Introduction

Breast cancer (BC) stands as the most prevalent female cancer, ranking among the leading causes of cancer-related deaths globally1. Additionally, with 2.4–7.3 cases in 100,000 pregnancies, breast cancer is notably prevalent among cancers occurring during pregnancy2,3,4. Previous reports from Western countries indicate growing numbers of pregnancy related breast cancer (PrBC) cases, supporting the hypothesis that delaying the reproductive phase to an older age may contribute to rising incidence2,4.

Previous studies predominantly referred to Pregnancy-associated breast cancer (PABC), defined as BC diagnosed during pregnancy or within the first postpartum year. Given the distinct prognosis and tumor biology, it is strongly recommended to conduct a separate evaluation for each condition5.

At the time of diagnosis, PrBC is often diagnosed at more advanced stages, characterized by larger tumors and nodal involvement6,7,8,9. Nevertheless, overall outcomes align with those of young non-pregnant BC patients when treatment is promptly administered in adherence to the standard of care for young BC patients7,9.

During pregnancy, elevated serum progesterone levels stimulate T-helper type 2 (Th2)-like-cytokine production in lymphocytes through Progesterone Induced Blocking Factor (PIBF), thereby diminishing cell-mediated and enhancing humoral immunity10,11,12. This induction of feto-maternal immune tolerance is primarily regulated by unique set of molecules expressed in the human placenta. Extravillous trophoblasts (EVT) that invade the decidua, display various of these immune checkpoint molecules on their cell surface, including PD-L1 and TIGIT, also found on malignant cells. Moreover, EVT exhibit minimal expression of classical HLA class I or II molecules but abundant non-classical HLA-G11,12. HLA-G diminishes uterine natural killer (NK) -cell cytotoxicity by binding to an Immunoreceptor with a tyrosin-based inhibitory motif (ITIM)11,12,13,14 on their cell surface. It further facilitates the differentiation of naïve T-cells into regulatory T-cells (Tregs), which in turn can suppress the activity of cytotoxic CD8 + T-cells (CTLs)15,16.

Despite this, pregnant women are often believed to be immunocompromised to some extent17. This coupled with delayed initial diagnosis in pregnant women7, may account for commonly observed advanced tumor stages in PrBC.

To characterize descriptively major differences in baseline characteristics as well as immune response markers between pregnant and non-pregnant BC patients, we compared the expression profiles of immune biomarkers in the BCP cohort (BCP registry study; NCT 00196833) with a matched cohort of non-pregnant BC patients.

Results

Comparative analysis—baseline characteristics

Pregnant BC patients were significantly younger than non-pregnant BC patients (median 34 [26–47 years] vs. 37 [27–47 years], P < 0.001). The predominant histologic subtype was ductal or ductal-lobular invasive BC (pregnant: 90.2% vs. non-pregnant: 80.8%) with high grading (G3) (pregnant: 68.0% vs. non-pregnant: 67.2%). At the time of diagnosis, the majority BCP tumors were locally advanced (T2; pregnant: 50.4%) and presented with nodal metastases (N+ pregnant: 53.6%). The BCP cohort had a significantly higher proportion of ER negative tumors (57.6% vs. 41.8%; P = 0.0016) and a significantly elevated expression of Ki-67 (>20%: 53.3% vs 38.1%, P = 0.025).

Categorial baseline characteristics of both cohorts are shown in Table 1.

Table 1 Baseline characteristics (categorial) of non-pregnant vs. pregnant BC patients
Full size table

Expression of immunological markers in both cohorts

The expression of HLA-G demonstrated significant differences in both cohorts, yielding opposing results with each quantification method. While the median HLA-G H-score was higher in the non-pregnant cohort (4.9 vs 10.6, P = 0.012), the median percentage of HLA-G expression was higher in the pregnant cohort (3.7% vs 2.0%, P = 0.007). A weak correlation between the H-scores of HLA class I and HLA-G was observed in the non-pregnant cohort, and a moderate correlation was noted for both markers in the pregnant cohort (Spearman’s ρ = 0.27 and 0.45, respectively). Except for Nectin-4, which exhibited significantly higher expression in both quantification methods for the BCP cohort (median 98.9% vs. 94.8%, P < 0.001 and H-score 156 vs. 113, P < 0.001) compared to non-pregnant cohort, no significant differences were found for TIGIT, PD-L1, the H-score of HLA class I and the number of TILs (Fig. 1; see also exemplary TMA slides in Figs. 2–4).

