Stromal Transcriptome Analysis of Human Lobular Breast Cancer Identifies a Role for Pregnancy-Associated-Plasma Protein-A


 Background: Invasive lobular carcinoma (ILC) is the second most common histological subtype of breast cancer and exhibits a number of clinico-pathological characteristics distinct from the more common invasive ductal carcinoma (IDC). Despite these differences, ILC is treated in the same way as IDC. We set out to identify alterations in the tumor microenvironment (TME) of ILC with potential clinical significance.Methods: We used laser-capture microdissection (LCM) to separate tumor epithelium from stroma in 23 ER+ ILC samples. Gene expression analysis was used to identify genes enriched in the stroma of ILC, but not IDC or normal breast. Results: 45 genes involved in regulation of the extracellular matrix (ECM) were enriched in the stroma of ILC, but not stroma from ER+ IDC or normal breast. Of these, 10 were expressed in cancer-associated fibroblasts (CAFs) and were increased in ILC compared to IDC in bulk gene expression datasets. PAPPA was the most enriched in the stroma compared to the tumor epithelial compartment in ILC. PAPPA encodes pregnancy-associated plasma protein-A (PAPP-A), a metalloproteinase that cleaves insulin-like growth factor-binding protein-4 (IGFBP-4), increasing IGF-1 bioavailability and downstream signaling. Analysis of PAPPA- and IGF1-associated genes identified a paracrine signaling pathway, and active PAPP-A was shown to be secreted from primary CAFs. Comprehensive survival analysis across 3,000 breast cancers identified PAPPA as a potential ILC-specific prognostic marker.Conclusions: This is the first study to demonstrate molecular differences in the TME between ILC and IDC, and it identifies PAPP-A, a CAF-derived proteinase, as a potential prognostic marker.


Background
Invasive lobular breast cancer (ILC) accounts for around 5-15% of breast cancers and is the second most common histological subtype after invasive breast cancer of no speci c type, commonly referred to as invasive ductal carcinoma (IDC). ILC is recognised to exhibit a number of clinico-pathological characteristics distinct from those of IDC [1,2]. It has an increased propensity for multi-centricity, multifocality and bilaterality, in addition to an unusual pattern of metastatic dissemination [3]. ILC is predominantly estrogen receptor (ER) and progesterone receptor positive, with low to absent expression of human epidermal growth factor receptor-2. Most patients with ILC are candidates for adjuvant endocrine treatment. Although response rates are initially good, an ILC diagnosis is associated with adverse long-term outcomes compared to IDC [4]. At the molecular level, ILC is de ned by a loss or reduced expression of E-cadherin, and several studies have further mapped the genomic landscape of ILC [5][6][7][8]. More recently, tumor-in ltrating lymphocyte populations have also been pro led [9]. ILC is characterized by having a dense stroma with a larger contact area with tumor cells than IDC, due in part to the difference in tumor growth pattern ("indian le" versus dense islands, respectively). However, little is known about the composition of the stroma or the role of the surrounding tumor microenvironment (TME). The TME plays a critical role in tumor behaviour by in uencing progression and metastatic spread, as well as therapeutic response [10,11], and in breast cancer, a stroma-derived prognostic predictor has been identi ed that strati es disease outcome independently of clinical prognostic factors [12].
In this study, we used laser-capture microdissection (LCM) to generate the rst human ILC stromal gene set. A number of genes were more highly expressed in the stroma of ILC, but not that of IDC or normal breast. These genes included PAPPA, which encodes pregnancy-associated plasma protein-A (PAPP-A), a metalloproteinase that cleaves insulin-like growth factor-binding protein-4 (IGFBP-4) when IGF-1 is bound to IGFBP-4 [13]. This results in a local increase in IGF-1 bioavailability and subsequent downstream signaling. Here, we further show that active PAPP-A is secreted from cancer-associated broblasts (CAFs), and also may represent an ILC-speci c prognostic marker.

Methods
All standard assays not detailed here are described in the Supplementary Methods (available online).
Tissue processing for LCM and gene expression analysis All samples were obtained from the McGill University Health Centre: Breast Cancer Functional Genomics Initiative Biobank, Montreal, Canada (study identi ers SUR-99-780 and SUR-2000-966). Tissue collection, LCM, sample isolation, RNA extraction and microarray hybridization were carried out as previously described [14] and analyzed using the SurePrint G3 Human GE 8x60K microarray kit. Processed and raw data are available from Gene Expression Omnibus (GSE148398).

