Evaluation of PD-L1 and tumor infiltrating lymphocytes in paired pretreatment biopsies and post neoadjuvant chemotherapy surgical specimens of breast carcinoma

Herein it was evaluated the impact of PD-L1 immunohistochemical expression and stromal tumor-infiltrating lymphocyte (sTIL) counts in pretreatment needle core biopsy on response to neoadjuvant chemotherapy (NACT) for patients with breast carcinomas (BC). In 127 paired pre- and post-NACT BC specimens, immunohistochemical expression of PD-L1 was evaluated in stroma and in neoplastic cells. In the same samples sTILs were semi-quantified in tumor stroma. Post-NACT specimens were histologically rated as having residual cancer burden (RCB of any degree), or with complete pathological response (pCR). PD-L1 expression and higher sTIL counts were associated with histological grade 3 BC. PD-L1 expression was also associated with the non-luminal-HER2+ and triple negative immunohistochemical profiles of BC. Pathological complete response was associated with histological grade 3 tumors, and with the non-luminal-HER2+ and triple negative profiles. Additionally, our results support an association between PD-L1 expression and pCR to NACT. It was also observed that there is a trend to reduction of sTIL counts in the post-NACT specimens of patients with pCR. Of note, PD-L1 was expressed in half of the hormone receptor positive cases, a finding that might expand the potential use of immune checkpoint inhibitors for BC patients.


Materials and methods
Study design, setting and subjects. The present cohort study was conducted at the Women's Hospital (Hospital da Mulher Professor José Aristodemo Pinotti, Centro de Atenção Integral à Saúde da Mulher-CAISM), State University of Campinas (UNICAMP), São Paulo, Brazil. Women with invasive breast carcinoma who underwent neoadjuvant chemotherapy (NACT) and post-NACT surgery between June 2016 and June 2019 were included. None of them received endocrine therapy in the neoadjuvant scenario. The study was approved by the Research Ethics Committee of Campinas State University (CAAE: 85013718.1.0000.5404). All procedures were carried out according to the Helsinki Declaration and its later amendments. Biological samples were stored at the Biobank number 56 from CAISM. All women signed the informed consent before having their biological samples deposited in the institutional biobank. The diagnosis of breast carcinoma was performed on specimens obtained by percutaneous needle (core) biopsy under ultrasound view. Criteria for indicating NACT included histological subtype, stage and age. NACT followed the institutional clinical protocol, considering the immunohistochemistry based molecular subtype of breast cancer.
The following regimens are indicated as NACT at our hospital: for luminal BC, four cycles of standard chemotherapy with usual anthracycline plus cyclophosphamide, with 21 days intervals, followed by weekly dose of taxane for 12 weeks, the (ACT) regimen 24 . For HER2 positive BC, the same ACT regimen is applied, plus four cycles of trastuzumab every 21 days, during NACT; then, trastuzumab is given as adjuvant therapy to complete 17 cycles 25 . Carboplatin is added to standard NACT for triple negative breast cancer (TNBC), to improve pathological complete response (pCR), although this indication is controversial in the literature 26 . After NACT all women underwent surgical treatment (mastectomy or quadrantectomy with sentinel lymph node biopsy or axillary lymph node dissection), according to the standard institutional protocol. Patients who abandoned treatment and those without available tumor tissue on pre-treatment archival specimens were excluded.
Among 127 women included in the study, three presented bilateral carcinoma. Clinicopathological data, as age, ethnicity, menopausal status, pregnancy, lactation, hormone replacement, family background of breast or ovarian cancer, smoking, clinical tumor stage, and clinical axillary lymph node status were obtained at medical consultation, registered in the medical records and listed on a data collection form. Clinical staging was performed according to the International Federation of Gynecology and Obstetrics system 27 . Response to chemotherapy was assessed in the surgical specimen post-NACT, according to the Residual Cancer Burden (RCB) online calculator. This reproducible and clinically validated approach allowed scoring of response to therapy as: complete pathological response (pCR), RCB-I (minimum residual disease), RCB-II (moderate residual disease), and RCB-III (extensive residual disease). The pathological reports provided the final residual tumor dimensions (in mm), the percentage of cancer cell areas in the residual tumor bed, and the proportion of in situ component. In addition, the number of positive lymph nodes and the diameter (mm) of the largest nodal metastasis had been informed 28 . For statistical analyses, cases were clustered into: (a) cases with complete pathological response (pCR), and (b) cases with residual breast carcinoma (RCB-I to RCB-III).
