Fibroblasts direct differentiation of human breast epithelial progenitors

Background Breast cancer arises within specific regions in the human breast referred to as the terminal duct lobular units (TDLUs). These are relatively dynamic structures characterized by sex hormone driven cyclic epithelial turnover. TDLUs consist of unique parenchymal entities embedded within a fibroblast-rich lobular stroma. Here, we established and characterized a new human breast lobular fibroblast cell line against its interlobular counterpart with a view to assessing the role of region-specific stromal cues in the control of TDLU dynamics. Methods Primary lobular and interlobular fibroblasts were transduced to express human telomerase reverse transcriptase (hTERT). Differentiation of the established cell lines along lobular and interlobular pathways was determined by immunocytochemical staining and genome-wide RNA sequencing. Their functional properties were further characterized by analysis of mesenchymal stem cell (MSC) differentiation repertoire in culture and in vivo. The cells’ physiological relevance for parenchymal differentiation was examined in heterotypic co-culture with fluorescence-activated cell sorting (FACS)-purified normal breast primary luminal or myoepithelial progenitors. The co-cultures were immunostained for quantitative assessment of epithelial branching morphogenesis, polarization, growth, and luminal epithelial maturation. In extension, myoepithelial progenitors were tested for luminal differentiation capacity in culture and in mouse xenografts. To unravel the significance of transforming growth factor-beta (TGF-β)-mediated crosstalk in TDLU-like morphogenesis and differentiation, fibroblasts were incubated with the TGF-β signaling inhibitor, SB431542, prior to heterotypic co-culture with luminal cells. Results hTERT immortalized fibroblast cell lines retained critical phenotypic traits in culture and linked to primary fibroblasts. Cell culture assays and transplantation to mice showed that the origin of fibroblasts determines TDLU-like and ductal-like differentiation of epithelial progenitors. Whereas lobular fibroblasts supported a high level of branching morphogenesis by luminal cells, interlobular fibroblasts supported ductal-like myoepithelial characteristics. TDLU-like morphogenesis, at least in part, relied on intact TGF-β signaling. Conclusions The significance of the most prominent cell type in normal breast stroma, the fibroblast, in directing epithelial differentiation is largely unknown. Through establishment of lobular and interlobular fibroblast cell lines, we here demonstrate that epithelial progenitors are submitted to stromal cues for site-specific differentiation. Our findings lend credence to considering stromal subtleties of crucial importance in the development of normal breast and, in turn, breast cancer.

