Prognostic significance of crown-like structures to trastuzumab response in patients with primary invasive HER2 + breast carcinoma

Obesity can initiate, promote, and maintain systemic inflammation via metabolic reprogramming of macrophages that encircle adipocytes, termed crown-like structures (CLS). In breast cancer the presence of CLS has been correlated to high body mass index (BMI), larger mammary adipocyte size and postmenopausal status. However, the prognostic significance of CLS in HER2 + breast cancer is still unknown. We investigated the prognostic significance of CLS in a cohort of 69 trastuzumab-naïve and 117 adjuvant trastuzumab-treated patients with primary HER2 + breast cancer. Immunohistochemistry of tumour blocks was performed for CLS and correlated to clinical outcomes. CLS were more commonly found at the adipose-tumour border (B-CLS) (64.8% of patients). The presence of multiple B-CLS was associated with reduced time to metastatic disease (TMD) in trastuzumab treated patients with BMI ≥ 25 kg/m2 but not those with BMI < 25 kg/m2. Phenotypic analysis showed the presence of CD32B + B-CLS was strongly correlated to BMI ≥ 25 kg/m2 and reduced TMD in trastuzumab treated patients. Multivariable analysis suggested that CD32B + B-CLS positive tumours are associated with shorter TMD in trastuzumab-treated patients (HR 4.2 [95%CI, (1.01–17.4). This study indicates adipose-tumour border crown-like structures that are CD32B + potentially represent a biomarker for improved personalisation of treatment in HER2-overexpressed breast cancer patients.

www.nature.com/scientificreports/ hsCRP High sensitivity C-reactive protein TMD Time to metastatic disease Adipose tissue is an important component of the healthy human breast and yet high body mass index (BMI) is associated with increased risk of developing breast cancer in post-menopausal women and with worse survival in all age groups compared to patients with healthy body weight 1 . There is consequently significant interest in understanding the dynamic endocrine and immunological activity of the breast and how high BMI impacts these systems and ultimately influences pathology. Adipose tissue is involved not only in the biosynthesis of important molecular mediators but is also linked to the metastatic potential of breast tumours 2 . It constitutes a dynamic endocrine and immunologically active organ that regulates energy homeostasis, as well as having important immunomodulatory properties 3 . White adipose tissue has evolved as a slowly mobilised energy store in the form of lipids 4 . During weight gain, adipocytes undergo structural changes resulting in the pathological growth of the adipose tissue that leads to adipocyte hypertrophy and adipocyte death 5,6 . The released intracellular contents of the dead adipocytes such as free fatty acids, cytokines and damage-associated molecular patterns, stimulate the recruitment and proliferation of metabolically activated macrophages, that surround and phagocytose the released cellular material, forming so-called crown-like structures (CLS) 7,8 . Crown-like structures induce a proinflammatory environment and constitute a biomarker of white adipose tissue inflammation and correlate with metabolic dysfunction 7,8 . Inflammation can occur concurrently in both breast and abdominal subcutaneous white adipose tissue as shown in women who underwent bilateral mastectomy and immediate autologous 'flap' reconstruction suggesting that CLS are a marker of low-grade systemic inflammation 9 . The presence of CLS is associated with raised levels of serum high sensitivity C-reactive protein (hsCRP), IL-6 and leptin indicating their pro-inflammatory nature 10 . In addition, CLS in obese adipose tissue have a mixed phenotype that is characterised by the concurrent expression of surface markers that describe M1-such as CD11c, and M2-like macrophages such as CD206 or CD163 11 .
In patients with primary breast cancer, CLS found at any location (Any-CLS; distant adipose tissue, within the adipose-tumour border or intratumoural) are present in 36-50% of patients and have been associated with clinical and body composition parameters 10,12 . Specifically, the presence of Any-CLS has been correlated to high BMI, larger mammary adipocyte size and postmenopausal status 10 . Although there is no clear evidence of association between the presence of CLS identified using the pan-macrophage marker CD68 and clinicopathological features 10,12,13 , the presence of Any-CLS is correlated to impaired relapsed free survival and overall survival in women with early breast cancer 10,13,14 . Moreover, CLS identified using the tumour associated macrophage marker CD163 have been correlated to triple negative and HER2 + breast cancer subtypes 13 .
