBCLs, we observed higher gene mutation frequencies in
PIM1, CD58, FAS, HLA-A, and
TNFRSF14 as compared to dDLBCL consistent with other studies analyzing global genomic sequencing data [
19,
27,
43] and illustrated the loss of immune surveillance (
CD58, HLA-A) as well as immune suppression (
FAS, TNFRSF14, PIM1) [
19,
43]. Recognition and elimination by T- and NK- cells are avoided by mutated
CD58 and
HLA-A genes [
27,
44], which both have increased incidence in rrDLBCLs compared to dDLBCLS suggesting decreased immunogenicity of the tumor cells in the refractory and relapsed situation. In addition, observations from matched diagnostic and relapsed samples of the same patient (
n = 3) depict decreased variant allele frequency in antigene presenting genes (e.g.,
HLA-A) but not vanishing after R-CHOP or R-CHOP-like treatment, suggesting a possible tumor evasion mechanism. This observation supports suggestion from Wise et al., 2020 that antigen presentation represents a key target for genetic alterations in rrDLBCL resistance [
27]. The loss of function mutations in
FAS gene lead to the suppression of Fas/FasL system responsible for activation-induced cell death [
45]. In contrast, the loss of function mutations in the
TNFRSF14 gene leads to B-cell autonomous activation as well as extrinsic activation of the lymphoma microenvironment through B and T-lymphocyte attenuator (BTLA attenuator) located on CD4+ T-helper cells [
46]. Specific missense mutations in
PIM1 are possibly activating. Along with PIM1 being overexpressed in DLBCL cells compared to normal B-cells, tumor cells are prevented from undergoing apoptosis inactivating proteins such as apoptosis signaling kinase 1 (ASK1), preventing further activation of FAS ligand [
47,
48]. However, neoantigen presentation is necessary for immune surveillance, and lack of expression of MHC molecules might be the reason for failed anti-PD1 immunotherapies [
10,
49]. In particular, if the cell is unable to present neoantigens in association with MHC molecules, there is no need for PD1/PD-L1 interaction [
10]. Therefore, these patients might be good candidates for other immune therapies such as CAR-T-cell therapy, which has shown long-term response in approximately 58% of rrDLBCLs [
27,
50], or bispecific CAR-T cell therapy that has approximately 80% overall response rate if patients receive freshly produced anti-CD19/20 [
51]. Nevertheless, it is interesting to speculate that clonal selection of genetic variants in antigen-presenting genes occurs during or after the treatment resulting in the development of rrDLBCL (Fig.
4) even if we cannot distinguish single cell double genetic events from polyclonal tumor formations.
Recently, several algorithms have been developed, providing a refined classification of DLBCL into five to seven distinct subtypes based on genetic features [
24,
52,
53]. As these genetic classes are based on global genetic analysis, and we use only a sub-selected set of genes in our analysis, we did not include refined genetic classification. However, it is observed that 73% of MCD genomes acquired genetic variants in genes affecting immune surveillance, thus becoming invisible to the host immune system, suggesting a crucial role in DLBCL pathogenesis, which is in agreement with cluster 5 described by Chapuy et al.,2018 [
24,
40].
Also, an important observation in our dDLBCLs and rrDLBCLs is that 16 and 29% of the patients, respectively, harbor mutations in antigene presenting genes excluding genes in immune suppression and exhaustion, while 13 and 12% harbor mutations in genes related to immune suppression and exhaustion but not in antigen- presentation, respectively (Table S2). Similar features are observed in external dDLBCL and rrDLBCL cohorts where 15 and 28% of patients, respectively, are affected by mutated antigen-presenting genes and 25 and 25% affected by genes involved in immune suppression and exhaustion, respectively. Thus, a higher mutational rate of antigene presenting genes in rrDLBCL than in dDLBCL can be observed, though findings are not significant in neither our nor external cohorts – perhaps due to small cohort sizes it may suggest their possible role in the development of resistance toward therapy. Additionally, decreased proportion of M1 in rrDLBCLs may suggest a less aggressive host immune response by M1 in rrDLBCLs compared to dDLBCLS, but not restricted to rrDLBCLs harboring genetic alterations in immune surveillance or MHC related genes since no difference was observed in M1 proportions between rrDLBCLs with and without mutated immune surveillance or MHC associated genes. Also, no significant difference was observed in paired samples suggesting that the host does not respond or detect a difference in the tumor even if the IS genes are mutated and become invisible. Due to the limited number of paired samples, this observation requires more samples and tumor microenvironment analysis for validation.