Higher tumor mutational burden and PD-L1 expression correlate with shorter survival in hematologic malignancies

Background: The prognostic implications of tumor mutational burden (TMB) and programmed death ligand 1 (PD-L1) expression are poorly studied in hematologic malignancies. Objectives: This study aimed to better understand the characteristics and prognostic value of TMB and PD-1/PD-L1 in hematologic malignancies. Design: This real-world study was conducted among patients with hematologic malignancies who had next-generation sequencing (NGS) (Foundation Medicine) at the University of California San Diego Moores Cancer Center (2014–2018). Methods: TMB was measured by NGS. PD-L1 expression (tumor proportion score, TPS) was measured by immunohistochemistry (classified as high (⩾50%), low (1–49%), and negative (<1%)). Data was curated from the electronic medical records. Results: In 388 evaluable patients, the most common diagnoses were B-cell non-Hodgkin lymphoma (NHL) (35%) and Philadelphia chromosome-negative myeloproliferative disorders (16%). Median TMB was 1.6 mutations/Mb (range, 0–46.83). Forty-eight patients (12%) had TMB ⩾10 mutations/Mb, 90% of which were B-cell or T-cell NHL. In 85 samples with available PD-L1 scores, 11 were high; 26, low; and 48, no tumor cell expression. PD-L1 TPS positive (⩾1%) was most common in T-cell NHL (7/9 (77%) cases) followed by B-cell NHL (21/51 (41%) cases). TMB ⩾4 mutations/Mb and PD-L1 score ⩾1% were significantly associated with shorter overall survival (OS) from diagnosis, with hazard ratio (HR) = 1.46 (p = 0.02, 95% confidence interval (CI) 1.05–2.03) and HR = 2.11 (p = 0.04, 95% CI 1.04–4.30), respectively; the relationship was more pronounced when PD-L1 ⩾50% versus <50% was used (HR = 2.80, p = 0.02, 95% CI 1.19–6.59). Higher TMB and higher PD-L1 positivity correlation were significant but weak (Pearson correlation coefficient R2 = 0.04, p = 0.04). Conclusion: TMB ⩾4 mutations/Mb and positive PD-L1 TPS are poor prognostic factors, correlating with shorter OS across hematologic malignancies. Trial registration: ClinicalTrials.gov NCT02478931.


Introduction
2][3][4][5][6] These biomarkers have the major histocompatibility complex (MHC), which may allow for better recognition by the immune system upon activation by ICIs.Meanwhile, the PD-1/PD-L1 pathway is one of the inhibitory receptors of the immune system that is exploited by cancer cells to evade immune surveillance; blocking this pathway leads to a cytotoxic antitumor response. 7TMB and PD-1/ PD-L1 expression are routinely assessed and used to guide therapeutic decisions in solid malignancies.TMB ⩾10 mutations/Mb is a tissue-agnostic biomarker that predicts response to pembrolizumab, as demonstrated in the pivotal clinical trial KEYNOTE-158, 8 resulting in its approval by the Food and Drug Administration for this purpose. 9TMB ⩾16 mutations/Mb appears to be predictive of response in patients treated with atezolizumab. 10e prognostic role of TMB and PD-1/PD-L1 expression is less clear.We previously demonstrated that low TMB of <5 mutations/Mb and high TMB of ⩾20 mutations/Mb portends better prognosis in ICI naïve patients while intermediate TMB, defined as >5 mutations/Mb and <20 mutations/Mb, correlates with worse prognosis. 11Studies demonstrate that the immune signature in cancers with low and high TMB can differ depending on the type of cancer, which could explain some inconsistencies in the association between TMB and ICI outcomes. 12urthermore, increased PD-L1 expression is correlated with worse prognosis in multiple solid cancer types. 13deeper understanding of the predictive role of TMB and PD-1/PD-L1 in treatment with ICI has led to breakthrough treatment options for solid tumors, but their efficacy in hematologic malignancies has been mixed.Currently, checkpoint inhibitors are only FDA-approved for two hematologic indications: relapsed/refractory classical Hodgkin lymphoma (R/R cHL) and relapsed/ refractory primary mediastinal B-cell lymphoma (PMBCL). 14In both of these conditions, checkpoint inhibitors have been demonstrated to be at least as effective, if not more, than in their solid organ counterparts.In R/R cHL clinical trials, CHECKPOINT 205 15 demonstrated an objective response rate (ORR) of 69% with singleagent nivolumab while KEYNOTE-087 16 yielded an ORR of 71.9% with pembrolizumab.The ORR observed in KEYNOTE-013 and KEYNOTE-170, 17 two clinical trials evaluating pembrolizumab monotherapy in R/R PMBCL, was 48% and 45%, respectively.[23] The TMB and PD-1/PD-L1 characteristics in hematologic malignancies remain poorly defined, let alone their prognostic and predictive roles.These have been best described in cHL in which 9p24.1 amplification in cHL is associated with higher expression of PD-L1 and worse progression-free survival (PFS). 24Similar to solid malignancies, higher PD-L1 expression in cHL has been associated with improved prognosis when treated with nivolumab. 18Therefore, it is important to understand the characteristics of TMB and PD-1/PD-L1 in hematologic malignancies, which we explore in this report.