Between 10-25% of patients with early breast cancer show overexpression of the HER2 protein which has been associated with a high risk of relapse in localised disease and poor overall survival in metastatic disease [15][16][17] . While the introduction of anti-HER2 monoclonal antibodies such as trastuzumab has significantly improved clinical outcomes in both early and metastatic breast cancer 18,19 , resistance to trastuzumab treatment, both primary and acquired, remains a significant clinical challenge 20 . Multiple potential mechanisms of action by which trastuzumab targets HER2 + breast cancers have been proposed 21 . These include its binding to the juxtamembrane portion of the extracellular domain of HER2 receptor 21 thus preventing the cleavage of the HER2 receptor and the formation of a signalling remnant p95. This consequently reduces the phosphorylation of p95 and blocks signal transduction 21 . Trastuzumab may also inhibit the HER2 signalling pathway by physically blocking either homodimerization or heterodimerization 20 . A third mechanism of action is the recruitment via the antibody Fc of innate immune effectors, leading to tumour-cell killing 21,22 . Additional mechanisms such as HER2 receptor degradation via endocytosis have been proposed.
Supporting a role for antibody effector function in the efficacy of trastuzumab, mice lacking the inhibitory CD32B (FcγRIIB) demonstrated enhanced trastuzumab-dependent cell-mediated cytotoxicity whereas mice deficient in activating FcγR were unable to inhibit tumour growth in mouse xenograft breast tumour model 23 .
In mouse xenograft tumour models the increased infiltration of macrophages in the tumour was associated with improved trastuzumab efficacy whereas depletion of macrophages led to reduced antitumor activity 22 . In addition, increased expression of human CD16a (FcγRIII) or mouse FcγRIV by macrophages activated with interferon-γ, induced trastuzumab-mediated antibody-dependent cellular phagocytosis of tumour cells 22 . A recent study that included 40 patients with HER2 + metastatic breast cancer who were treated with trastuzumab, showed that increased numbers of iNOS + M1-like macrophages and CD8 + T-cells within the tumour was associated with improved survival, indicating that responses to trastuzumab can be affected by the immunological status of the tumour 24 .
Adiposity has been shown to affect the local immune environment of solid tumours in both pre-clinical and clinical studies where several studies suggest that obesity potentially enhances responses to immune checkpoint inhibitors 25,26 . Nevertheless, the evidence for the role of adiposity in therapeutic response in HER2 + breast cancer is conflicting which may be attributed to the heterogeneity of breast tumours. In the metastatic setting, a multicentre retrospective cohort study showed that BMI was not associated with clinical outcomes in patients with metastatic HER2 + breast cancer 27 . In contrast, Krasniqi E et al., showed that BMI ≥ 30 kg/m 2 was correlated with poor overall survival in patients with HER2 + metastatic breast cancer treated with pertuzumab and/or trastuzumab emtansine 28  www.nature.com/scientificreports/ not significantly differ between the different BMI groups 30 . Together, these findings indicate that adiposity and hormonal profile may explain the differential responses to anti-HER2 agents in patients with HER2 + primary breast cancer. The spatial distribution of immune and tumour cells has been correlated to clinical outcomes in solid tumours. In patients with adenocarcinoma of the lung, immune geospatial variability was associated with clinical outcomes 31 . Similarly, in patients with triple negative breast cancer, multiplexed imaging of immune cells and tumour markers revealed that tumour expression and immune composition were interrelated and associated with overall survival 32 . The geospatial distribution of immune cell composition was also investigated in patients with HER2 + breast cancer treated with neoadjuvant HER2-targeted therapy 33 . This study demonstrated the importance of spatial proteomic profiling of tumour-immune microenvironment in capturing regional tumour heterogeneity and treatment-associated changes. Currently, there is no clear evidence on the effect of the spatial distribution of CLS on therapeutic responses in patients with HER2 + breast cancer.
