Plasma cell infiltration and treatment effect in breast cancer patients treated with neoadjuvant chemotherapy

This study included 148 patients with invasive breast cancer who received NAC and underwent curative surgery at Juntendo University Nerima Hospital between 2006 and 2010. Of these, we retrospectively examined 146 patients whose clinical records and tumour samples were available. Clinicopathological features of the 146 patients are shown in Additional File 1. Mean age was 54.1 years. Subtype distributed as follows: luminal HER2-negative 53%, luminal HER2-positive 9%, HER2 type 19%, and TN 19%. NAC regimens: 132 (90.4%) patients received four cycles of CEF (C: cyclophosphamide 500 mg/m², E: epirubicin 75 or 100 mg/m², F: fluorouracil [5-FU] 500 mg/m²), followed by taxane (12 weeks of paclitaxel: 80 mg/m² or four cycles of docetaxel: 75 mg/m²) prior to surgery; five (3.4%) patients received only CEF; and nine (6.2%) patients were given only taxane, due to conditions such as cardiac dysfunction. For patients with human epidermal growth factor receptor 2 (HER2)-positive tumours, trastuzumab was also administered simultaneously with taxane. Following surgery, HR-positive patients received adjuvant endocrine treatment. To test the association of PCs with pCR in HR-negative cases, another cohort of HR-negative tumours (n = 71), treated at Juntendo University Hospital, was also employed (see details in the Results section). This study was carried out with approvals from the ethics committee of Juntendo University Nerima Hospital (no.2020035) and Juntendo University Hospital (no.19–182). Patients could see the research plan on the website of the hospitals and were offered the choice to opt out of the study at any time.

Pathological examinations were carried out by two pathologists at our hospital. Tumour grade was judged based on the Nottingham Histologic Score system and a grade III tumour was defined as high grade. Chemotherapy effects were determined employing surgical specimens, based on the General Rules for Clinical and Pathological Recording of Breast Cancer (the 18th edition published by the Japanese Breast Cancer Society) [21]. Briefly, grade 0 (no effect): no histological findings of treatment effect are observed, grade 1 (slightly effective): treatment changes in less than two-thirds of the invasive cancer tissue are seen, grade 2 (markedly effective): treatment changes in more than two-thirds of the invasive cancer tissue are seen, grade 3 (pCR): all invasive nests disappeared. In the current study, we defined pCR based only on the primary breast tumour, that is, without lymph node evaluation.

TIL amounts were determined using haematoxylin and eosin-stained tumour biopsy specimens, based on recommendations made by the International TILs Working Group [22]. Briefly, TILs in the stromal compartment (% stromal TILs), using the area of stromal tissue as a denominator, were determined semi-quantitatively. TILs were examined within the borders of the invasive tumour, and average TIL numbers in the tumour area, not focusing on hotspots, were assessed.

IHC was performed on biopsy specimens before NAC. ER and PgR were assessed based on the Allred scoring system [23]. Because the global cut-off value for HR at that time was 10%, HR-positive tumours at our institution were defined as having a total score of 4 or more. Considering that our patients received systemic treatments based on these criteria and recent studies suggest that tumours with less than 10% positivity respond poorly to endocrine therapies as they have a different molecular phenotype [24, 25], we retained our rules in the current study. Therefore, a tumour with less than 10% staining of cancer cell nuclei was considered HR-negative. HER2 was considered positive if the entire cell membrane of more than 10% of tumour cells showed strong staining, or HER2/neu gene amplification was confirmed by fluorescence in situ hybridisation. We used mouse monoclonal anti-Ki67 antibody, clone MIB-1 (Dako, Tokyo, Japan). The Ki67 labelling index was calculated for each biopsy specimen from a hotspot within a high-powered field (× 400). To achieve higher reproducibility, we counted cells with e-Count software (e-path, Kanagawa, Japan), developed for assessing nuclear IHC staining. Briefly, this software automatically counts the total number of positive and negative cells in the field and calculates the positive rate. For CD8 and PC marker CD138, clone C8/144B (Dako) and MI15 (Dako) were used, respectively. Positive immune cells in stromal areas were counted manually within a hotspot in a high-powered field (× 400), since the aforementioned software works only for nuclear staining. The researchers conducting these assessments were blinded to the effect of chemotherapy on the patients.

For immune cell profiling we conducted multiplexed fluorescent immunohistochemistry (mfIHC). We used tyramide signal amplification with an Opal IHC kit (PerkinElmer, Waltham, MA, USA) according to the manufacturer's instructions. The paraffin-embedded block was cut into 4 μm sections. Primary antibodies used were: CD3 (clone SP7, Abcam, Tokyo, Japan), CD4 (clone 4B12, Leica Microsystems, Tokyo, Japan), CD8 (clone 4B11, Leica Microsystems), FoxP3 (D6O8R, Cell Signaling Technology, Danvers, MA, USA), T-bet (clone 4B10, Santa Cruz Biotechnology, Dallas, CA, USA), cytokeratin (clone AE1/AE3, Dako), CD20 (clone L26, Thermo Fisher Scientific, Tokyo, Japan), CD79a (clone JCB117, Dako) CD38 (clone SPC32, Leica Microsystems), PD-1 (clone EH33, Cell Signaling Technology), and PD-L1 (clone E1L3N, Cell Signaling Technology). We constructed two sets of panels, a T cell and a B cell panel. The former comprised CD3, CD4, CD8, Foxp3, and T-bet, while the latter comprised CD20, CD79a, CD38, PD-1, and PD-L1. Both panels also included cytokeratin to differentiate tumour and stromal areas, and DAPI for staining nuclei. For mfIHC, we employed CD38 and a combination of CD20 and CD79a for PC, i.e., PCs were defined as CD79a⁺CD20⁻CD38⁺ cells, since CD138 can be detected on immune cells as well as some cancer cells.