Population doubling level (PDL) was calculated as: PDL = 3.32 (log I -log Y) +X, where I is the cell number of the inoculum, Y is the cell yield and X is the population doubling of the inoculum. The hTERT immortalized breast broblasts have currently been propagated for more than 80 passages (available through Ximbio, UK, IAHF, cat. no. 153783 and IEHF, cat. no. 153784).
Subcon uent broblast cultures in passage eight were transduced with the viral supernatant supplemented with 8 μg/mL polybrene at serial dilution over-night upon when the medium was replaced. At 90% con uency, the transduced cells underwent antibiotic selection with medium containing 300 μg/mL G418 (Life Technologies) for nine days until non-transduced control cells showed no signs of cell survival. The concentration of antibiotic used was determined prior to transduction by testing different concentrations of G418 and choosing the dose of 300 μg/mL G418, which eliminated all cells within one week. The transduction e ciency was not more than 15%, in which the majority of cells were transduced by one copy of retroviral particle [20]. RNA extraction, RT-qPCR, and Next generation sequencing To measure hTERT expression, total RNA was extracted from hTERT-transduced HBFCs, iHBFCs, and empty vector-transduced HBFCs, evHBFCs, in passage 11 according to the manufacturer's instructions (Sigma, GenElute, RTN70) and the RNA was reverse transcribed to cDNA using the High Capacity RNA-to-cDNA Kit (Applied Biosystems). Real-time quantitative polymerase chain reaction (RT-qPCR) was performed as described [11] using TaqMan Gene Expression Assays (Applied Biosystems) and the TaqMan primers: human telomerase reverse transcriptase (hTERT, Hs00972656_m1), Glyceraldehyde-3-phosphate-dehydrogenase (GAPDH, Hs02758991_g1), hypoxanthine phosphoribosyltransferase 1 (HPRT1, Hs99999909_m1) and phosphoglycerate kinase 1 (PGK1, Hs00943178_g1). Gene expression was determined using the formula 1/(2 ΔCT ), in which ΔCT represents the difference between the target and the geometric mean of reference genes. GAPDH, HPRT1 and PGK1 served as reference genes for normalization.
For next generation sequencing, total RNA was extracted using Trizol (Thermo Fischer) and a spin column method according to the manufacturer's instructions (Zymo Research) from subcon uent duplicate cultures of HBFC CD105 and HBFC CD26 in passage 9, and from duplicate cultures of passage 24 iHBFC CD105 and passage 25 iHBFC CD26 . RNA sequencing and bioinformatics analysis was performed by the Beijing Genomics Institute (BGI), Hong Kong as previously described [11]. In brief, sequencing was performed using BGISeq 500 and 13.7 M clean reads were generated for each sample. Mapped clean reads to reference using Bowtie 2 tool [21] were then used to calculate gene expression with the RSEM package [22]. To identify differentially expressed genes (DEGs) between groups, the DESeq2 method was used [23]. A Venn diagram (https://bioinfogp.cnb.csic.es/tools/venny/index.html) was used to depict the overlap of DEGs with a 2 fold difference between broblast populations.
For analysis of cluster of differentiation (CD) molecular signature, a comprehensive list of 453 unique CD molecules and their gene names was retrieved from the Uniprot database (https://www.uniprot.org/docs/cdlist) and applied to lter DEGs with a 2 fold difference and FPKM larger than 5. The R software (v3.2.2) was used to plot gene expression values in a heatmap.
Freshly isolated myoepithelial cells were plated (2,500 -5,000 cells/cm 2 ) onto con uent broblasts feeder layers of iHBFC CD105 and iHBFC CD26 , respectively. Myoepithelial/ broblast co-cultures were maintained in a specialized culture medium, Myo medium [11], supplemented with 5% FBS (Myo 5%). In one experiment, cultures were maintained in DMEM/F12-5%, which gave a similar result. Primary myoepithelial cells were also plated on collagen coated plastic in Myo medium and expanded to passage 2 before use in co-cultures with broblasts in passage 3 using Myo 5% medium.
To isolate myoepithelial cells from co-cultures, the cell cultures were trypsinized (0.25% trypsin/1 mM EDTA), counted using a Burker-Türk chamber and stained for CD271-APC at 4°C for 30 minutes followed by two washes in HEPES/BSA/EDTA buffer. Fixable Viability Stain 780 (1:1000, BD Biosciences) live-dead discriminator was added prior to analysis and sorting on FACS ARIA-II or FACS Fusion (BD Biosciences). FACS data analysis was performed with FACS DIVA and FlowJo software.
In a cross-over test, myoepithelial cells in primary culture were isolated from co-cultures with iHBFC CD105 and iHBFC CD26 , respectively, and from each, 1,600 myoepithelial cells/cm 2 were re-plated onto con uent broblast feeders of both iHBFC CD105 and iHBFC CD26 . To account for variance in absolute CD271 levels and for normalization purposes, myoepithelial CD271 levels were divided by the mean background CD271 uorescence of the co-cultured broblasts.
For assessment of epithelial morphogenesis, FACS sorted primary MUC1 high luminal cells (6,000 cells/ cm 2 ) were seeded in Myo medium onto con uent feeder layers of iHBFCs and observed for up to three weeks using a phase contrast microscope and imaged (Leica DM IL).
In 15 tests using TGF-β signaling inhibition by SB431542 (Axon 1661, Axon Medchem), HBFCs representing four biopsies were allowed to grow to con uence over 7 days and were then treated with 10 μM SB431542 for 3 days before plating of MUC1 high luminal cells at day 10 from ve biopsies.
In two tests, MUC1 high luminal cells from two biopsies were plated onto con uent HBFCs from two biopsies in Myo medium. From day 2-9 the co-cultures were exposed to 10 μM SB431542 or vehicle (DMSO).

