CD274 (PD-L1) Copy Number Changes (Gain) & Response to Immune Checkpoint Blockade Therapy in Carcinomas of the Urinary Tract

INTRODUCTION

Immune checkpoint inhibitors (ICIs) induce durable responses in a minority of patients with locally advanced and metastatic urothelial carcinoma (UC), but the degree and durability of response varies in those that do benefit clinically [1]. There is thus an urgent need to identify pre-treatment biomarkers to identify patients likely to respond to ICI or who exhibit intrinsic resistance. One potential biomarker of ICI response is copy number change at 9p24.1, which includes the loci for both PD-L1/CD274 and PD-L2/PDCD1LG2 [2–6]. In tumors with PD-L1 amplification, increased production of PD-L1 is thought to result in its greater binding to the PD-1 receptor on cytotoxic T-cells, leading to immune evasion [7]. While prior studies have indicated that amplification of PD-L1 may be associated with response to antibodies targeting PD-L1/PD-1 in a variety of cancer types, the association between PD-L1 copy number alterations and PD-L1 expression and the clinical implications of PD-L1 amplification in UC remain poorly characterized [2–4, 6].

MATERIALS AND METHODS

This study was approved by the institutional review board (12–245, 06–107) and informed consent was obtained from all participants. To define the prevalence of PD-L1 copy number alterations in tumors of the urinary tract, we analyzed a cohort of 1050 carcinomas of the bladder and upper urinary tract that were prospectively sequenced using the MSK-IMPACT next generation sequencing (NGS) assay [8]. Amplifications were defined as a fold change≥2.0 and deletions were defined as fold change≤–2.0 (ref 5). Fluorescence in situ hybridization (FISH) for JAK2/INSL6 and PD-L1/PD-L2 genes was performed, as previously described [3], on 3 of the tumors with PD-L1 amplification by MSK-IMPACT that had tissue available and on 2 additional cases, due to high expression of PD-L1 by immunohistochemistry (IHC). An amplified signal was defined as > 10:1 ratio of PD-L1 signal to the reference centromeric probe. IHC for PD-L1 (clone E1L3N) was performed on 8 and 4 tumors with PD-L1 amplification and deletion, respectively, and scored using the combined positivity score (CPS), which is calculated by the sum of PD-L1-expressing tumor and immune cells, divided by the total number of tumor cells and multiplied by 100. A cutoff of 10 is considered clinically actionable [1]. We also performed PD-L1 IHC staining on 74 UC tumors that did not harbor CD274/PD-L1 amplification by MSK-IMPACT. The cBioPortal.32e34 platform was used to analyze data from The Cancer Genome Atlas [9].

RESULTS

We identified 7 tumors (0.66%) with PD-L1 amplification and 9 (0.86%) with PD-L1 deletion (Fig. 1, S-Tables 1 and 2, S-Molecular Data). Two additional tumors that did not undergo MSK-IMPACT were found to harbor PD-L1 amplification by FISH after diffuse PD-L1 expression was identified by IHC. The histopathologic tumor types with PD-L1 amplification included UC, not otherwise specified (n = 5), small cell/neuroendocrine carcinoma (n = 3), and a poorly differentiated carcinoma with clear cell adenocarcinoma features (n = 1). Within the same cohort of 1050 tumors, 25 additional tumors were of small cell/neuroendocrine carcinoma histology that did not harbor CD274/PD-L1 amplification.

Fig. 1

Molecular profiling and immunohistochemistry. This oncoprint depicts the results of molecular profiling and IHC for PD-L1, as well as other relevant alterations. Tumors with PD-L1 amplification were associated with better response to ICI treatment. The one tumor with PD-L1 deletion with durable response is an HPV-associated carcinoma.

The CPS range in PD-L1 amplified tumors was 13 to 120 (median: 104, n = 8) versus UC tumors without PD-L1 amplification (range 0–70, median: 2, n = 74; p < 0.0001, S-Figure 1). Notably, high CPS in CD274/PD-L1-amplified tumors was primarily driven by PD-L1 expression on tumor cells, not immune cells, as six of the eight tumors tested using IHC expressed PD-L1 on nearly 100%of tumor cells. For comparison, from the 74 tumors without CD274/PD-L1 amplification that were stained with PD-L1, only 14 (19%) had a CPS≥10. In six of these 14 tumors, PD-L1 expression was identified nearly exclusively on immune cells, not on tumor cells (CPS range: 10–20). Moreover, three tumors with the highest levels of PD-L1 expression (CPS of 40, 65 and 70) and no CD274/PD-L1 amplification exhibited sarcomatoid (n = 2) and squamous differentiation (n = 1), which are known to be associated with high PD-L1 expression levels [10, 11]. The remaining five tumors expressed PD-L1 on 45%, 15%, 15%, 10%and 8%of tumor cells, corresponding to a CPS of 50, 20, 18, 15 and 11, respectively. Our results suggest that tumors with very high levels of PD-L1 expression on tumor cells may be enriched for CD274/PD-L1 gene amplification, particularly those that don’t have squamous or sarcomatoid morphology. Furthermore, our findings indicate that there are other mechanisms of PD-L1 overexpression beyond CD274/PD-L1 gene amplification that may be secondary to tumor-intrinsic factors or related to tumor microenvironment, which require further investigation.