Primary CAF dataset generation
All samples were obtained from the NHS Lothian Tissue Governance Committee, Edinburgh, United Kingdom (approval number 15/ES/0094). CAFs were isolated from ve ILC and three IDC samples, and total RNA was extracted (Qiagen). RNA was biotinylated using the Illumina TotalPrep RNA Ampli cation kit (Ambion). Samples were run on Illumina HT-12 v4 BeadChips. Quantile normalization was performed using the Lumi package. Processed and raw data are available from Gene Expression Omnibus (GSE148156).

Statistical analysis
All statistical analyses were two-sided, and p < 0.05 was considered statistically signi cant. Differential gene expression analyses from LCM and CAF datasets were calculated using rank products in MeV [15]. Differences in gene expression in whole tissues were assessed by Wilcoxon test, in primary CAFs by Mann-Whitney-Wilcoxon test and in KEP tumor and CAFs by t-test. Mann-Whitney-Wilcoxon test and fold change analysis were used to assess differences in gene expression between tumor and stroma in LCM datasets. RNAScope data were assessed by paired t-test. Correlation between PAPPA and IGF1 was assessed by Pearson correlation and linear regression analysis. Comprehensive survival analysis was performed using the survivALL R package [16].

Results
Generation of an LCM-ILC dataset LCM was performed on 23 ILC fresh frozen human samples: 17 were Grade 2 (74%), ve Grade 1 and one Grade 3. RNA was isolated from tumor epithelium (TE) and tumor stroma (TS) compartments (TS de ned here as primarily CAFs and matrix proteins, with the majority of immune cells being excluded). Gene expression data were generated for a total of 22 TE and 18 TS samples ( Figure 1A; Supplementary Figure   1), including matched pairs from 17 samples. 17 matched TE and TS samples from the same patient. Two-class paired rank product analysis (percent false positive (pfp) < 0.01) identi ed 1,082 genes consistently more highly expressed in the TS and 837 in the TE. These genes clustered the samples by compartment type (epithelium/stroma), showing successful microdissection of TE and TS compartments. Biomolecular pathway annotation revealed up-regulation of genes involved in extracellular matrix (ECM) remodeling, collagen degradation and integrin cell surface interactions in TS compared to TE, while genes related to cell cycle, DNA replication and methylation were up-regulated in TE compared to TS compartments ( Figure 1B).

