The role of infiltrating lymphocytes in the neo-adjuvant treatment of women with HER2-positive breast cancer
Written by A. J. Eustace, S. F. Madden, J. Fay, D. M. Collins, E. W. Kay, K. M. Sheehan, S. Furney, B. Moran, A. Fagan, P. G. Morris, A. Teiserskiene, A. D. Hill, L. Grogan, J. M. Walshe, O. Breathnach, C. Power, D. Duke, K. Egan, W. M. Gallagher, N. O’Donovan, J. Crown, S. Toomey & B. T. Hennessy
Introduction:
HER2-positive breast cancer accounts for approximately 20% of all breast cancers and prior to the clinical development of trastuzumab, had the worst outcome of any breast cancer subtype [1]. However, the development of trastuzumab and the subsequent clinical trials which have tested newer HER2-targeted therapies (including lapatinib and pertuzumab) in combination with trastuzumab, have significantly improved the outcomes of women with early-stage HER2-positive breast cancer [1]. Trastuzumab, a humanized monoclonal antibody, is known to have both cytotoxic and immunological effects on tumour cells [1, 2]. In the last decade studies have identified that the localized immune environment plays an important role in determining the outcome of women with non-metastatic HER2-positive breast cancer [3, 4]. In fact studies have shown pre-treatment tumour infiltrating lymphocytes (TILs) [5] and more recently stromal lymphocytes (SLs) [4] have been shown to be independent predictive markers of future pathological complete response (pCR). Whilst many studies have correlated baseline lymphocyte levels with the likelihood of subsequent pCR, very few have studied the impact of HER2-targeted therapy on the immune environment of the tumour itself. In the TCHL clinical trial (NCT01485926), which assessed TCH (docetaxel, carboplatin, and trastuzumab) and TCHL (TCH and lapatinib) in stage II-III HER2-positive breast cancer patients, we obtained core biopsy samples from the primary tumour from consenting patients at pre-treatment and at 20-days post-cycle 1 of trastuzumab based treatment. Using these tumour samples, we conducted TIL analysis and assessed the impact of a single dose of TCH/L chemotherapy treatment on the numbers of infiltrating lymphocytes in breast tumours. For the frst time, our study identifies that immune contexture is significantly modulated in breast
tumours after only 1 cycle of TCH/L chemotherapy, and this may provide clues as to how and why some patients achieve a subsequent pathological complete response (pCR).
Materials and methods
Patient population and samples
TCHL (ICORG10-05) (NCT01485926) is a Phase-II neoadjuvant study run by Cancer Trials Ireland (formerly All Ireland Co-Operative Oncology Research Group (ICORG)) assessing TCH (docetaxel, carboplatin, and trastuzumab) and TCHL (TCH and lapatinib) in stage II-III HER-2-positive breast cancer patients [6]. Full details of the trial are available at www.clinicaltrials.gov. pCR was determined in the TCHL clinical trial by the absence of invasive carcinoma. Of the 88 patients enrolled we were able to obtain lymphocyte information for 68 patients. Of those 68 patients, 20 had a core biopsy taken by an interventional radiologist, 20-days post-cycle 1 of either TCH/TCHL therapy (Ontreatment samples). Samples were snap frozen and stored at −80 °C until required. Full clinicopathological details of patients involved in this study are including in Table 1, and Fig. 1 represents a consort diagram of samples used in the analysis.
Sample processing
Baseline tumour biopsies obtained prior to neo-adjuvant chemotherapy were formalin fxed and parafn embedded (FFPE). Haematoxylin and Eosin (H&E) staining was performed on 3 µM sections of biopsies and assessed for invasive tumour epithelial cellularity by a Histopathologist. Only samples with greater than 10% tumour cellularity were used for further analysis. On-treatment samples were embedded in optical coherence tomography and the samples were cryosectioned. A single 3 µM section was taken for H&E staining and analysis, and the adjacent ten 10 µm sections were cut and stored in a chilled cryovial. Following this, a second 3 µM section was then cut for H&E staining. Cut sections were stored at −80 °C.