Histopathology and immunohistochemical evaluation. Tissue samples were collected either by percutaneous needle core biopsy (pre-treatment specimens), or during surgical resection after NACT (posttreatment specimens). All samples were fixed in 10% buffered formalin and paraffin-embedded. Hematoxylin-eosin-stained sections were reviewed to confirm the histological diagnosis in the core needle biopsy and in post-NACT tissues. In every case, histological diagnoses were established according to the World Health Organization (WHO) criteria 29 . Histological non special and special types were included; tumor grade and the presence of vascular invasion were revised. www.nature.com/scientificreports/ Stromal tumor-infiltrating lymphocytes (sTILs) were assessed on hematoxylin-eosin-stained slides from matched pre-treatment core needle biopsies and post-neoadjuvant chemotherapy resection specimens. sTILs were assessed according to the guidelines of the Immuno-Oncology Biomarker Working Group 29 .Although the guidelines for sTILs evaluation have been established for initial diagnostic specimens of tumors, they were also used for residual disease. Stromal TILs were defined as the percentage of tumor stroma area occupied by mononuclear inflammatory cells in the tissue samples. After NACT, in women with residual cancer, sTILs were evaluated in the residual tumor bed, defined as the largest cross-sectional area of residual invasive tumor cells 30 . In women with pCR, sTILs were assessed in the tumor bed. The criteria described by Dieci et al. 30 were used to group sTILs intensity: absent, mild (> 1% to 10%), moderate (11% to 60%), and intense (> 60%). Only sTILs in association to invasive tumors were considered; areas of ductal carcinoma in situ and normal breast lobules were not addressed for sTILs scoring 31 .
Conventional manual immunohistochemical technique was performed, as follows: tissue sections were deparaffinized in xylene and rehydrated in alcohol gradient; antigen retrieval was made in citrate buffer (pH 6.0) in a regular microwave at 95º C for 30 min; endogenous peroxidase activity was blocked with 0.3% hydrogen peroxidase; then, slides were incubated overnight with the primary antibodies at 8° C; the next morning, primary antibodies were rinsed away with phosphate buffer saline, and the biotin-free peroxidase-conjugated polymer system was applied for one hour at room temperature; freshly prepared 3-3′-diaminobenzidine tetrahydrochloride was used to reveal peroxidase bound to immunocomplexes for 25 min; counterstain was achieved using Harris hematoxylin, followed by dehydration, clearing, and mounting. Positive and negative controls were run in each batch.
Evaluation of ER and PR was performed according to Allison et al. 33 , on the diagnostic (pre-treatment) specimens. Cases were considered positive if 1% or more tumor cells were positive. For Ki67, results were represented by the percentage of positive tumor cells in hot spots 34 . Human epidermal growth factor receptor 2 (HER2) staining was scored as 0 + /1 + (negative), 2 + (equivocal), or 3 + (positive). Equivocal (2 +) cases were further submitted to fluorescence in situ hybridization (FISH) to detect gene amplification in a reference Laboratory, according to the recommendations of the American Society of Clinical Oncology/College of American Pathologists (ASCO/CAP) 35 . To address tumor heterogeneity, in cases negative for hormone receptors and/or HER2 in immunohistochemistry performed on the pre-treatment tissue sample, immunohistochemistry was remade on the surgical specimens with residual disease for subtype confirmation 36 .
PD-L1 assessment was visually estimated on both pre-and post-NACT tissue specimens. Membranous and cytoplasmic staining of epithelial tumor cells and stroma cells were evaluated. A case was considered positive if more than 1% of positive cells were present, in one or both components 37,38 . In the presence of residual tumor, comparison of PD-L1 expression between pre-and post-NACT specimens was made in tumor and stromal cells. In patients with pCR, only stromal components were compared between pre and post NACT specimens.
As evaluation of PD-L1 immunostaining is not part of the pathologist's customary routine, it was independently carried out by two pathologists for all 127 matched specimens (LH and KA). After a period of co-observation training, both pathologists estimated the expression of this marker individually. Then, the interobserver agreement was calculated. For discordant cases, a consensus of both pathologists was reached at the coobservation microscope. The same routine was adopted for post-NACT specimens. All other microscopical evaluations were made as standardized in pathology practice. Statistical analysis. Data were analyzed using the R Environment for Statistical Computing Software 39 .