Fig. 1: Boxplots of Continous biomarkers (HLA, HLA-G, TIGIT, Nectin-4).
figure 1

Boxplots for HLA class I, HLA-G, TIGIT,Nectin-4 in non-pregnant and pregnant cohort.

Full size image
Fig. 2: TMA slides of biomarkers HLA class I, HLA-G, PD-L1.
figure 2

A–C TMA slide example with <10% (A) and > 50% positive (B) membranous staining for HLA class I (clon EMR8-5); TMA slide example with <10% (C) and >50% positive (D) membranous staining for HLA-G (clon 4H84); TMA slide example with TPS <1% (E) and with TPS > 50% positive (F) membranous staining for PD-L1 (clon 22C3). TPS tumor proportion score.

Full size image
Fig. 3: TMA slides of biomarker Nectin-4.
figure 3

Exemplary TMA slide with <10% (A) and >50% (B) positive membranous staining for Nectin-4 (clon EPR15613-68).

Full size image
Fig. 4: TMA slides of biomarker TIGIT.
figure 4

Exemplary TMA slide with <10% (A) and >50% positive cytoplasmatic staining for TIGIT (clon BLR047F) on tumor cells (B) and TILs (C).

Full size image

Given the positive correlation observed between the expression of Nectin-4 and Ki-67 in other studies18,19, we performed a post-hoc analysis and found that in both cohorts, patients with Ki-67 > 20% exhibited a significantly increased Nectin-4 expression (Table 2).

Table 2 Post-hoc analysis Wilcoxon–Mann–Whitney tests for Nectin-4 (%, H-score) between pregnant and non-pregnant patients with a low ( ≤20%) vs. high ( >20%) Ki-67
Full size table

In a further post-hoc analysis, we compared HR- (defined as ER and PR negative) and HR+ (defined as ER and/ or PR positive) subgroups within both cohorts examining their Nectin-4 expression. Within the non-pregnant cohort, the HR− subgroup exhibited a significant overexpression of Nectin-4 in both quantification methods (% and H-score). In contrast, within the pregnant cohort, only the H-score of Nectin-4 was significantly higher in the HR− subgroup (Table 3).

Table 3 Post-hoc analysis Wilcoxon–Mann–Whitney tests for Nectin-4 (%, H-score) between HR- vs. HR+ pregnant and non-pregnant patients
Full size table

Subgroup analysis: pregnant cohort and non-pregnant cohort

Comparisons between pregnant patients with T1/2 vs. T3/4, N0 vs. N+ and 1st vs. 2nd vs. 3rd trimester revealed no significant differences in the expression of HLA class I, HLA-G, PD-L1, TIGIT and Nectin-4. Among non-pregnant patients, those with more advanced tumor stages (T3/4) exhibited a significantly higher Nectin-4 H-Score than patients with T1/2 tumors (Median 129 vs. 107, P = 0.049).

Pregnant patients with TILs ≤25% exhibited a significantly lower expression of HLA class I compared to pregnant patients with TILs >25% (mean 36.4% vs. 80.9%, P = 0.003, H-Score 68.0 vs. 178, P = 0.002). Similar results were found in the non-pregnant cohort, where patients with TILs ≤25% (median 12.9% vs. 73.3%, P = 0.004, H-score: 24.6 vs. 110.0, P = 0.010) demonstrated a significantly reduced expression of HLA class I compared to non-pregnant patients with TILs >25%.

Additionally, the count of PD-L1 negative IC in pregnant BC patients with TILs ≤25% was significantly higher than in pregnant BC patients with TILs >25% (PD-L1 negative IC 84.6% vs. 28.6%, P = 0.004). Corresponding results were found in the non-pregnant cohort, where patients with TILs ≤25% had a significantly higher number of PD-L1 negative IC than patients with TILs >25% (PD-L1 negative IC 89.0% vs. 33.3%, P <0.001).

Significant difference can be demonstrated of combined PD-L1 and PD-L1 IC concerning the ER-status in HER2− pregnant cohort. In TNBC the combined and positive PD-L1 IC value was higher than in ER + /HER2− BC (PD-L1 combined 30.4% vs 10.8%, P >0.016 and PD-L1 IC 32.8% vs 10.8%, P > 0.041; data not shown).