Immunohistochemistry (IHC) and TIL counting
H&E staining was performed on a Thermo Shandon Varistain Gemini stainer using Harris haematoxylin (CellPath, RBA-4213-00A) and alcoholic eosin Y (Thermo Scientifc, 6766008) before being cover slipped (Thermo Shandon Consul). 4 µm serial tissue sections were cut using a Leica RM2135 microtome. IHC analysis was carried out on a Bond-III immunostainer (Leica Biosystems, Newcastle, UK). Primary antibodies CD45, Common Leukocyte Antigen (Dako, Clones 2B11+PD7/26, M0701) and Cytokeratin (Dako, Clone AE1/3, M3515) were diluted in Bond Primary Antibody Diluent (Leica, AR9352) at 1/500 and 1/400, respectively. Pre-treatment of samples was carried out on the Bond-III using Bond Epitope Retrieval Solution I (Leica, AR9961) for 30 min (CD45) and Bond Enzyme Pre-treatment solution (Leica, AR9551) for 10 min (AE1/3). Detection and visualisation of stained cells was achieved using the Bond Polymer Refne Detection Kit (Leica, DS9800) with Bond DAB Enhancer (Leica, AR9432). Tissues were counterstained with haematoxylin and cover slipped. Slides were scanned at 40X using a Philips 2.0 scanner, viewed
with Philips Image Management System 2.2 and analysed as per current guidelines [7]. As per the recommendations of the TIL working group [7, 8] which stated that TILs at the invasive edge or intra-tumoural TILs can still be included for research purposes, we proceeded with a research study to assess the impact of TCHL treatment on TILs in HER2-positive breast cancer. To that end, four random areas the size of 1 high power microscope feld (between 100,000 and 100,500uM2) were selected in each case. CD45+cells were counted in each of the four areas. Cytokeratin AE1/3
was used to assess the location of tumour cells relative to the CD45+cells in each of the areas counted. These IHC stains were completed on FFPE baseline biopsy samples (n=68/88) and on fresh frozen (FF) biopsies taken 20-days post-cycle 1 (Day-20) of TCH/TCHL (n=20/88). A lymphocyte was counted as a TIL if it was observed to be in direct contact with an invasive tumour epithelial cell [7]. A stromal lymphocyte (SL) was determined if it was dispersed in the stroma, with no contact between the tumour epithelium and the lymphocyte [7]. Overall Lymphocyte count (OL) was the combined TIL and SL count. TIL analysis was independent of treatment groups. In samples where the tumour had completely regressed following treatment, the number of lymphocytes were assessed by counting four random high power felds. In the instances of no residual tumor in on-treatment biopsy samples, it is important to note that the biopsy samples were small. Whilst we report no residual tumor it maybe that any residual tumor was so scattered and minimal, that it was not captured in the small biopsy.
T‑cell IHC and image analysis
We had previously shown from MCP counter analysis [9] a small subset of TCHL patient samples that increased levels of T-cells were associated with response to TCHL-based therapy [10]. We had sufcient material from 13 patients who had matched pre and on-treatment biopsies to perform T-Cell IHC and image analysis. 3 µm serial tissue sections were cut using a Leica RM2135 microtome. IHC analysis was carried out on a Bond-III immunostainer (Leica Biosystems, Newcastle, UK). Primary antibodies for CD3 (Leica, NCL-L-CD3-565), CD4 (Leica, NCL-CD4-368)
and CD8 (Leica, NCL-CD8-4B11) were diluted in Bond Primary Antibody Diluent (Leica, AR9352) at 1/40, 1/100 and 1/100, respectively. Pre-treatment of samples was carried out on the Bond-III using Bond Epitope Retrieval Solution I (Leica, AR9961) for 20 min (CD3, CD8) and using Bond Epitope Retrieval Solution II (Leica, AR9640) for 20 min (CD4). Detection and visualization of stained cells was achieved using the Bond Polymer Refne Detection Kit (Leica, DS9800) with Bond DAB Enhancer (Leica, AR9432). Tissues were counterstained with haematoxylin and cover slipped. The CD3, CD4 and CD8 stained slides for 13 cases (pre-treatment and on-treatment) were scanned at 40X using a Philips 2.0 scanner and the whole section analysed using the open access image analysis software QuPath [11]. The positive cell detection tool was used to measure the number of positive cells per square millimeter of tissue and compared against the assessment of a Histopathologist. Two comparisons were made using both QuPath and the Histopathologist: Firstly, for each antibody, the number of positive cells in the pre-treatment biopsy was compared
to the post-treatment biopsy and secondly the number of CD4+and CD8+cells were compared between the pretreatment biopsy and the post-treatment biopsy. Due to the large number of positive cells in most samples the pathologist score could not be given as a numerical value but was noted as a comparative statement between the samples being analyzed. The QuPath results for all samples were then compared to the pathologist score to ensure accuracy of the software, and QuPath results were then used for quantitative analysis.
Statistical analysis
The non-parametric Wilcoxon signed-rank test was used to determine if there was a signifcant diference between pathological complete response (pCR) and no-pCR for the three comparison groups (TILs; SLs and overall lymphocytes). The test was paired when comparing baseline and on-treatment groups. The paired test was also used when comparing pre versus On-Treatment CD3+, CD4+and CD8+counts. T-cell markers and tumour content were correlated using the non-parametric Spearman’s rank correlation. Tumour content versus T-cell markers was plotted and loess regression was used to ft a smooth line to illustrate the relationship between the two variables. P-values of less than 0.05 were considered statistically signifcant.
Results
Pre‑treatment TIL levels correlate with a better pCR rate
We determined the number of both SLs and TILs in the baseline pretreatment FFPE tumors of 68/88 patients who were recruited to the TCH/L trial (Fig. 2a, b). Our study demonstrated that patients who achieved a pCR at surgery post-chemotherapy had signifcantly higher numbers of TILs (p=0.05) in their baseline pre-treatment tumour samples, relative to those patients who did not achieve a pCR postchemotherapy (Fig. 2c). We also observed that pre-treatment SL counts may be predictive of a better chance of achieving a pCR post-chemotherapy but did not reach statistical signifcance (p=0.08). While larger studies have shown estrogen receptor status is predictive of pCR, it did not have an impact on rates of pCR in the TCHL study (p=0.2141) [12].