ILC-speci c expression of primary cancer-associated broblasts
In order to perform functional studies, gene expression pro ling of primary CAFs from ILC and IDC tumors was performed (Figure 2A and B; Supplementary Figure 4). Rank product analysis identi ed 1,027 genes differentially expressed between ILC-and IDC-derived CAFs (pfp < 0.05); approximately half (485) were up-regulated in ILC CAFs compared to IDC CAFs ( Figure 2C) and were enriched for ECM-associated genes, along with genes involved in glycolysis, focal adhesions and members of the TGF signaling pathway ( Figure 2D).
Of the 45 ILC-speci c stromal genes identi ed, 28 were expressed in the CAF dataset (Table 1; Figure 2E).
Functional network analysis (http://genemania.org/) identi ed that 24 of these 28 genes were in the same pathway or are linked by known genetic or physical interactions (Supplementary Figure 5). The majority (14/24) were signi cantly up-regulated in ILC compared to IDC (p < 0.05) in at least one of the published bulk datasets (Table 1). Clustering the 10 genes with a TS/TE fold change >2 in the LCM-ILC dataset clearly showed increased expression in the TS of ILC, but not in IDC or normal breast ( Figure 2F).
PAPPA is predominantly expressed in the stroma of ILC PAPPA showed the greatest fold change expression in the stromal compared to epithelial compartments in ILC (log2(FC(TS/TE)) = 2.6) ( Table 1), so it was selected for further analysis. PAPPA encodes PAPP-A, a secreted metalloproteinase that cleaves IGFBP-4, releasing bioactive IGF-1 [20]. PAPP-A activity can be inhibited by non-covalent or covalent complex formation with endogenous inhibitors stanniocalcin-1 (STC1) or -2 (STC2), respectively [21,22] ( Figure 3A). To verify that PAPPA was expressed predominantly by CAFs, we analyzed PAPPA transcripts by RNAScope. Results con rmed higher levels of PAPPA expression in CAFs compared to tumor cells, although epithelial PAPPA transcripts were also seen in some tumors ( Figure 3B). We then examined expression of PAPPA and functionally related genes in the LCM-ILC, -IDC and -normal datasets ( Figure 3C). In ILC, both PAPPA and IGF1 were signi cantly more highly expressed in the stroma compared to the epithelium (p < 0.0001), while IGF1R was found predominantly in the tumor epithelium (p < 0.05), suggesting the presence of a potential paracrine activation loop. In IDC and normal breast tissue, IGF1 was also predominantly expressed in the stroma (p < 0.0001 and p < 0.01, respectively), whereas PAPPA was expressed in both stromal and epithelial compartments ( Figure 3C). Interestingly, we found a clear positive correlation between PAPPA and IGF1 in ILC (r = 0.64, p < 0.0001), which was not observed in IDC and normal tissue ( Figure 3D). As there are few cell lines that represent ILC, we rst examined PAPPA across three integrated breast cancer cell line datasets [23]. PAPPA was low or undetectable in all luminal cell lines, including two reported ILC lines, SUM44-PE and MDA-MB-134VI ( Figure 4A). qPCR con rmed that PAPPA was not expressed in SUM44-PE and MDA-MB-134VI cells or the T47D and MCF-7 ER+ IDC lines. Analysis of 11 primary patient-derived ILC CAFs, 5 IDC CAFs, and HCI 013, an ILC patient-derived xenograft, showed that PAPPA and IGF1 were expressed exclusively in CAFs ( Figure 4B; Supplementary Figure 6). In contrast, IGF1R was mostly expressed by tumor cells ( Figure 4B). We also separated tumor epithelial cells from CAFs in tumors derived from a mouse model of ILC driven by loss of Trp53 and Cdh1 (Supplementary Figure 7A) [24]. qPCR results showed that Pappa, Igf1 and Stc1 were only expressed in the CAFs, while Igf1r, Stc2 and Igfbp4 were expressed in both tumor cells and CAFs (Supplementary Figure 7B). Together, these data support a wider paracrine signaling role for PAPP-A in luminal tumors.

PAPP-A secreted by CAFs is active
We analyzed conditioned media (CM) and con rmed that PAPP-A was secreted by CAFs but not the tumor cells ( Figure 5A). PAPP-A needs to be active in order to cleave IGFBP-4 and liberate IGF-1. CM from the CAFs was able to cleave recombinant IGFBP-4, indicating that non-complexed, active PAPP-A was present in the media ( Figure 5B). To con rm that the IGFBP-4 fragments generated by the CAF CM were a result of PAPP-A activity, the CM was treated with a PAPP-A inhibitory antibody, (mAb 1/41) [25]. Preincubation with mAb 1/41 reduced levels of the cleaved IGFBP-4 fragment to those in the control lane, showing that the observed cleavage of IGFBP-4 is due to PAPP-A present in the CM ( Figure 5C).
PAPPA expression is positively correlated with IGF1 and negatively with IGF1R and is elevated in CDH1−, claudin-low tumors Investigation of PAPPA and related genes in large cohorts of breast cancers con rmed positive correlations between PAPPA and IGF1 and negative correlations with IGF1R and CDH1 ( Figure 6A and 6B). An integrated compendium of 17 Affymetrix datasets representing 2,999 breast cancers [23] revealed that PAPPA is detectably expressed in 36% of tumors, the highest proportion of which were of the claudinlow subtype, which also had signi cantly higher levels of PAPPA expression ( Figure 6A). Somewhat surprisingly, 7% of ILCs in the METABRIC dataset are classi ed as claudin-low, compared to only 4% of IDCs, and PAPPA expression was signi cantly higher in tumors where CDH1 was undetectable ( Figure  6B).
Low PAPPA expression in ILC, but not IDC, is associated with poor outcomes To examine whether PAPPA relates to prognosis of breast cancers, we performed comprehensive survival analysis using the survivALL R package [16]. Low PAPPA expression was signi cantly associated with worse overall survival for a large proportion (807/1,903) of all possible (n − 1) cut-points for 1,904 breast cancers in the METABRIC cohort (p < 0.05). However, none of the cut-points for PAPPA expression were signi cantly associated with overall survival across the 1,098 breast cancers of the TCGA dataset (Supplementary Figure 8A). Restricting analysis to the 142 and 206 ILCs of the METABRIC and TCGA cohorts, respectively, identi ed a number of cut-points that were signi cantly associated with overall survival in both cohorts ( Figure 6C; Supplementary Figure 8B). To compare this with IDCs from the two cohorts, analysis was limited to the same numbers (142 and 206) of ER+ IDC tumors with 10-fold random sampling, but on no occasion were any signi cant cut-points identi ed ( Figure 6D; Supplementary Figure   8C). Taken together, these results suggest that reduced or absent PAPPA is an ILC-speci c prognostic marker.