The prognostic significance of CLS and consequently of white adipose tissue inflammation in patients with HER2 + primary breast cancer is largely unknown. In the current exploratory study, we investigated the phenotype and spatial distribution of CD68 + CLS in a small cohort of 188 HER2 + breast cancer patients and correlate these phenotypes to clinical outcomes. We showed that the presence of CD32B + CLS at the adipose-tumour border are associated with shorter time to metastatic disease in patients with HER2-overexpressed breast tumours who were treated with adjuvant trastuzumab. Thus, CD32B + adipose-tumour border CLS could potentially be used as a biomarker to optimise patient stratification and personalisation of treatment in HER2-overexpressed breast cancer patients.

Materials and methods
Study population. This study was conducted in a retrospective series of 188 patients with HER2 + invasive breast carcinomas who were diagnosed with breast cancer between 1982 to 2012 at the University Hospital Southampton NHS Foundation Trust, UK. Inclusion criteria were histological confirmation from clinical pathology reports of HER2 positivity (immunohistochemistry score 3 + or immunohistochemistry score 2 + with positive in-situ hybridisation), and surgical excision of the tumour at University Hospital Southampton NHS Foundation Trust. This cohort includes a HER2 + trastuzumab naïve group of 69 unselected primary operable (stage I-III) HER2 + invasive breast carcinomas presenting between 1982 and 2004 and a HER2 + adjuvant trastuzumab group consisting of 117 primary operable HER2 + breast tumours from patients diagnosed between 2005 and 2012. All patients were treated in a single institution with surgery (mastectomy or wide local excision and axillary surgery), followed by adjuvant chemotherapy and/or radiotherapy as per standard surgical and oncological protocols. Adjuvant trastuzumab was given intravenously with a loading dose of 8 mg/kg and maintenance dose of 6 mg/kg every 3 weeks. The planned duration of adjuvant trastuzumab was 12 months. Clinicopathological parameters for both case series are summarised in Supplementary Table 1. Patients with BMI < 18.5 kg/m 2 or de novo metastatic disease or patients who were treated with neoadjuvant chemotherapy were excluded from the study.
Survival data, including time to metastatic disease and development of loco-regional and distant metastases were extracted from clinical records. Time to metastatic disease was defined as the number of months from first operation to the occurrence of distant recurrence. Breast cancer distant disease-free survival was censored if the patient was still alive and progression-free at the time of analysis, lost to follow-up, or died by other causes.
Ethical approval was obtained from South Central -Hampshire B Research Ethics Committee (10/H0504/73). Tumour Marker Prognostic Studies (REMARK) criteria informed the reporting of this study.
Immunohistochemistry. Immunohistochemistry for CD16, CD32B and CD68 were performed on 4 µm sections of formalin-fixed paraffin-embedded breast tumours using BOND MAX Fully Automated Research Stainer, Leica Microsystems, U.K. using BOND reagents according to the manufacturer's instructions as previously reported 34 . Sections were deparaffinized, pre-treated for heat-induced Ag retrieval (BOND ER1/ER2 protocol) and incubated with hydrogen peroxide followed by primary antibody. Mouse or rabbit Abs were subsequently bound to the Post Primary IgG linker reagent or the Poly-HRP IgG reagent, respectively, before incubation with 3,3′-diaminobenzidine (DAB). The sections were subsequently incubated with the other primary antibody, which was then bound to either the Post Primary IgG linker reagent or the Poly-HRP IgG reagent. The substrate chromogen Fast Red was applied and sections counterstained using haematoxylin and mounted in CV Ultra mounting media (Leica Microsystems).