Significance was set at p < 0.05, with 95% confidence intervals (CIs). First, we assessed the relationship between key clinical and pathological features and PD-L1, sTILs, pathological response status, using a bivariate binomial model; odds ratios and p-values for the associations were calculated. Next, we compared the pre-treatment status (biopsy fragment) and post-neoadjuvant chemotherapy, paired with the sTIL counts (evaluating in both cases with residual disease and with complete pathological response to treatment) and PD-L1 expression (evaluating in both cases with expression only in the lymphoplasmocytic infiltrate, and with expression in histiocytes), using the McNemar's paired chi-squared test.
Four scenarios were possible in relation to sTIL counts: (1) Increased, when sTIL counts were low/negative before and moderate/intense after NACT; (2) Positive stable, when sTILs were moderate/intense before and after NACT; (3) Negative stable, when sTILs were low/negative before and after NACT, and (4) Decreased, when sTILs were moderate/intense before NACT and turned out to be low/negative after NACT. Likewise, four scenarios were possible in relation to PD-L1 expression: (1) Increased, when PD-L1 was negative before and positive after NACT; (2) Positive stable, when PD-L1 was positive before and after NACT; (3) Negative stable, when PD-L1 was negative before and after NACT, and (4) Decreased, when PD-L1 was positive before NACT and turned out to be negative after NACT. Then, we fit a logistic regression model for the association of these four scenarios (using "Increased" as reference) with pathological response. Finally, we produced Box plots for
Complete pathological response was achieved in 25 tumors (19.7%), while in 102 there was residual disease (Table 3).
Among the 127 core needle biopsies collected before NACT, 75 (59.0%) expressed PD-L1 either in epithelial or stromal cells. High sTIL counts were present in 55 tumors. None of the clinical characteristics was related to both parameters (Fig. 1, Table 1).
Agreement between the two observers for the evaluation of PD-L1 status in pre-treatment biopsy was reached in 63 positive cases, and in 37 negative cases, resulting in an overall agreement (Cohen's Kappa) of 0.54. For post-NACT specimens, agreement for positive PD-L1 status was reached in 65 cases, and in 36 negative cases, resulting in an overall agreement (Cohen´s Kappa) of 0.55. Discordant cases were mostly represented by those in which the degree of positivity was borderline.
Patients were followed up for at least 26 months, up to 56 months: 94 are alive, disease free; 11 are alive with disease and 22 died of disease. Survival analyses according to PD-L1 expression and sTIL counts in pretreatment core biopsy did not reach significant results (data not shown).

Statistical analyses.
In univariate analysis, we observed correlation between higher histological grade (Nottingham) with PD-L1 expression and high sTIL counts. The variables with significant association with PD-L1 in univariate analysis were selected to enter a multivariate regression model; after multivariate adjustment, triple negative BC remained significantly associated with PD-L1 (Table 2).
Twenty-five tumors (19.7%) of the entire cohort evolved with pCR and 102 (80.3%), with residual disease after NACT. In univariate analysis, high histological grade, and PD-L1 overall expression in core needle specimens were associated with complete response rate. After multivariate adjustment, non-luminal-HER2+ and triple negative cases were significantly associated with pCR (Table 3). www.nature.com/scientificreports/ Stromal TIL counts in the core needle biopsy did not show significant change in relation to the post-NACT sTIL counts in the surgical specimen; this was valid for both, women with or without residual disease (McNemar test for paired sample p = 0.17 and p = 0.20; Table 4). Table 5 shows that in patients with residual tumor, 64% remained with low sTIL counts or decreased sTIL numbers and 36% remained with high sTIL counts or increased. Although not statistically significant, there was a trend to decrease in sTIL counts among patients with pCR (p = 0.07).
Comparing PD-L1 status between core needle biopsy and post-NACT surgical specimens, expression status was significantly maintained: positive cases remained mostly positive and negative cases remained mostly negative. This was valid for both evaluations, considering the expression in histiocytes or disregarding these cells (McNemar test for paired sample, p < 0.01; Table 6).
There was a significant association between stromal and epithelial PD-L1 expression with sTIL counts in pretreatment specimens ( Fig When expression of PD-L1 is considered only on epithelial cells and when it is evaluated in epithelial + stromal components of core needle biopsies, it is associated with pCR (p = 0.02 and p = 0.03, respectively; Table 7). Table 8 shows that there was no significant change of PD-L1 expression on core needle biopsy versus post-NACT surgical specimens, comparing women with residual cancer or pCR. This is valid either considering PD-L1 expression on histiocytes in the total score, or disregarding them.

