Genomic and Immune Profiling of a Patient With Triple-Negative Breast Cancer That Progressed During Neoadjuvant Chemotherapy Plus PD-L1 Blockade

Preliminary results from neoadjuvant trials combining immune checkpoint blockade (ICB) with standardof-care chemotherapy suggest high pathologic complete response rates that range between 50% and 80% in triple-negative breast cancer (TNBC). Adding atezolizumab to NAB-paclitaxel for first-line treatment of metastatic TNBC significantly improved response rate and progression-free survival compared with NABpaclitaxel alone, suggesting synergy between ICB and chemotherapy. However, not all patients respond to ICB, and a minority exhibit rapid progression of their disease. Patients who experience exceptionally favorable or unfavorable responses provide unique opportunities for studying disease biology and for identifying response markers. Progression during neoadjuvant chemotherapy is a rare event in TNBC. Here, we report results from the molecular analysis of a TNBC that rapidly progressed during neoadjuvant chemotherapy plus programmed death-ligand 1 (PDL1) blockade in a clinical trial Neoadjuvant MEDI4736 Concomitant With Weekly NAB-paclitaxel and Dosedense AC for Stage I-III Triple Negative Breast Cancer (ClinicalTrials.gov identifier: NCT02489448). Among the first 30 patients in this ongoing trial, no other participant experienced disease progression. We also hoped to identify potentially actionable genomic alterations or immunologic features.


INTRODUCTION
Preliminary results from neoadjuvant trials combining immune checkpoint blockade (ICB) with standardof-care chemotherapy suggest high pathologic complete response rates that range between 50% and 80% in triple-negative breast cancer (TNBC). [1][2][3] Adding atezolizumab to NAB-paclitaxel for first-line treatment of metastatic TNBC significantly improved response rate and progression-free survival compared with NABpaclitaxel alone, suggesting synergy between ICB and chemotherapy. 4 However, not all patients respond to ICB, and a minority exhibit rapid progression of their disease. 5 Patients who experience exceptionally favorable or unfavorable responses provide unique opportunities for studying disease biology and for identifying response markers. Progression during neoadjuvant chemotherapy is a rare event in TNBC. Here, we report results from the molecular analysis of a TNBC that rapidly progressed during neoadjuvant chemotherapy plus programmed death-ligand 1 (PD-L1) blockade in a clinical trial Neoadjuvant MEDI4736 Concomitant With Weekly NAB-paclitaxel and Dosedense AC for Stage I-III Triple Negative Breast Cancer (ClinicalTrials.gov identifier: NCT02489448). Among the first 30 patients in this ongoing trial, no other participant experienced disease progression. We also hoped to identify potentially actionable genomic alterations or immunologic features.

CASE REPORT
The 41-year-old premenopausal woman presented with a self-palpated lump in the right breast, and mammogram revealed multifocal T1N1 disease. Core needle biopsy (CNB) of the largest breast lesion (1.7 cm) and of an enlarged right axillary lymph node showed high-grade TNBC. Systemic staging was negative for distant metastases, and germline cancersusceptibility panel testing revealed no deleterious mutations. Baseline tumor-infiltrating lymphocyte (TIL) count was 10% (CD4, 5%; CD8, 5%; and CD20, 1%), macrophage (CD68) was 1%, and tumor cellularity was 50%. TILs were both intratumoral and stromal (ie, this was not a T-cell excluded tumor). The patient agreed to participate in a neoadjuvant phase I/II clinical trial that combined durvalumab (10 mg/kg once every 2 weeks) with once-per-week NABpaclitaxel (100 mg/m 2 ) for 12 cycles and dosedense doxorubicin plus cyclophosphamide (ddAC) for 4 cycles. After 8 weeks of NAB-paclitaxel plus durvalumab, physical examination showed increased tumor size and new skin edema confirmed by repeat mammogram and ultrasonogram. Repeat CNB showed 60% tumor cellularity but also an increase in TIL count to 20% (CD4, 10% to 15%; CD8, 5%; and CD20, 0%) and an increase in macrophages (CD68, 20%). TILs were again noted in the stroma and intratumorally. NAB-paclitaxel was stopped but because of the increased immune infiltration, durvalumab was continued, and the patient was administered ddAC, which led to disease stabilization after four courses of therapy. Because of the apparent clinical benefit, she received two additional courses of ddAC without durvalumab off protocol and underwent right skin-sparing mastectomy and lymph node dissection. Pathology showed extensive multifocal disease (largest focus, 2.4 cm; tumor cellularity, 40%) with lymphovascular invasion in the breast and more than 10 positive axillary lymph nodes (ypT2, ypN3). Immune cell proportions in the mastectomy were TILs, 20%; CD4, 10%; CD8, 10%; CD20, 1%; and CD68, 10%.
DNA and RNA were extracted from formalin-fixed paraffin-embedded sections of tumor samples obtained at baseline, at week 8, and from the mastectomy. Blood DNA served as the reference for somatic mutation calling for whole-exome sequencing. Immune gene messenger RNA expression analysis was performed by using the Nanostring PanCancer Immune Profiling assay (Nanostring Technologies, Seattle, WA) as described in the Data Supplement and as previously reported. 6,7 The patient's gene expression levels were compared with those from ASSOCIATED CONTENT