CLS quantification method. CLS were defined as a structure consisting of an adipocyte encircled by CD68 + macrophages by ≥ 50%. The presence and number of CD68 + , CD16 + CD68 + and CD32B + CD68 + CLS were evaluated manually in full face sections using a Nikon Eclipse 80i microscope (Nikon, Tokyo, Japan) by two independent scorers (CB and AL) who were blinded to patients' clinicopathological data and outcomes. CLS were defined as CLS within distant adipose tissue (D-CLS) if the CLS were located > 1 mm away from the adipose tissue-tumour interface into the adipose tissue, CLS within the adipose-tumour border (B-CLS) if the CLS were located ≤ 1 mm away from the adipose tissue-tumour interface into the adipose tissue and intratumoural CLS (T-CLS) if the CLS were located in the tumour/stromal tissue and not in the adipose tissue. Any CLS was www.nature.com/scientificreports/ determined if a patient had a CLS in at least one compartment. The cut-off of CLS ≤ 1 per full-face section was used to differentiate low from high CLS.
All available data were analysed to increase power and minimise the introduction of selection bias. Baseline characteristics of both cohorts by detection of CLS at the adipose-tumour border were summarised by mean and standard deviation for continuous variables and frequencies and percentages for categorical variables. Where appropriate, Pearson's Chi-square, Fisher's exact, Student's t and ANOVA one-way tests were used. For multiple comparisons, p values were adjusted according to Benjamini and Hochberg multiple p value adjustment method 35 . Cumulative survival probabilities were estimated using the Kaplan-Meier method, and differences between survival rates were tested for significance using the log-rank test. Univariate analysis was performed and where variance was significant, this was carried forward to multivariable analysis. Multivariable analysis for time to metastatic disease against covariates CD32B + B-CLS and white cell count was performed using the Cox proportional hazard model. Potential confounders were selected for inclusion in multivariable model based on known associations in the literature. Hazard ratios (HR) and 95% confidence intervals (95% CI) were estimated for each variable. All tests were two-sided with a 95% CI and a p value < 0.05 considered significant.
Ethical approval. All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Ethical approval was obtained from South Central -Hampshire B Research Ethics Committee (10/H0504/73). All data were anonymised, and all the personal identifiers were removed. Informed consent was obtained for research use of tissue surplus to diagnostic requirements as part of the routine surgical consent form.  Tables 3 and 4

respectively).
Frequency and spatial distribution of CLS by BMI. We evaluated the presence of CLS in the adipose (D-CLS), border (B-CLS) and tumour (T-CLS) as shown in representative images in Fig. 1a. In the whole cohort, at least one CLS per full-face section (Any-CLS) was detected either within the distant adipose tissue, the adipose-tissue border, or the tumour in 129 (73.3%) tumour sections (Fig. 1b). CLS were more commonly found in the adipose-tumour border (64.8%) rather than in the distant adipose tissue (38.6%) or intratumorally (15.9%) (Fig. 1c). The presence of Any-CLS was significantly associated with BMI ≥ 25 kg/m 2 (p = 0.01) (Fig. 1d). We then assessed the spatial distribution of CLS and found that detection of D-CLS was associated with BMI ≥ 25 kg/ m 2 (p = 0.005) (Fig. 1e). Although there was a trend for positive correlation between individual or multiple B-CLS and BMI ≥ 25 kg/m 2 this was not statistically significant (p = 0.103). A similar trend was observed for T-CLS which was also non-significant (p = 0.085). To control for bias in scoring, we also evaluated the correlation between the proportion of adipose tissue per tissue section and BMI, which showed no evidence of association (Fig. 1f). Finally, we assessed the BMI distribution across the whole cohort ( Fig. 1g) which demonstrated that the BMI distribution of this cohort is representative of the general population 36 .