Immunohistochemistry and immunocytochemistry
Cryostat sections of normal breast tissue biopsies, xenografts as well as cultured cells and cell smears were stained essentially as previously described after xation in either methanol (M in Table 1) or formaldehyde (F1 in Table 1) or formaldehyde followed by methanol:acetone (F2 in Table 1) and included controls without primary antibody [12,27,28]. Blocking was performed for 5 minutes in 10% goat serum in PBS or Ultra V Block (Lab Vision Corporation TA125-UB). Cells were incubated with primary and secondary antibodies for 60 and 30 minutes respectively (Table 1). For immunoperoxidase staining, the secondary antibody was UltraVision ONE HRP Polymer (Thermo Fisher, TL-125-PHJ) and for uorescence, isotype-speci c goat anti-mouse IgG AlexaFluor (AF, Life Technologies) secondary antibodies were used. Nuclei of immunoperoxidase-or uorescence stained sections and cells were counterstained with hematoxylin or ProLong Gold Antifade reagent with 4, 6-diamino-2-phenylindole (DAPI, Life Technologies), respectively. 11 pairs of iHBFC CD105 and iHBFC CD26 spanning passages 11-50 were stained by immunoperoxidase for CD105 (Abcam, SN6) and CD26 (Abcam, 202-36). Photographs were acquired with Leica DM5500B.
MUC1 high -luminal/ broblast co-cultures were immunoperoxidase-stained on day 9-12 for Keratin 19 (GenWay or abcam, BA16) and images acquired on Leica Z6 AP0 at 1.25 magni cation. The images were analyzed with ImageJ software (v1.52a) in batch mode using a macro previously established [10] counting the number of epithelial structures larger than 0.0026 mm 2 .
Myoepithelial/ broblast co-cultures were co-stained for K14 (Monosan, LL002), K17 (Dako, E3) and K19 (Abcam, BA16), followed by AF488 (IgG3), AF568 (IgG2b) and AF568 (IgG1). Images of three co-cultures representing three different biopsies were acquired with Leica DM5500B and K17 intensity measured with image analysis software, ImageJ (1.52a). For this, segmentation was rst performed on K14 using the ImageJ functions Multiply, Median and Make Binary providing the outline of the myoepithelial cells. This segmentation was then applied to corresponding images of K17 in which uorescence intensity was measured.
Cultures subjected to luminal differentiation conditions were stained for K19 (Abcam, BA16) by immunoperoxidase on day 8-12, evaluated and imaged using Leica DM5500B.
For a quantitative assessment of CD271 as a marker for ductal myoepithelium, cellular smears were prepared from FACS-isolated CD271 high versus CD271 low myoepithelial cells from four different biopsies. The smeared cells were xed at room temperature for 10 minutes in 3.7% paraformaldehyde, washed three times in PBS and permeabilized in 0.01% Triton X-100 for 10 min followed by three washes in PBS. The xed smears were blocked by 5 minutes incubation in Ultra V Block followed by 5 minutes in 10% goat serum before staining with K17 (Dako, E3) antibody, followed by AF488 (IgG2b) and DAPI. Images of stained smears were acquired with Leica DM5500B and a minimum 100 cells per cell preparation was counted using ImageJ (v1.52a) Cell Counter plugin.

In vivo morphogenesis
From primary co-culture with iHBFC CD105 or iHBFC CD26 , approximately 500,000 myoepithelial cells, with or without removal of co-cultured CD271 low broblasts by FACS, representing two biopsies, were admixed with 125,000 or 500,000 irradiated (~20Gy) iHBFC CD105 or iHBFC CD26 cells and suspended in cold 1:1 collagen gel: with Tukey's test, Kruskal-Wallis rank-sum test, Wilcoxon signed-rank test or nested t-test, as indicated.

Immortalization of human breast broblastic cells (HBFCs)
We previously puri ed broblasts from reduction mammoplasty specimens and sorted them into lobular CD105 high /CD26 low and interlobular CD105 low /CD26 high lineages which could be propagated in culture [10].
Under these conditions, HBFCs senesce after more than 80 days and approximately 16 passages [10]. To generate lines of HBFCs, we here examined whether retroviral transduction with the hTERT gene would render HBFCs immortal. HBFCs in passage eight were infected with retrovirus encoding hTERT together with a neomycin drug resistance marker or an empty vector. Whereas the empty vector cells did not exhibit extended lifespan over what is expected for HBFCs, the hTERT transduced cells generated populations of infected HBFCs with no signi cant growth arrest and an apparent in nite life span ( Fig. 1a and Additional le Fig. 1).
Interestingly, the CD105 high -and CD26 high -derived cell lines given identical growth conditions, stably exhibited different growth properties (Fig. 1a), and have currently been grown for more than 80 passages. Thus, immortalization was successful, and in the following, we refer to the hTERT transduced breast broblasts as iHBFC CD105 and iHBFC CD26 , respectively.