Correlation between on‑treatment lymphocyte counts and pCR
We have previously shown that tumour epithelial cells are undetectable in the day-20 On-treatment biopsies of some patients who go on to achieve pCR at subsequent surgery [13]. Tumour biopsy samples were obtained from 20 patients 20-days after they had undergone cycle 1 of neo-adjuvant chemotherapy treatment (On-treatment samples). Analysis of both SLs and TILs in these On-treatment tumour biopsy samples identifed that, in contrast to the pre-treatment tumour biopsies, SL, TIL and OL counts are not signifcantly diferent between the two groups defned by pCR versus nopCR at subsequent surgery (Fig. 1d). In our pCR group we observed that after 1 cycle of therapy 70% (7/10) of biopsies had no residual tumour remaining (<5% residual tumour). When we compared TIL counts in the pCR group we observed a non-signifcant trend whereby TIL numbers were lower in the biopsies with no residual tumour relative to the remaining biopsies where residual tumour remained (p=0.14). Of the 20 available on-treatment biopsy samples, eight (which includes a sample from the nopCR group) had no residual tumour left in the biopsy after
20-days of starting neo-adjuvant treatment. Upon excluding these cases in which the immune response is possibly already subsiding, we observed that OL (p =9.09× 10–3) counts were signifcantly higher in on-treatment tumour biopsies from patients who subsequently achieved a pCR relative to those who failed to achieve a pCR at subsequent surgery (Fig. 2e). When we stratifed the lymphocyte counts into either TILs or SLs, patients who achieved a subsequent pCR after completing neo-adjuvant chemotherapy treatment had signifcantly higher SL counts (p=9.09× 10–3) in their On-treatment tumour biopsy samples than those patients who did not achieve a subsequent pCR, but this efect was not seen for TILs (p=0.1).
Level of lymphocytes increase with neo‑adjuvant TCHL chemotherapy treatment
Of the 20 fresh frozen On-treatment patient samples available, 16 had matched baseline infltrating lymphocyte information allowing for an analysis of changes in infiltrating lymphocyte levels in paired pre- and on-treatment samples. Examination of these 16 samples (pCR n=9 vs No-pCR n=7), irrespective of residual tumour status, determined that 1 cycle of neo-adjuvant TCHL treatment was associated with changes in levels of infltrating lymphocytes in patient tumours when they were stratifed on the basis of subsequent pCR. There was no consistent significant difference in TIL,
SL or OL levels between baseline and day-20 tumour biopsies when the tumours in the group that attained a pCR at subsequent surgery were analysed (Fig. 3a). However, in patients who did not achieve a subsequent pCR, we observed a trend from the matched baseline to the day-20 On-treatment samples whereby lymphocyte numbers in the tumours signifcantly increased (TILs, p=0.05; SILs, p=0.08; OLs, p=0.05)
Neo‑adjuvant TCHL treatment reduces number of tumour‑related T‑cells
Given the key role of T-cells in regulating the immune response, we performed IHC analysis on 13/16 paired preand on-treatment fresh frozen samples for which we had sufcient material. The T-cell markers CD3 (pan T-cell), CD4 (T helper cells) and CD8 (cytotoxic T-cell) were examined. No distinction was made between stromal and tumour infltrating T-cell populations. When analysing all patients, we observed that levels of CD3+, CD4+or CD8+T-cells did not signifcantly increase or change from the baseline to the On-treatment tumour biopsy samples.
To further analyse the efect of neo-adjuvant treatment on T-cell numbers we observed the changes in CD3 + , CD4+and CD8+T-cells in the matched baseline and day-20 On-treatment tumour biopsy samples of individual patients (Fig. 4). Interestingly, in those patients who achieved a subsequent pCR, we found a decrease in levels of CD4+or CD8+T-cells in 4/5 patients at day-20. However, in those patients who did not achieve a subsequent pCR, only 4/8 patients had a decrease in CD4+T-cells at day-20, whilst 5/8 had a decrease in CD8+T-cells.
A reduction in tumour volume correlates with decreased numbers of CD4+and CD3+T‑cells
As outlined above, neo-adjuvant TCH/L-based treatment results in a reduction of tumour volume after 1 cycle of treatment, and this tumour reduction correlates with a greater chance of a patient achieving a pCR [13]. However, we aimed to further understand if there was a correlation between loss of lymphocytes, either CD3+, CD4+or CD8+T-cells in the day-20 On-treatment tumour biopsy and a reduction in tumour volume in the biopsy (Fig. 5). Using a Spearman rank correlation, we found a signifcant correlation whereby a reduction of CD3+(p=0.04, rho=0.60) and CD4+(p=0.01, rho=0.72) T-cells was associated with decreased residual tumour content post-1 cycle of treatment. We also observed a similar positive trend for CD8+T-cells, but the results did not reach statistical signifcance (p=0.08, rho=0.52). We, however, did not see a positive trend for OLs (p=0.1, rho=0.50).
References available on request