Association between CLS phenotype and BMI. CLS are characterised by a mixed macrophage phenotype with CD16 expression being predominantly expressed 37 . For this reason, we assessed the expression of CD16 (FcγRIII), an activating Fcγ receptor, and CD32B (FcγRIIB), the inhibitory Fcγ receptor, by CLS macrophages in the different tissue compartments. Figure 2a shows representative images of CD16 + and CD32B + CLS from the distant adipose tissue and adipose-tumour border. Evaluation of the phenotype of CLS showed no evidence of association between the expression of CD16 in CLS and BMI in the distant adipose tissue, the adipose-tumour border or in the tumour when CLS were present in those locations (Fig. 2b). Although the expression of CD32B in D-CLS and T-CLS were not correlated with BMI ≥ 25 kg/m 2 , there was a positive association between CD32B expression in B-CLS and BMI ≥ 25 kg/m 2 (p = 0.015) (Fig. 2c). Subsequently, we assessed the co-expression of CD16 and CD32B in CLS in distant adipose tissue, adipose-tumour border, and tumour. This demonstrated that double positive expression of CD16 and CD32B in B-CLS was positively associated with BMI ≥ 25 kg/m 2 (p = 0.002) (Fig. 2d). Whilst we had found no association between the presence of B-CLS and BMI in the whole cohort (Fig. 1e), consistent with this BMI-associated change in the CD32B status of the B-CLS, there was a positive association between the presence of CD32B positive B-CLS and increased BMI ≥ 25 kg/m 2 (p = 0.001) (Fig. 2e).
Prognostic significance of CLS on time to metastatic disease. Next, we investigated the potential prognostic significance of CLS stratified by trastuzumab treatment status. Any-CLS and D-CLS were not associated with time to metastatic disease, neither in the trastuzumab naïve, nor in the trastuzumab treated patients (Fig. 3a- www.nature.com/scientificreports/ naïve patients. Conversely in the trastuzumab treated patients there was a significant reduction in time to metastatic disease in those with more than one adipose-tumour border CLS (B-CLS > 1) (p = 0.004) (Fig. 3e,f).   (Fig. 3f). T-CLS > 1 was significantly associated with shorter time to metastatic disease (p = 0.008) in the trastuzumab naïve patients whereas there was a non-significant trend for this association in those that received trastuzumab treatment (p = 0.062) (Fig. 3g,h). However, these data are based on a relatively small number of patients with T-CLS present, n = 6 and 14 for T-CLS > 1 in trastuzumab-naïve and trastuzumab-treated patients respectively. www.nature.com/scientificreports/ Overall though, these findings indicate that the spatial distribution of CLS may determine their prognostic impact.
Correlations between B-CLS and clinical parameters. As shown above, CLS were frequently detected at the adipose-tumour border compared to other tissue compartments (Fig. 1c) and B-CLS were correlated with clinical outcomes in the trastuzumab treated patients (Fig. 3f). In addition, CD32B + expression in B-CLS was correlated to BMI ≥ 25 kg/m 2 (Fig. 2e). For this reason, the subsequent analysis focused on B-CLS. Patient characteristics stratified by B-CLS status in trastuzumab-naive and trastuzumab-treated cohort arms are shown in Supplementary Tables 5 and 6. In the trastuzumab-naive arm, B-CLS ≤ 1 and B-CLS > 1 were present in 31 (47%) and 35 (53%), respectively. In the trastuzumab-treated arm, B-CLS ≤ 1 were identified in 47 (43.5%) of the tumours whereas B-CLS > 1 were found in 61 (56.5%). There was no evidence of association between the presence of B-CLS and clinicopathological characteristics in either cohort arms (adjusted p > 0.05). The mean duration of adjuvant trastuzumab was 10.8 and 10.2 months in the B-CLS ≤ 1 group and B-CLS > 1 group, respectively. The presence of multiple B-CLS was associated with a higher peripheral blood white cell count both in the trastuzumab-naïve (p = 0.04) and the trastuzumab-treated patients (p = 0.06). However, this became non-significant after controlling for multiple comparisons.

Prognostic significance of B-CLS by BMI.
In both the trastuzumab-naïve and the trastuzumab-treated cohort arms, BMI was not associated with time to metastatic disease (Fig. 4a,b). Stratified analysis of trastuzumab-treated patients by BMI showed that BMI may have an impact on the prognostic effect of B-CLS. Specifically, there was no statistically significant difference between B-CLS status and time to metastatic disease in patients with BMI < 25 kg/m 2 (p = 0.137) (Fig. 4c). However, in the BMI ≥ 25 kg/m 2 group, patients with B-CLS > 1 had shorter time to metastatic disease compared to patients with B-CLS ≤ 1 (p = 0.01) (Fig. 4d).