Discussion
In this study, we provide the rst comprehensive analysis of the TME in ILC. This identi ed a set of 45 genes enriched in the stroma of ILC, but not in the stroma of IDC or normal breast, that were associated with ECM regulation. CAFs are an important component of the TME that regulate ECM deposition [26].
Little is known about the involvement of CAFs in the biology of ILC, although one study has reported an increased CAF density in ILC compared to matched IDC samples [27], while differences in collagen deposition and alignment have also been reported [28,29]. Our analysis of primary CAFs from ILC and IDC shows differences in a number of genes involved in ECM interactions and signaling pathways, indicating that CAFs are able to differentially in uence the TME in ILC compared to IDC. A wider analysis of CAFs from ILC will be required to fully understand their biological and clinical importance and also whether the heterogeneity and different sub-populations of CAFs that exist in other breast cancers are pertinent to ILC [30,31]. Interestingly, four of the stromal-derived ECM-associated genes (PRKCA, ITGA10, NOV, WNT5B) that we identi ed in ILC were found in the reactive-like ILC subtype described by Ciriello and colleagues [5]; these reactive-like ILC tumors largely associate with the reactive subgroup identi ed by TCGA that are characterized by strong microenvironmental and CAF signaling [32].
We focused here on PAPPA, which encodes an IGF-promoting proteinase. Further analysis of PAPPA and functionally associated genes showed that both PAPPA and IGF1 were predominantly expressed in the stroma of ILC, while IGF1R was expressed within the tumor epithelium. Together with the demonstration that active PAPP-A is secreted from CAFs, this strongly supports the existence of a paracrine signaling pathway in ILC. However, of the few immunohistochemical studies reporting on PAPP-A in breast tumors, only expression in the tumor epithelium has been recorded [33,34]. Of note, PAPPA was elevated in the reactive ILC breast cancer subtype (characterized by proteins produced by the TME and CAFs) when analyzing both TCGA and METABRIC datasets (Supplementary Figure 8D). Further analysis of a panel of breast cancer cell lines showed that PAPPA was low or undetectable in all luminal cell lines, in support of a wider PAPP-A paracrine, rather than autocrine, signaling axis in other breast cancer subtypes.
Previously, we have shown that loss of E-cadherin promotes hypersensitization of PI3K/Akt pathway activation in response to IGF1, independent of PAPP-A [35], and of oncogenic mutations in the PI3K/Akt pathway that are prevalent in ILC [5]. This is consistent with other reports demonstrating that E-cadherinmediated adhesion negatively regulates IGF1R activation [36,37]. Furthermore, in breast cancer models, loss of E-cadherin and the subsequent activation of IGF1R signaling results in increased sensitivity to dual IGF1R/Insulin receptor inhibitors, and Akt inhibitors that target downstream receptor pathway activation, even in the presence of activating PIK3CA mutations [35,36]. Interestingly, increased expression of IGF1 is seen in ILC compared to IDC [27,35,38], consistent with reported pathway activation [35,36]. Together, these data suggest that patients with ILC may bene t from treatments targeting the IGF1 signaling pathway.
Although a number of strategies to target the IGF1/IGF1R signaling axis have been tested, results in the clinical setting have been disappointing, with a number of contributory factors including effects on systemic glucose metabolism and associated metabolic toxicities and lack of predictive biomarkers [39].
Indirect targeting of IGF1R signaling via inhibition of PAPP-A proteolytic activity may provide a viable alternative by reducing the levels of bioactive IGF1 speci cally in the local microenvironment of tumors that express high levels of PAPP-A. Consistent with this are reports that treatment with a monoclonal antibody that blocks the proteolytic activity of PAPP-A reduces tumor growth in both lung cancer and Ewing sarcoma models, which express high levels of PAPP-A within the tumor cells [25,40]. In addition, in a series of patient-derived ovarian tumor xenografts, antitumor activity of the PAPP-A inhibitory antibody correlated with PAPP-A expression [41]. Interestingly, in the 4T1 murine model of breast cancer, expression of a protease-resistant IGFBP-4, or treatment with recombinant PAPP-A-resistant IGFBP-4, inhibited tumor growth and metastasis via inhibition of PAPP-A secreted by the stroma [42,43].
A recent study found that circulating levels of PAPP-A in breast cancer patients were independently prognostic for recurrence-free and overall survival [44]. Thus, circulating levels of PAPP-A may provide a route to patient strati cation when considering targeting PAPP-A. However, PAPP-A has also been proposed to be a tumor suppressor following the discovery that it is epigenetically silenced in breast cancer precursor lesions [33], and comprehensive survival analysis in this study of the two largest publicly available breast cancer gene expression datasets indicate that reduced PAPPA is associated with worse outcomes in ILC but not IDC. The reason(s) for this apparent discrepancy is not clear. The link between PAPPA and clinical outcomes may re ect the increased CAF density that has been reported in ILC, and further understanding of the different CAF subtypes in ILC is required. It is also important to consider that PAPPA levels do not re ect active proteinase activity of the secreted PAPP-A. PAPP-A is predominantly found bound to its endogenous inhibitors STC1 and STC2, and it will be important in the future to consider measurements of active PAPP-A when considering PAPP-A as a potential therapeutic target. Interestingly, PAPPA levels were negatively associated with CDH1, and PAPPA was signi cantly higher in tumors where CDH1 was undetected, while the highest proportion of PAPPA positive tumors were of the claudin-low subtype, which are characterized by low or absent expression of CDH1 [45], suggesting a potential functional link between PAPP-A and E-cadherin.
Overall, this study demonstrates that PAPP-A is an important stromal factor in ILC. Intriguingly, the correlation between E-cadherin and PAPP-A, together with the reported role for E-cadherin in regulating IGF1R activation, indicates that there is a wider role for PAPP-A in regulating breast cancer growth.