Discussion

The primary objective of the analysis was to compare the expression of HLA class I, HLA-G, PD-L1, TIGIT and Nectin-4 exploratively between pregnant and young, non-pregnant BC patients. Our results revealed a significantly higher expression of Nectin-4 in the pregnant cohort.

Differences in proliferation rates between the cohorts could be partially attributed to the age difference between both cohorts, as higher Ki-67 rates are more common in younger BC patients20,21,22. However, there are also biological explanations. Elevated Ki-67 rates could result from a general increase in ductal proliferation in mammary tissue during pregnancy due to elevated levels of estrogen in preparation for lactation23.

Interestingly, tumors in the pregnant cohort exhibited significantly higher proliferation rates according to Ki-67 rates. High proliferation is associated with faster progression24. In addition to the recognized challenges of detecting malignancies in the pregnant breast, the notably higher proliferation activity observed in the BCP cohort as measured by the expression of Ki-67 expression may further explain why PrBC tends to present at a more advanced stage at diagnosis6,7,8,9. Tumors with high proliferation rates, tend to be more aggressive but also demonstrate a better response to neoadjuvant chemotherapy24, emphasizing the importance of timely treatment for pregnant BC patients.

Another factor could be the high Nectin-4 expression in the pregnant cohort. Nectin-4 overexpression is associated with various adverse conditions in malignancies, including distant metastasis, locally advanced disease, and overall poorer survival18,19. Nectins are Ca2+ -independent immunoglobulin-like cell adhesion molecules that are involved in regulating diverse cell functions and communications25. While Nectin-1, -2 and -3 are widely expressed on healthy adult tissue, Nectin-4 expression is primarily restricted to fetal or placental tissue but is also present in various human cancers26.

Nectin-4 is recognized for its ability to activate the Pi3K/Akt-signaling pathway, thereby enhancing cell proliferation and promoting endothelial-to-mesenchymal transition (EMT)27,28. While, the interactions between Nectin-4 and the Pi3K/Akt-signaling pathway were not specifically investigated this study, their potential connection is intriguing, especially considering their shared expression on trophoblasts. Trophoblastic migration, cell proliferation and invasion are regulated through Pi3K/Akt29. Abnormal placentation has been associated with disrupted Pi3K/Akt-signaling30, and the upregulation of Nectin-4 expression on trophoblasts31. Therefore, it is plausible that Nectin-4 also interacts with Pi3K/Akt-pathway on trophoblasts, potentially influenced by pregnancy-specific factors.

One of these pregnancy-specific factors is the elevated level of steroid hormones. As mentioned earlier, estrogen promotes ductal proliferation in the pregnant breast23. It is known that in the presence of high estrogen levels estrogen receptors (ERs) are typically downregulated32. Consequently, this downregulation could potentially result in an upregulation of estrogen-related receptors (ERR). ERRs are speculated to directly compete with ERs, leading to alterations in other’s functions33. Unlike ERs, ERRs do not directly bind to natural estrogen but exert influence on the transcription of the same estrogen-responsive elements. For instance, they may enhance the expression of Nectin-4 through the Pi3K/Akt pathway34.

Previous studies have highlighted an association between Nectin-4 overexpression and a negative HR status19,26. Given this, along with the notably increased prevalence of ER negative tumors in the pregnant cohort (57.6% vs. 41.8%; P = 0.0016), we were prompted to examine the consistency of our findings with existing literature and performed a post-hoc analysis. It is essential to acknowledge that the results from our post-hoc analysis regarding HR-status and Nectin-4 expression may be subject to distortion. A notable aspect is that 16% of the pregnant patients are ER negative and PR positive, whereas only 1.6% of the non-pregnant patients are ER−/PR + . This imbalance in the ER status between the cohorts raises the possibility that ER negativity might leads to an upregulation of ERRs potentially causing Nectin-4 overexpression. Consequently, the 16% of ER−/PR+ pregnant patients might impact the accuracy of our analysis. This presents a limitation of our study and calls for further investigation.