Methods
This study was conducted in accordance with the University of California San Diego Internal Review Board-approved PREDICT study (NCT02478931) and any investigational interventions for which patients consented.All patients had hematologic malignancies treated at the University of California San Diego Moores Cancer Center and had next-generation sequencing (NGS) on the FoundationOne ® Heme panel performed between 2014 and 2018.A review of the electronic medical records was performed.This was a real-world retrospective study and included all patients with available clinical-grade TMB data who participated in the PREDICT study.The reporting of this study conforms to the Strengthening the Reporting of Observational Studies in Epidemiology statement 25 (Supplemental Table 1).
Tumor tissue profiling by massively parallel NGS of 406 genes and select introns of 31 genes involved in rearrangements as well as RNA sequencing (cDNA) of 265 commonly rearranged genes fusions were performed using the FoundationOne Heme assay (Foundation Medicine, Cambridge, MA, USA) as described previously. 26,27TMB was determined on 1.2 megabases of DNA sequence, excluding germline and driver alterations from the calculation. 28icrosatellite instability (MSI) was determined by examining homopolymer repeat regions. 29-L1 scores of the tumor were classified as high (⩾50%), low (1-49%), and negative (<1%).PD-L1 tumor proportion score (TPS) was determined by immunohistochemistry (IHC) performed at Foundation Medicine.The assay utilized Dako (22C3) or Ventana (SP142) anti-PD-L1 antibodies and was performed according to manufacturer instructions on Formalin-fixed, paraffin-embedded (FFPE) tissue of the tumor.
Descriptive statistics were utilized to summarize patient characteristics.PFS and overall survival (OS) were visualized using Kaplan-Meier and quantified with Cox proportional hazards regression reporting two-sided p-values.Correlation of the TMB and PD-L1 score was assessed by Pearson correlation coefficient, and the difference between the medians of TMB by disease group was calculated using one-way analysis of variance (ANOVA).A confidence interval (CI) of 95% was used for all analyses.

Results
A total of 480 patients were identified.After applying the exclusion criteria, 388 patients were included in final analysis (Supplemental Figure 1).The median age at diagnosis was 61 years and 55.8% were men.The most common diagnosis was B-cell NHL (35%) followed by Philadelphia chromosome-negative myeloproliferative neoplasm (MPN) disorders (16%) and myelodysplastic syndrome (MDS) (12%).None of the 302 patients who were tested for MSI status were MSI-High.TMB and PD-L1 scores are shown in Table 1.TMB differed by hematologic disease type: The median TMB of the entire cohort was 1.6 mutations/Mb.The median TMB of B-cell NHL was 4.0 mutations/Mb, and when chronic lymphocytic leukemia (CLL), which had a median TMB of 0.8 mutations/Mb, was excluded, the median was 7.7 mutations/Mb.The median TMB in T-cell NHL was 1.6 mutations/Mb and that of MM was 4.0 mutations/Mb.There was a significant difference in median TMB by disease type (one-way ANOVA, F = 11.7,p < 0.001) (Figure 1).Post-transplant lymphoproliferative disorder had the highest median TMB (14.5 mutations/ Mb) while MDS/MPN overlap syndrome had the lowest median TMB (0.8 mutations/Mb).
Correlation between TMB and PD-L1 expression: Similar to a prior study, 31 the correlation between higher TMB and higher PD-L1 positivity was statistically significant, but weak (Pearson correlation coefficient R 2 = 0.04, p = 0.04) (data not shown).