Conclusions
This is the rst detailed analysis of the tumor microenvironment in ILC following laser-capturemicrodissection (LCM) of human ILC samples to separate tumor epithelium from surrounding stroma. Gene expression analysis of this LCM-dataset, and cancer-associated broblasts isolated from both ILC and IDC, demonstrate novel molecular differences in the tumor microenvironment between ILC and IDC and identify genes that are enriched speci cally in the stroma of ILC. We identi ed pregnancy-associatedplasma protein-A, a secreted CAF-derived proteinase, as the rst gene to be a potential ILC prognostic marker following comprehensive survival analysis of large patient cohorts.

Declarations
Ethics approval and consent to participate Samples for LCM were obtained from the McGill University Health Centre: Breast Cancer Functional Genomics Initiative Biobank, Montreal, Canada (study identi ers SUR-99-780 and SUR-2000-966). For CAF generation samples were obtained from the NHS Lothian Tissue Governance Committee, Edinburgh, United Kingdom (approval number 15/ES/0094).

Consent for publication
Not applicable Availability of data and materials Processed and raw data are available from Gene Expression Omnibus GSE148398 and GSE148156.

Competing interests
The authors declare that they have no competing interests.   PAPPA in in vitro models of ILC A) Heatmap representing the expression of PAPPA and associated genes across three integrated panels of breast cancer cell lines following batch correction, ranked by CDH1 expression. Blue, luminal; red, basal; yellow, claudin-low. Gray bars indicate samples where the detection call for PAPPA is assigned as 'present' and those tumors where CDH1 is 'absent'. B) Expression of PAPPA, IGF1, IGF1R, SCT1, SCT2 and IGFBP4 by qPCR in ILC and IDC human cell lines and primary CAFs. Each sample was analyzed at three different passage numbers, and its average represented as the relative mRNA expression to ED30 primary ILC CAFs. Line represents the mean with SEM.  Low PAPPA expression in lobular, but not ductal, breast cancers is associated with poor outcomes A) Breast tumors of the METABRIC cohort ranked by PAPPA expression indicates positive correlation with IGF1, negative correlation with IGF1R and highest expression in claudin-low subtype tumors. ****p ≤ 0.0001. Heatmap shows log2-transformed mean-centered expression values (red, high; blue, low).
Subtypes are indicated by the colored bar above the heatmap (red, basal; purple, ERBB2; dark blue,