Prognostic significance of B-CLS in trastuzumab-treated patients by Hormone Receptor status.
The association of BMI and B-CLS with clinical outcomes in the trastuzumab-treated patients was stratified by hormone receptor status. As demonstrated in Supplementary Fig. 1a, Fig. 1c), there was evidence of association between the presence of B-CLS > 1 and reduced time to metastatic disease in the HER2 + ER + cohort (p = 0.019) ( Supplementary  Fig. 1d). In patients with BMI ≥ 25 kg/m 2 , the presence of B-CLS > 1 in HER2 + ER− breast tumours was associated with shorter time to metastatic disease (p = 0.031) (Supplementary Fig. 1e) whereas there was a trend for a possible association in the HER2 + ER + patients (p = 0.097) (Supplementary Fig. 1f). These findings suggest that the prognostic association of B-CLS on clinical outcomes to trastuzumab treatment may be influenced by HER2/ ER co-expression profile and BMI.
Prognostic significance of CD32B + B-CLS by BMI. As previously shown, the presence of CD32B + B-CLS was correlated with BMI ≥ 25 kg/m 2 , thus we wanted to investigate the potential prognostic significance of the presence of CD32B + B-CLS in the tumours. Survival analysis revealed that the presence of CD32 + B-CLS was not associated with reduced time to metastatic disease across the whole cohort (p = 0.143) (Fig. 5a). However, stratification of the cohort by BMI, revealed a significant association between the presence of CD32B + B-CLS and reduced time to metastatic disease in patients with BMI ≥ 25 kg/m 2 (p = 0.009) but not with BMI < 25 kg/m 2 (p = 0.779) (Fig. 5b,c). Patients with BMI ≥ 25 kg/m 2 were further stratified on trastuzumab treatment status. This showed no evidence of association between the presence of CD32B + B-CLS and time to metastatic disease in trastuzumab naïve patients (p = 0.449) (Fig. 5d). In the trastuzumab-treated group though,  Table 7).

Discussion
Adipose tissue is infiltrated by significant numbers of macrophages via two main mechanisms: either they can differentiate from bone-marrow-derived monocytes that migrate to the adipose tissue by diapedesis from the systemic circulation, or they can transdifferentiate from pre-adipocytes and mesenchymal stem cells within the adipose tissue 38 . Phenotypic differences exist in adipose tissue macrophages between those found in lean and obese adipose tissue. Differing immune phenotypes are driven by a range of activating stimuli such as metabolic ligands including free fatty acids, high-density lipoproteins, and glucose; along with cytokines and pattern recognition receptor ligands 37,39 . In obesity, adipose tissue macrophages are metabolically activated, and their main function is to clear necrotic adipocytes via either phagocytosis or lysosomal activation 37,40 . Free fatty acids released from ruptured adipocytes can initiate the intracellular NF-κB-mediated signalling pathway via the TLR4 receptor, leading to the activation of innate immune cells in the presence of interferon gamma 41,42 . Proinflammatory macrophages are recruited around necrotic or pyroptotic hypertrophic adipocytes forming CLS that secrete proinflammatory mediators such as TNF-α, IL-6, IL-1b, MCP-1, MIF and nitric oxide (NO). Previous studies demonstrated varying degrees of association between the presence of CLS and clinical outcomes in breast cancer 10,13,14 . The study population of these cohorts was predominantly patients with ER + breast cancer. Hence, the prognostic significance of CLS, and consequently of white adipose tissue inflammation, in patients with primary HER2 + breast cancer has not been previously assessed. The formation of B-CLS are not unique to HER2 + breast cancer and have also been observed in other subtypes including luminal A, luminal B and TNBC 13 . The mechanism of CLS formation has been extensively reviewed elsewhere 7,8 , thus in this exploratory study, we focused our investigation into the phenotype and prognostic role of CLS in 188 patients with HER2 + primary breast cancer. To our knowledge, this is the first study that reports the role of CLS on therapeutic responses in patients with HER2 + breast cancer. CLS were more frequently detected at the adipose-tumour border (Fig. 1c) and were significantly correlated to BMI ≥ 25 (Fig. 1d,e) which is consistent with previous studies 9, 10,12,13,[43][44][45] . Adipose tissue macrophages