Within TILs, T-cells are the predominant cell-type, followed by B- and NK-cells35. BC is generally considered to be less immunogenic compared to other tumors such as melanoma36. This notion is supported by our data where 92.6% of the non-pregnant and 86.9% of the pregnant patients had low TILs (≤25%). The upregulation of HLA class I expression improves T-cell recognition37. Accordingly, the subgroup “TILs >25%” in both cohorts exhibited a significantly higher expression of HLA class I. While the overall expression of HLA class I molecules in percentage was significantly higher in the pregnant cohort (27.7% vs. 13.8%, P = 0.029), there were no significant differences in categorial TILs between the two cohorts. A recent study by Sajjadi et al. obtained similar results showing no significant differences regarding the number of TILs in PrBC and early-onset BC (EOBC)20. These findings contradict previous observations of reduced TILs in pregnant BC patients compared to non-pregnant BC patients, suggesting a pregnancy-induced reduction in maternal immunity38. Our data, however, indicates that TILs, although generally low in both cohorts, can vary across biological subtypes. The immune cell infiltration is associated with the biological subtype, typically being highest in TNBC and lowest in HR + /HER2− BC20,39,40. Missing differences in our cohort could therefore be explained by our matching process which, in contrast to the other study38, included “biological subtype” as a matching variable. Nevertheless, it is important to note that these results are limited by the fact that 64 patients in the pregnant cohort have missing data for TILs.

Breast cancer incidence tends to increase with age41, while fertility experiences a decline42. The age group 30–34 is characterized by high birth rates43 and exhibits a relatively increased incidence of BC compared to younger women44. The highest incidence of PrBC was observed in the age group 30–34 years, aligning with previous studies45,46. Despite our matching process, pregnant BC patients were younger than non-pregnant BC patients, presenting a limitation of our study as age at diagnosis influences tumor biology21,22. The correlation between the age group with the highest birth rate and the most cases of PABC was also been noted by Andersson et al.2 It is conceivable that if women continue to postpone childbearing into their late thirties, the incidence of PrBC is likely to increase further. Additionally, our matching process using the Mahalanobis distance accommodates this age difference, pairing two matching partners closest to each other while considering all variables47.

The main limitation of this study results from the 100% nodal positivity observed in the GAIN cohort. It is well-established that the TME varies between primary and metastatic disease48,49. While the majority (53.6%) of the pregnant patients in our study exhibited nodal positivity, the inclusion of node negative non-pregnant patients could have enhanced the validity of our results and potentially influenced the expression patterns of immunomarkers. Another limitation is the lack of information on parity status and period between BC diagnosis and latest pregnancy, if applicable, in the GAIN cohort. In addition, the results of this study are subject to some irregularity as BC risk, tumor biology and thereby tumor aggressiveness may vary. Moreover, these three factors are most unfavorable shortly after birth and only improve in the following postpartum years, so the results obtained here should be viewed with caution5,50. The cohort was selected primarily due to the availability of extensive pre-existing TMAs and associated data from GAIN patients.

For future BC studies an inclusion of parity status and time of last pregnancy in clinical data collection is advisable.

Our results indicate that BC during and outside of pregnancy exhibit similarities but also notable differences. The PrBC appears to manifest with higher proliferation rates. Furthermore, the overexpression of Nectin-4 and its association with disease progression suggest that the typically more advanced stages observed in PrBC might not solely result from delayed diagnosis. Elevated estrogen levels during pregnancy could potentially influence tumor biology and induce alterations in the TME of PrBC, warranting investigation. Despite the immune alterations associated with, our study found no evidence of a diminished anti-tumor immune response in the pregnant cohort.

Nevertheless, as in any malignant disease timely diagnosis and treatment of PrBC is crucial to optimize the outcome for mother and child. Further information on cancer management during pregnancy can be found on https://esgo.org/network/incip/51.

Establishing Nectin-4 as a standard marker in PrBC or any malignancy occurring during pregnancy might improve our understanding of its expression patterns and may represent a potential targeted treatment option to modify standard chemotherapy approaches.

Methods

Patient population

Our pregnant cohort comprised pregnant BC patients enrolled in the Breast Cancer in Pregnancy (BCP) registry study (GBG-29/NCT00196833) by the German Breast Group (GBG)9. The trial was conducted according to the Declaration of Helsinki. Between April 2003 and March 2019, a total of 2831 patients were registered. Out of these, tumor samples with representative and untreated tumor regions were available for 127 early breast cancer patients. The study was approved by the ethics committee of Goethe University Frankfurt (Reference No. 254/02), Germany. Written informed consent was obtained prospective from patients who give their willingness to participate in this analysis.