Discussion
In the present study PD-L1 expression evaluated in initial biopsies correlated with pathological complete response after NACT. Significance was found either when both neoplastic and stromal cells or when just epithelial cells were considered. Similar results were reached in most studies enrolled in a systematic review by Miglietta et al. 40 . Only rare studies associated PD-L1 positive expression with the presence of residual disease after NACT 41 . According to the review by Stovgaard et al. 42 , all six studies addressed for PD-L1 expression in the NACT setting found significant association between pre-treatment PD-L1 expression and higher pCR. In this review, one of the studies showed association of PD-L1 expression in tumor cells and pCR but did not establish this marker as predictive for complete response in individual subtypes of breast carcinoma 43 . Another one found association between PD-L1 expression and pCR only in univariate, but not in multivariate analysis 44 .
Wimberly et al. 16 favored a predictive value of PD-L1 expression for pCR applying a particular method, that is, quantitative immunofluorescence evaluation by image analysis and signal amplification with thyramide. As the values were continuous, the cut point to consider PD-L1 positive was statistically set. Cerbelli et al. 45 showed association of PD-L1 expression in triple negative breast cancers with pCR, when cut off was set at ≥ 25% positive Table 2. Correlation between overall PD-L1 expression and high sTIL counts on diagnostic core needle biopsies and tumor parameters. Values in bold indicate data of statistical significance. *Three women had bilateral breast carcinoma; sTILs: stromal tumor infiltrating lymphocytes; Ref: reference value. In one case immunohistochemistry and FISH for HER2 was inconclusive. In ten cases HER2 was not expressed through immunohistochemistry and or FISH, in percutaneous needle core biopsies (pre-treatment specimens), and was positive in the surgical specimens after NACT (post-treatment specimens). The variables with significant association with PD-L1 in univariate analysis were selected to enter a multivariate regression model; after multivariate adjustment, triple negative remained significantly associated with PD-L1 expression (p < 0.01; OR = 0.12, 95%CI 0.03 to 0.46). # p value calculation using Fisher's exact test and logistic regression. www.nature.com/scientificreports/ tumor cells. This controversy shows that studies on PD-L1 expression in breast cancer patients submitted to NACT are still needed, with special emphasis on methodological standardization of immunohistochemical evaluation and correlation with other interfering clinicopathological features. In this respect, the analyses performed in the present study, associating multiple clinical and pathological parameters with PD-L1 expression, were not homogeneously addressed in previous investigations. According to the systematic review from Stovgaard et al. 42 , PD-L1 expression in breast carcinomas variated between 0 to 83% in all different intrinsic subtypes, and most studies included in the review presented values around 50%, a figure similar to that evidenced in the present study (59%). Comparison of IHC evaluation of PD-L1 in cancer has been problematic, as different studies use diverse anti-PD-L1 antibodies or platforms. To circumvent this issue there are currently three diagnostics automated IHC assays approved by the Food and Drug Administration (FDA), one using the SP142 clone, the other, the 22C3 clone, and the third, the SP263 clone. Not only the anti-PD-L1 antibody and the platform used are different, but also the scoring systems, with three variations: scoring on tumor cells, on stromal cells and a combination of both, tumor and stromal cells 46 . The platform using the SP142 clone is scored by ICa (percentage of tumor area covered by PD-L1 positive immune cells) 47 , and the platform using the 22C3 clone, by CPS (positive tumor or immune cells as percentage of all tumor cells), without possibility of interchange between both scoring systems 48 . For breast cancer, the ICa score has been suggested, but in a recent evaluation of the three scoring systems, Guo et al. 38 advocated the cutoff 1% using the combined score, in tumor and immune cells.