constitute a diverse population that is characterised by a mixed inflammatory phenotype 37 . In high-fat diet mice, obesity was shown to switch the polarisation of the adipose tissue macrophages from an M2-to M1-like phenotype 46 . However, human CLS have a mixed phenotype that is characterised by the concurrent expression of surface markers that describe M1-and M2-like macrophages 11 . CD16 (FcγRIII) is an activating receptor while CD32B (FcγRIIB) is inhibitory, and these tend to be negatively associated whereby in inflammatory M1-like macrophages CD16 is high while CD32B will be low, and visa-versa in M2-like macrophages. Here, we showed that distant adipose-CLS are characterised by being double positive for both CD16 and CD32B, independently of BMI (Fig. 2d). In patients with BMI < 25 kg/ m 2 , B-CLS tended to be CD16+/CD32B−. However, in patients with BMI ≥ 25 kg/m 2 , B-CLS were predominantly characterised by a CD16 + CD32B + phenotype (Fig. 2d) suggesting that the higher level of adiposity is potentially driving metabolic dysfunction within CLS at the adipose tumour border. In overweight and obese patients, the www.nature.com/scientificreports/ presence of CD32B + B-CLS was strongly associated with shorter disease-free survival compared to patients with normal BMI (Fig. 5). Subgroup analysis showed that the presence of CD32B + B-CLS was associated with worse disease-free survival in trastuzumab-treated patients with BMI ≥ 25 kg/m 2 . Multivariable analysis revealed that the presence of CD32B + B-CLS is an independent prognostic factor for metastatic disease (Table 1). Previous studies have shown varying degrees of association between the presence of CLS and poor clinical outcomes in patients with breast cancer 10,13,14 . This discrepancy in the clinical outcomes in the different cohorts can be attributed to patient and tumour heterogeneity, small sample size and the different quantification methods and outcome assessments. Specifically, these cohorts included patients of any hormonal status and evaluated the presence of CLS within the whole observed adipose tissue area. Geospatial proteomic and transcriptomic profiling of the immune-tumour microenvironment has been associated with clinical outcomes in patients with breast cancer 32,33 . Keren L et. al., showed that the tumour-immune border is a unique site of immune inhibition with altered expression profiles by both tumour and immune cells in patients with triple-negative breast cancer 32 . Hence, the spatial study of immune cell phenotypes and composition as well as their interactions with tumour cells is important in detecting tumour heterogeneity in relation to therapeutic responses. Despite this, to our knowledge no studies have so far directly assessed how the spatial proximity of breast CLS to the adipose/ tumour border may impact prognosis and response to therapy 8 . For this reason, in this study, we set out to investigate the significance of spatial distribution of CLS on clinical outcomes. The spatial distribution of CLS were classified as either being located within the distant adipose tissue (D-CLS); at the adipose-tumour border (B-CLS); or intratumoural CLS (T-CLS) if the CLS were located in the tumour/ stromal tissue and not in the adipose tissue. This investigation has highlighted the potential impact of spatial distribution of CLS within the distant adipose tissue versus the adipose-tumour border revealing clinically important differences compared to these previous studies. Thus, we hypothesise that the presence of CD32B + B-CLS induces, or are at least indicative of, a chronic inflammatory environment that is associated with resistance to anti-HER2 therapy in patients with BMI ≥ 25 kg/m 2 . Due to the multi-modal nature of the mechanism of action of trastuzumab, either by direct cell targeting or engagement with effector cell function 21,47 , the biological mechanism underlying this association between CD32B + B-CLS and trastuzumab resistance is currently unclear and requires further investigation. We can speculate that this may be via a direct effect on the immune system or instead could be a marker of the overall inflammatory state of the tumour microenvironment. Future investigations including the analysis of other immune cell populations such as T-infiltrating lymphocytes, may help to further elucidate the immune mechanism involved.