Formalin-fixed paraffin-embedded (FFPE) tissue samples with untreated tumor material from either core biopsies for patients with neoadjuvant treatment (n = 23) and unknown therapy setting (n = 6) or surgical specimens for patients with adjuvant treatment (n = 97) were used to construct tissue microarrays (TMA). Patients receiving palliative treatment (n = 1) were excluded. The pregnant cohort was matched with a suitable non-pregnant BC cohort with existing TMAs drawn from the GAIN study (n = 1305), which recruited node-positive, high-risk early BC patients randomized to two distinct dose-dense regimens52. Matching criteria included age, grading, biological subtype, tumor and nodal stage. Among them, 125 BCP patients and 522 patients from the GAIN study had available information in all matching variables. Additionally, patients from the GAIN study were excluded if postmenopausal and/or older than >47 years (maximum age in the BCP cohort).

Statistical analysis

The nearest neighbour matching in a 1:1 ratio and by using the Mahalanobis distance without replacement47, was selected as an appropriate method for multivariate cohort matching after a comparison with different other matching methods (propensity score matching, coarsened exact matching (CEM), weighted regression analysis). Although similar results were obtained from other methods, the Mahalanobis distance without replacement proved superior because it allowed the inclusion all 125 pregnant patients and simultaneously provided an acceptable balance of all matching variables between both cohorts. The matching process was executed in R, version 4.1.0, especially employing the R package MatchIt, version 4.1.253. Statistical comparisons between the pregnant and non-pregnant cohorts, as well as between the different subgroups were performed using the Wilcoxon test for two-group comparisons, the Kruskal–Wallis test for comparisons between three groups (for continuous parameters) as well as Fisher’s exact test and Pearson’s χ2 test (for binary and categorical parameters). Correlations between markers were analysed by Spearman’s rho. All statistical tests were descriptive, and data analysis was performed with SAS® version 9.4 using SAS Enterprise Guide Version 8.3 on Microsoft Windows 10 Enterprise.

Methods

We evaluated classical prognostic breast cancer biomarkers (ER, PR, HER2, Ki-67) and immune response markers (HLA class I, HLA-G, PD-L1, TIGIT and Nectin-4) in a cohort of BC patients. TMAs underwent immunohistochemistry (IHC) staining to evaluate estrogen and progesterone receptor (ER, PR) status and human epidermal growth factor receptor 2 (HER2) status in accordance with the ASCO/ CAP guidelines54,55. Additionally, Ki-67 and markers relevant to immune response including HLA class I, HLA-G, PD-L1, TIGIT and Nectin-4, were assessed.

Hematoxylin-eosin-stained slides were analyzed to assess the count of stromal TILs in accordance with the guidelines of the International Immuno-Oncology Biomarker Working Group56. The TILs were categorized as low ≤25%, intermediate 26–60% and high >60%.

Positive nuclear staining in tumor cells was employed to assess the expression of Ki-67 as percentage, using the MIB-1 antibody (1:100 dilution, EDTA AG-retrieval, Dako/Agilent). The staining was then categorized as ≤20% vs. >20%. For PD-L1 expression assessment on tumor (TC) and immune cells (IC), the 22C3 antibody (1:50 dilution, EDTA AG-retrieval, Dako/Agilent) was used (see Fig. 1G–I). Positive expression was defined as ≥1% of TC or IC (proportion score) exhibiting membranous staining.

In addition to assessing the percentage of positive membranous staining for HLA class I (MHC I, HLA-ABC) with the EMR8-5 antibody (1:10,000 dilution, Citrat AG-retrieval, Abcam) (see Fig. 1A–C), HLA-G with the 4H84 antibody (1:100 dilution, Citrat AG-retrieval, Abcam) (see Fig. 1D–F), Nectin-4 with the EPR15613-68 antibody (1:2000 dilution, Trilogy AG-retrieval, Abcam) (see Fig. 2A–C) and TIGIT with the BLR047F antibody (1:500 dilution, Trilogy AG-retrieval, Abcam) (see Fig. 3A–D), H-scores for these markers were calculated as continuous variables. The formula used for calculation was: 3× percentage of strongly staining tumor cells + 2× percentage of moderately staining tumor cells + percentage of weakly staining tumor cells, ranging from 0 to 30057. Experimental biomarkers, including TIGIT, Nectin-4, HLA-A, and HLA-G were assessed using QuPath’s (V0.2.2)58 digital semi-automatic positive cell detection. Ki-67 on BCP/TMA was assessed using the Ki-67 Quantifier Tool from the Cognition Master Professional Suite (VMScope, Berlin, Germany).