Discussion
6]32 We noted that the TMB in hematologic malignancies tends to be low, especially among the myeloid malignancies.The median TMB was higher in B-and T-cell NHL.If the conventional cutoff of TMB ⩾10 mutations/Mb is used, as therapeutically relevant given the FDA approval for pembrolizumab, 9 only 12% of all tumors were TMB-high, 90% of which were B-cell or T-cell NHL.Similar to TMB-high status, a positive PD-L1 status in our study was observed mostly in patients with lymphoma.Notably, two of two cHL were positive for PD-L1, which is predicted by the 9p24.1 copy number alteration characteristically seen in cHL. 33B in various lymphoid malignancies has been poorly described in the literature, but there are a few notable studies.Cho et al. 34  We also investigated the prognostic value of PD-L1 TPS for survival.In our entire cohort of samples, a positive PD-L1 TPS score was correlated with worse prognosis compared to a negative PD-L1 TPS score; the patients with PD-L1 ⩾50% had a particularly worse prognosis even when limited by the small number of patients in this group.It is plausible that this worse prognosis is due to the immune system shield that PD-L1 provides.There are several other studies that describe PD-L1 expression in lymphoma, which vary widely in their results.Tumor PD-L1 score is reported to be positive in 10-50% of DLBCL, 70-100% of PMBCL, 0-10% of CLL, and 7-80% of T-cell NHL. 37The prognostic value of PD-L1 score in NHL is also unclear, with some studies reporting worse OS and some suggesting no difference. 37total of 17 patients who had positive biomarkers (either high TMB and/or positive PD-L1 expression) were treated with ICI-based therapy.Notably, all but one patient received combination therapy with other agents.Only 4 of these 17 patients achieved an objective response.The four patients who responded had PFS values ranging from 193 to 951 days.Three of the four patients had B-cell NHL and 1 had T-cell NHL.Importantly, the patient with GZL received brentuximab vedotin and a patient with PCNSL received temozolomide, which are active agents in their respective diseases.It appears that a select group of patients derive benefit from ICI combination therapy.
There are several limitations to this study.Since NGS is not routinely obtained for many hematologic malignancies, patients who were enrolled in this study and had NGS data available are likely to suffer from selection bias.NGS was also obtained at different time points of the disease, with some obtained at the time of diagnosis and others at the time of relapse, although our previous work suggests that TMB remains constant from diagnosis to relapse. 38Another limitation is that the PD-L1 IHC testing was performed using two different antibodies, Dako (22C3) and Ventana (SP142), which may lead to discrepancies in results.Our dataset is also limited by the relatively small sample size; as such, detailed assessments stratified by individual cancer types were not able to be conducted, though our results may suggest applicability across tumor types, especially since similar results (i.e.better prognosis with lower TMB) have been reported in solid tumors. 11Another limitation of the current work is that PD-L1 expression was examined in tumor cells; further studies should also interrogate the microenvironment.Finally, the ICI-treated patient number was small and heterogeneous, allowing for descriptive but not statistical analysis.

Conclusion
In summary, this study describes detailed characteristics of TMB and PD-L1 expression for various hematologic malignancies.Acute leukemias and myeloid malignancies had low TMB while a subset of lymphomas had higher TMB and/or positive PD-L1 expression.TMB cutoff of ⩾4 mutations/Mb and positive PD-L1 TPS scores (especially scores ⩾50%) were each associated with significantly shorter survival.These data suggest that TMB and PD-L1 expression may be important biomarkers for the prognosis of hematologic malignancies in addition to their implications for solid tumors. 11,39,40clarations

Table 1 .
TMB and tumor PD-L1 scores of all patients.

Table 2 .
Characteristics and outcomes of patients who received immune checkpoint inhibitor therapy.