Epidemiological studies have revealed that the presence of CLS in breast tumours is associated with raised inflammatory markers. Specifically, these studies showed strong evidence of association between high levels of high sensitivity CRP and IL-6 with the presence of CLS in patients with breast cancer 10,44 . The effect of BMI in peripheral inflammation was shown in a cohort of breast cancer patients with healthy BMI where these correlations were either weak or non-significant 43 . These findings indicate that peripheral inflammation may be prominent in patients with high BMI and CLS which in turn may potentially play a critical role in therapeutic responses.
There is also strong evidence of correlation between elevated levels of serum leptin and adiponectin with the presence of CLS 10 . Preclinical studies revealed that leptin activates RAS-dependent MAPK pathway and upregulates the levels of chaperone heat-shock protein 90 (Hsp90) by inducing JAK2/STAT3 activation 48 . These lead to an enhanced HER2 receptor overexpression that reduces sensitivity of breast cancer cells to anti-oestrogen tamoxifen treatment 48 . Leptin has also been implicated in trastuzumab resistance whereby differentiated adipocytes negatively affected trastuzumab-induced growth inhibition of HER2 + BT474 and SKBR3 cell lines, an effect that was mediated via AKT phosphorylation 49 . Inhibition of PI3K suppressed the ability of differentiated adipocytes to promote resistance to trastuzumab 10,49 . Leptin also induces overexpression of HER2 receptor and cell proliferation via the phosphorylation of both epidermal growth factor receptor and Janus-activated kinase 2 (JAK2) in SKBR3 cells 50 .
Together with our data, these findings suggest that a possible underlying mechanism is that adiposity in patients with high BMI induces white adipose tissue inflammation, increasing adipokine expression and metabolic dysfunction. This may be mediated via a paracrine interaction between macrophages, adipocytes and tumour cells leading to the secretion of pro-inflammatory cytokines and adipokines in the breast tissue in obese or overweight patients. This in turn may induce trastuzumab resistance via the activation of MAPK or PI3K signalling pathways. Our observation that this association is driven by CD32B positivity of B-CLS rather than the number of B-CLS alone in high BMI patients may thus potentially act as a novel biomarker allowing for stratification of patients to determine more effective therapeutic strategies. For example, high BMI patients identified to have CD32B + B-CLS are likely to have a reduced response to trastuzumab therapy. Thus, we hypothesise that this subset of patients may benefit from more intensive anti-HER2 therapy such as dual antibody therapy using trastuzumab and pertuzumab in combination. Alternatively, if CD32B + B-CLS are a marker of an immune suppressive tumour environment, then the trastuzumab drug conjugate ado-trastuzumab emtansine (Kadcyla) may be more effective due to its cytotoxic mechanism of action. Conversely, this study highlights how effective trastuzumab treatment is in patients that do not have CD32B + B-CLS, and so these patients could benefit from deescalation of anti-HER2 therapy thereby minimising potential side-effects. However, it is important to highlight the exploratory nature of this present study and that the results are based on a relatively small number of patients. Thus, further validation studies using additional cohorts of patients with primary HER2 + breast cancer treated with trastuzumab is required to confirm these findings. Also, the median follow-up of 36.6 months in our cohort does limit the results' interpretation and so future studies should thus aim to include longer follow-up times.
In conclusion, we provide evidence which indicates that the presence of B-CLS correlates with clinical outcomes and therapeutic responses in patients with HER2-overexpressed breast cancer. Also, CD32B positivity of B-CLS may represent a predictive biomarker which could potentially be used to optimise the stratification and personalisation of treatment in HER2-overexpressed breast cancer patients.