Few studies have explored the correlation between CD274/PD-L1 copy number gain and response to ICIs [2–4, 6]. Some reports suggest that CD274/PD-L1 amplification may be a predictor of favorable response to ICI therapy independent of tumor mutational burden [2–4]. In our cohort of the 9 patients whose tumors harbored CD274/PD-L1 amplification, 6 received ICIs (Fig. 2A-D, S-Figures 2, 3, S-Table 1), five of whom had recurrent or metastatic disease at the time of initiation of ICI therapy. Clinical benefit, defined as time on ICI treatment of 180 days or greater was achieved in 4 of the patients, including 2 patients with durable responses of 68 (Case 1, S-Figure 2) and 29 months (Cases 2) [12]. More details about the cohort are available in S-Table 1. While our findings indicate that patients whose tumors harbor CD274/PD-L1 amplification may have higher response rate to ICI, our cohort of CD274/PD-L1-amplified tumors is very small and cannot explain the response to ICI in patients whose tumors don’t harbor such genetic alterations. We further explored whether tumor mutation burden (TMB) was associated with response to ICI therapy in this cohort beyond PD-L1 copy number alterations. TMB data were available for of 7 of 9 PD-L1-amplified and all 9 PD-L1-deleted tumors (see supplementary molecular data table). For the 4 patients who derived clinical benefit, TMB was 7.9, 9.7, 11.2 and 2.6 mutations/megabase, respectively. Using the current cutoff of 10 mutations/ megabase based on recent FDA approval of pembrolizumab in TMB-high tumors, two of these four cases would not be considered TMB-high, while a third case has a borderline TMB value (9.7 mutations/ megabase). Conversely, at least 2 patients with TMB high tumors and PD-L1 gene deletion (15.8 and 25.5 mutations/ megabase, respectively) exhibited progressive disease while on immunotherapy. These findings suggest that TMB by itself does not explain the response to immunotherapy in the 4 patients whose tumors harbored PD-L1 gene amplification.

Fig. 2

Copy Number Assessment, Histopathology & Imaging. A copy number plot (A, Case 4) with relative (Log₂) tumor/normal ratios (y-axis) and corresponding chromosomes (x-axis) is displayed, with each blue dot representing an individual probe region. Amplified regions are shown in red and the inset shows a 20-fold gain based on the FACETS algorithm (inset; chromosome 9, arrow indicates 9p24.1 amplicon). Representative histopathology (B), PD-L1 IHC (C) and FISH using probes for PD-L1/PD-L2 (D) is shown. A copy number plot for a second patient is depicted (E, Case 10) and genomic deletions are represented with red dots, confirmed using the FACETS algorithm (inset; chromosome 9, arrow indicates 9p24.1 amplicon). Baseline CT imaging for this patient (Case 10) prior to initiation of immunotherapy (F) and at the time of disease progression, at 735 days of follow up (G, H), is shown.

Studies of acquired resistance to ICIs in melanoma have implicated genes that regulate tumor cell interaction with T-cells (B2M) and genes involved in upregulating PD-L1 expression (JAK1, JAK2) [13]. It has thus been hypothesized that deletions encompassing the 9p24.1 locus (PD-L1, PD-L2 and JAK2) may also lead to ICI resistance. In our cohort, we identified 9 UC tumors with PD-L1 copy number losses (Fig. 1, Fig. 2E-H, S-Tables 1 and 2, S-Molecular Data). IHC was performed on five such tumors and no PD-L1 expression was detected on the tumor cells in any of them. Of the 4 patients who were treated with ICIs, 3 developed progressive disease (Case 11, 12 and 16; Fig. 2 E-H). The one patient with evidence of CD274/PD-L1 copy number loss who derived durable clinical benefit to ICI had an HPV-associated bladder carcinoma (Case 10). Whether this HPV association contributed to the durable response in this case is not clear as studies on HPV-associated carcinomas in other sites and their response to immunotherapy have yielded mixed results[14, 15]. Other co-alterations of note included a loss of function alteration in JAK2 in one patient who did not receive ICIs (Case13).

Notable recurring molecular co-alterations in tumors with PD-L1 loss included activating FGFR3 alterations in 4 cases (Case12, 14, 15, 17). FGFR pathway alteration has been associated with a poor prognosis and resistance to immunotherapy [16]. Thus PD-L1 deletions should be further investigated for their co-occurrence with FGFR3 alterations as a possible contributory mechanism of ICI resistance.

CONCLUSION

CD274/PD-L1 amplification was associated with high PD-L1 expression as compared to non-amplified UC tumors. Combining our single-institution experience with the TCGA dataset (10 amplified and 5 deleted cases), the combined incidence of amplification events at the 9p24.1 locus was 1.16%(17 of 1456) and that of corresponding deletion events was 0.96%(14 of 1456; see S-Table 3 for a list of the TCGA cases with amplification and deletion) [17]. As four of six patients with CD274/PD-L1 amplification treated with ICI derived clinical benefit, our results suggest that CD274/PD-L1 amplification may contribute to ICI sensitivity and warrants further investigation as a potentially predictive biomarker of response to ICI.

Limitations of the current study include its retrospective design and the fact that all patients studied were treated at a single tertiary institution, factors that may have led to selection bias.

As the incremental cost of profiling tumors for PD-L1 amplification and deletions is minimal for patients undergoing comprehensive NGS-based molecular profiling, PD-L1 copy number analysis may be an easily measured and cost effective biomarker that, in combination with other tumor and host factors, may provide better prediction for response to ICI therapy. Given the rarity of CD274/PD-L1 amplification in UC, future studies on the influence of CD274/PD-L1 copy number alterations on ICI response will need to combine prospective genomic datasets with detailed demographic and treatment data generated at multiple institutions.