Data Supplement
Author affiliations and support information (if applicable) appear at the end of this article. Licensed under the Creative Commons Attribution 4.0 License a reference cohort (n = 31) of untreated primary breast cancers analyzed with the same platform. 6 The patient's result was considered to be a significant outlier if her expression levels fell outside the 2.5th to 97.5th percentile range of the bootstrap distribution of the same gene in the reference cohort. TIL counts and immune cell subtypes were determined by routine pathology and immunohistochemistry.
When we compared our patient's baseline expression of 26 immune cell types and 100 immune function metagenes with those in the reference cohort, her neutrophil metagene expression was below the 2.5th percentile, whereas the cell cycle and immunosuppression metagenes were above the 97.5th percentile of the reference distribution (Fig 1A), indicating a highly proliferative tumor with an immunosuppressed microenvironment. All other metagene categories were within the 2.5th to 97.5th percentile range (Data Supplement). The tumor also showed low expression of two previously validated immunotherapy predictive gene signatures, 8,9 indicating low probability of response to ICB ( Fig 1B). Single-gene level analysis revealed significantly low expression of 27 genes including PDCD1, CD8A, and KLRC2 ( Fig 1C), suggesting impaired T-cell and natural killer (NK) cell activity. In addition, we found high expression of 33 genes, including TGFB, HLA-G, CD63, CCL28, and CXCL16 (Fig 1D), which are associated with an immune evasive phenotype, increased cell motility, and invasion. Baseline expression status of all genes is provided in the Data Supplement.
Next, we examined expression changes across the three time points. Overall, greater changes were observed between week 8 and mastectomy (ddAC treatment) than between baseline and week 8 (NAB-paclitaxel treatment). We observed upregulation of FOS, ABCB1, KIR3DL3, GZMA, and GZMK in the mastectomy (single-gene and metagene expression changes are provided in the Data Supplement). Most immune metagenes increased during the second period (Fig 2A), consistent with an increase of immune cell (mainly T and NK cells) infiltration. The upregulation of ABCB1, a multidrug resistance drug efflux pump, was also apparent and may have contributed to chemotherapy resistance in tumor cells. We also observed a concurrent decrease in immunosuppression, leukocyte functions, and leukocyte migration metagene categories, suggesting a potentially ineffective immune response. To study this further, we next analyzed T-cell exclusion and dysfunction features using the tumor immune dysfunction and exclusion (TIDE) method (Data Supplement). 10 In concordance with our previous findings, cytotoxic T-cell infiltration increased during the second period. However, this was accompanied by an increase in T-cell dysfunction score, indicating an ineffective immune response (Fig 2B).
We also performed whole-exome sequencing of the baseline and mastectomy specimens. We identified 80 somatic mutations, including one cancer driver gene, TP53 (R175H). Other mutations with the highest variant allele frequency at baseline included ZNF385C, CACNA1E, NXPE1, and DYNC1H1, which have poorly understood functions in cancer (Fig 2C; Data Supplement). We also detected amplifications in FGFR1, FGF2, FGF3, MYC, MCL1, CCND1, and TGFB2 and deletions in CDKN2A and CDKN2B (Data Supplement). By using Sciclone (Waltham, MA) 11 and ClonEvol, 12 we identified four distinct tumor clones, all present at baseline and in the mastectomy specimen, with little evidence for clonal selection during therapy (Figs 2C-D). In the germline, 32 highly functional impact variants were detected by using Ingenuity Variant Analysis (QIAGEN Bioinformatics, Redwood City, CA; Data Supplement). Of note, the patient was heterozygous for an 11-nucleotide frameshift-indel + insertion in the MH2 domain of the SMAD6 gene (NC_000015.9:g.67073715_ 67073726delinsA). This domain is essential for the function of SMAD6 as a negative regulator of TGFβ signaling. SMAD6 sequesters SMAD2 in the cytoplasm and prevents its translocation to the nucleus. 13 Another potentially relevant germline variant was a heterozygous JAK3 frameshift mutation in the JH5 domain (NC_000019.9:g. 17953278delG). JAK3 is expressed in T and NK cells and mediates signal transduction by different interleukin receptors. Compound heterozygous JAK3 inactivating mutations cause severe combined immune deficiency. 14

DISCUSSION
This cancer showed primary resistance to NAB-paclitaxel and ddAC chemotherapy concurrent with an anti-PD-L1 agent. The tumor harbored several poor prognostic genomic alterations at diagnosis, including a p53 mutation and coamplification of MYC and MCL1, both of which are implicated in chemotherapy resistance. 15,16 The expression of permeability glycoprotein (P-glycoprotein; ABCB1), a drugefflux transporter that mediates taxane and anthracycline resistance in vitro, also increased during treatment. 17 We observed a measurable increase in intratumor inflammatory response by the end of the treatment (Data Supplement), but unfortunately this did not translate into clinical antitumor activity. The cancer had a persistently high TGFβ expression, and the patient also carried a germline heterozygous deletion of the negative regulator of TGFβ signaling, SMAD6. TGFβ activation in cancers has been linked to primary chemotherapy resistance and also to immune evasion and resistance to PD-L1 blockade in multiple experimental systems. [18][19][20][21] Furthermore, the R175H p53 mutation that this cancer harbored has been shown to render cancers insensitive to the growth inhibitory effects of TGFβ. 22 On the basis of these findings, we hypothesize that the high expression of TGFB2 by this tumor may have contributed to its immune escape and its resistance to chemotherapy. Another notable feature of this cancer was the low level of expression of programmed cell death protein 1 (PD1) at the time of diagnosis and the high level of expression of HLA-G. Low PD-L1 expression has been consistently linked to lesser benefit from ICB, and HLA-G is a self-signal that shields placental cells from immune attack, which mediates immune tolerance in pregnancy. 23 It is impossible to ascertain which of these mechanisms played the dominant role in mediating treatment resistance in this case, or if the multiple anomalies collectively