For each case, a whole slide was scanned at × 100 with an automated imaging system (Vectra ver. 3.0, PerkinElmer). They were exposed to the five filters (DAPI, FITC, CY3, TEXAS RED, and CY5) to ensure that each slide was in focus. Phenochart was used to annotate the tumour and stromal fields and whole specimens were captured with an average of 20 areas at × 200 magnification (sized 669 × 500 μm each). An image analysing software program (InForm, PerkinElmer) was used to segment cancer tissue into cancer cell nests (intra-tumoural) and the framework (stromal) region and to detect immune cells with specific phenotypes. Following the manufacturer’s instructions, manual training sessions for tissue segmentation and phenotype recognition using representative images of mfIHC were conducted. Then, automatic machine learning was repeated until the algorithm reached the necessary level of confidence before performing the final evaluation. Representative images of tissue segmentation and cell phenotype recognition are shown in Additional File 2. Infiltrating immune cells were quantified using an analytic software program (Spotfire, TIBCO, Palo Alto, CA, USA) and then calculated per area.

The overall pCR rate was 17% (25 cases). High grade, high Ki67 labelling index, PgR-negative and HER2-positive tumours showed a significantly higher pCR rate (Table 1; p = 0.017, p < 0.001, p = 0.033, and p = 0.001, respectively). Tumours with a high infiltration of TIL and CD138 also had a higher frequency of pCR (p = 0.002 and p = 0.039, respectively). On multivariate analysis, HER2 status and TIL remained as independent factors associated with pCR (p = 0.002 and p = 0.003, respectively).

To focus on the relationship between PC infiltration and chemo-effect, we further examined the distributions of PCs in relation to chemo-effect. Comparing mean values of PC, we observed significantly more PC infiltration in the pCR group (Fig. 1a, p = 0.028). When non-pCR cases were further classified based on chemo-effect, we found PC number was dependent on chemo-effect, where mean values decreased from Grade 3 to 0 (101, 78, 57, and 48, respectively; Fig. 1b). When PC infiltrations were assessed according to HR status, no differences were observed in HR-positive tumours (p = 0.764; Fig. 1c), whereas a trend of correlation between PC infiltration and chemo-effect was observed in HR-negative cases (p = 0.052; Fig. 1d). To test whether this correlation would be observed in another cohort, we used HR-negative tumours (n = 71) from Juntendo University Hospital (JUH). Clinicopathological features of these patients are shown in Additional File 3. In this cohort, the correlation between PC infiltration and chemo-effect was confirmed (p < 0.001), as more PC infiltrations were observed in pCR cases (Fig. 1e).

To elucidate roles of PCs, we selected cases with Grade 3 (pCR; n = 25) and Grade 0 (n = 10) for further examination with mfIHC (indicated by green rectangles in Fig. 1b). On the T cell panel, we observed significantly more CD3⁺CD8⁺ T-lymphocytes in pCR cases, both in cancer and stromal areas (Fig. 2a) (p < 0.001 and p = 0.005, respectively). Representative images of CD3⁺CD8⁺ T cell infiltrations are shown in Fig. 2b. CD3⁺CD4⁺T-bet⁺ T-lymphocytes were also more frequently observed in the cancer area of pCR cases (p = 0.024), as well as CD3⁺CD8⁺FOXP3⁺ T cells (p = 0.012). On the B cell panel, significantly more total B-lymphocytes and PCs were observed in pCR cases, both in cancer and stromal areas (Fig. 3a). Representative images of PC infiltrations by mfIHC are shown in Fig. 3b. Notably, tumours of the pCR group frequently had more infiltration of both PD-1⁺ and PD-L1⁺ B cells and PCs. Significantly higher expression of PD-L1 in cancer cells was also observed in the pCR group (p = 0.005).

During the median 78-month observation period (range: 14–168) for the 146 patients, 44 (30%) developed distant metastases and 34 (23%) died due to breast cancer. To examine the relationships between clinicopathological factors and patient outcomes, a Cox proportional hazard model was constructed. As shown in Table 2, remnant tumour size and lymph node metastasis were independent factors associated with shorter DFS (p = 0.048 and p = 0.007, respectively). The hazard ratio of having high PC infiltration was 0.36, indicating that patients with high PC tumours might survive longer. Therefore, we further focused on the relationship between PC expression and DFS and drew Kaplan–Meier curves. The mean value of PC numbers of 64 was employed as cut-off to distinguish PC-high and -low groups. Interestingly, Kaplan–Meier curves show that patients with HR-negative tumours with high PC infiltration had significantly longer DFS (p = 0.034), a trend not observed in patients with HR-positive tumours (Fig. 4). No statistically significant differences were identified concerning OS (Additional File 4). For reference, Kapan–Meier curves of patient outcomes according to TIL are also shown in Additional File 5.