For our study, we performed manual IHC detection of PD-L1 using the ZR3 rabbit monoclonal antibody. The ZR3 clone has been indicated as one of the biosimilar diagnostic antibodies used in clinical practice (Sorokin et al. 32 ; Gonzales-Ericsson et al. 49 ). Scoring was performed by a consensus of two pathologists assessing tumor epithelial cells and stromal immune cells in the tumor microenvironment by visual analysis. The scoring method used in this study for PD-L1 expression is somewhat similar to that of the SP142 assay, using a 1% cutoff, but herein this cutoff was applied to staining of both epithelial and stromal cells in combination 38,48 . Table 3. Correlation between pretreatment tumor parameters and pathological response post neoadjuvant chemotherapy. Values in bold indicate data of statistical significance. *Three women had bilateral breast carcinoma; pCR: pathological complete response; RCB: residual cancer burden; sTILs: stromal tumor infiltrating lymphocytes; Ref: reference value; In one case immunohistochemistry and FISH for HER2 was inconclusive. In ten cases HER2 was not expressed through immunohistochemistry and or FISH, in percutaneous needle core biopsies (pre-treatment specimens), and was positive in the surgical specimens after NACT (post-treatment specimens). Of the 75 positive cases for PD-L1, 35 expressed as much in the epithelium as in the stroma, 5 expressed only in the epithelium and 35 only in the stroma. We next fit a regression model including all variables with p < 0.10 (Grade-Nottingham, molecular subtype, sTILS and PD-L1); after adjustment, high histological grade, non-luminal HER2+ and triple negative carcinomas remained significantly associated with pCR: OR = 3.50, 95%CI = 1. 28  The predominant association of PD-L1 expression with non-luminal breast carcinomas (27 out of 36, or 75%) and high histological grade (51 out of 65, or 78%) found in our patients is in accordance with previous reports 16,40,42,43 . PD-L1 expression is particularly frequent in triple negative breast carcinomas, a subtype that most benefits from immunotherapy. Interestingly, among our cases of hormone positive carcinomas PD-L1 was expressed in about half of the cases (48 out of 94, or 52%). This figure is higher than the frequency reported   42 . This figure suggests that immunotherapy might also be appropriate for luminal and PD-L1 positive breast cancers, a proposition which warrants further studies. Among pathological parameters, complete pathological response (20% of the cases) was associated with grade 3 and absent expression of hormonal receptors, results already reported before 45 . sTIL counts on initial diagnostic biopsy did not influence pathological response, in opposition with previous reports 16,18,50,51 . This discrepancy could be attributed to the different proportion of molecular subtypes of breast carcinoma in our study, or to factors related to patient's biological characteristics. Comparing the initial biopsy with post-NACT specimens in patients with both residual tumor and complete pathological response, we observed a considerable number of cases with pCR which presented a reduction of sTILs after neoadjuvant chemotherapy. This difference, however, did not reach statistical significance (p = 0.07). In a report on 59 paired samples, a significant reduction in sTILs was seen in post-chemotherapy specimens in comparison with baseline samples; PD-L1 expression remained similar in the paired samples 18 . In this study the greatest decrease in sTILs was associated with pCR. In more recent studies, reduction in sTILs was also detected in post-NACT samples 50,51 . This decrease could reflect the removal of antigenic stimuli exerted by neoplastic cells, which would be followed by reduction of lymphocyte infiltrates in patients with pCR. An inverse relation, that is, an increase in sTILs in post-NACT surgical specimens, was observed by others 52,53 , who interpreted their observation as an attractant effect on lymphocytes exerted by chemotherapy. In regard to the maintenance of PD-L1 expression in residual carcinomas, it is possible that such patients could also benefit from inclusion of immune checkpoint inhibitors to conventional therapy 18 .
Our study presents some limitations, as the relatively small number of the different immunohistochemically defined subtypes of breast carcinoma; the evaluation of more patients with hormone receptor expression, HER2 positive and triple negative would allow the analysis of the clinical impact of PD-L1 expression and sTIL counts within each group. Second, the evaluation of sTILs can be made using markers for different lymphoid subtypes, allowing the determination of which could exert impact on clinical outcomes. In the present report, we used exclusively H&E, according to the guidelines of the International Immuno-Oncology Biomarker Working Group 27,28 . Finally, long term follow up could allow the evaluation of the influence of such parameters in disease free and overall survivals. However, the patients were enrolled prospectively and had a standardized treatment and evaluation.

Conclusion
The present study supports the role of PD-L1 immunohistochemical expression evaluated in initial biopsies in predicting pCR after NACT. This is valid when only epithelial cells are considered and when both epithelial and stromal cells are included. The value of sTILs in predicting pCR was not confirmed by our data. However, we could support the notion that, in general, post-NACT specimens present decreased numbers of sTILs in    www.nature.com/scientificreports/ patients with pCR. PD-L1 associated with higher histological grade (Nottingham) at diagnosis, with non-luminal-HER2+ and triple negative BC, and with sTIL counts. Besides, our data also support the association of pathological complete response rate with high histological grade, non-luminal-HER2+ and triple negative subtypes of BC. Of note, the higher number of patients with HR + subtype in our study might indicate a potential use of immunotherapy in this group of women. Finally, the methods used herein are applicable in most routine pathology laboratories, supported by consensus studies on scoring of sTILs and PD-L1 expression.