Differentiation state of iHBFCs
To characterize the hTERT immortalized lines, we rst examined their staining pattern with CD105 and CD26.
As seen in gure 1b, iHBFC CD105 and iHBFC CD26 maintain high expression of CD105 and CD26, respectively ( Fig. 1b). In order to further investigate the differences between the two cell lines and in parallel the nite lifespan HBFCs, we next examined the mRNA expression pro les of the iHBFC CD105 and iHBFC CD26 . We found that there were approximately 850-900 transcripts in each population that were >2 fold differentially expressed compared to the other population, and that in general, the iHBFCs remained well differentiated along lobularand interlobular-broblastic pathways, respectively (Fig. 2a). Thus, in contrast to previous attempts to culture and maintain lobular and interlobular breast broblast [8,9] and dermal broblast subpopulations [14], the lineages in the present study remain phenotypically distinct in extended culture and upon immortalization.
Since multiple broblast subpopulations have been characterized in human dermis based on expression of different combinations of cluster of differentiation (CD) genes [14] we next speci cally extracted this information from the mRNA arrays of the iHBFCs (Fig. 2b). The list of 34 genes contained several well-known broblast markers, including CD248 (endosialin/TEM1, [32]), CD36 (scavenger receptor class B member 3, SCARB3, [33]), CD34 [34], CD140b (platelet-derived growth factor receptor-beta, PDGFRb; [35]), CD138 (syndecan-1, [36]), CD90 (Thy-1, reviewed in [37]), and CD13 (aminopeptidase N, ANPEP, [8,38]). Among these, CD90 and CD140b have been de ned as pan-broblast markers, that is genes expressed at a high level in both papillary and reticular dermal broblasts and all cultured broblast lines [14]. In the present study, however, the expression levels of these markers appear to distinguish lobular and interlobular iHBFCs, since CD140b is expressed at a higher level in the former, and CD90 is expressed at a higher level in the latter (Fig. 2b). Upon further comparison with human dermis the most obvious equivalent expressing CD26 is the papillary broblast, while CD105 expression concurs with CD36, which is expressed in both lobular breast broblasts and lower reticular dermis [39]. Indeed, the iHBFCs serve as a sensible model with relevance to the in vivo setting, which was further illustrated in a series of 10 specimens, where, in addition to CD26 and CD105, two of the identi ed markers of iHBFCs, CD140b and CD248, recognize the cells in situ which they are supposed to represent (Fig. 2c). This pattern was observed in 8/10 cases. In 2/10 cases no difference in staining was observed between lobular and interlobular stroma.
Next, we analyzed whether the two cell lineages had also retained critical functional properties in spite of immortality. We have previously shown that CD105 high as opposed to CD26 high HBFCs in several respects behave like MSCs [10]. Here, we conducted a series of experiments between passage 22 and passage 50 to reveal the potential of the iHBFCs with respect to functional differentiation towards adipocyte and osteoblast lineages. Indeed, the iHBFCs remained discernably stable for the entire culture period with respect to their differentiation potential as demonstrated by accumulation of lipid droplets in adipogenic cultures and formation of mineralized matrix in osteoblastic cultures of iHBFC CD105 only (Additional le Fig. 2a and b). Also in this respect, iHBFC CD105 show similarity to reticular broblasts, which readily undergo adipogenic differentiation [39]. iHBFC CD105 do not, however, exhibit the entire differentiation repertoire of MSCs, since they differ from bone marrow-derived MSCs by lack of ability to form bone in vivo ( Additional le Fig. 2c). Hence, the iHBFC CD105 and iHBFC CD26 retain critical properties of primary cells and of their putative cells of origin and share lineage relationships with broblasts from other tissues.

Fibroblast cell type and impact on breast epithelial progenitors
With a reproducible source of lobular-and interlobular-like HBFCs in hand, we assessed their impact on the neighboring breast epithelium. Firstly, we looked at the luminal epithelial compartment characterized by a high cellular turnover in vivo [12]. Here we took advantage of a heterotypic co-culture assay designed for measuring branching morphogenesis [10,40]. As seen in Fig. 3, the readout from this assay was an unequivocal high level of branching morphogenesis supported by iHBFC CD105 . This difference between iHBFC CD105 and iHBEC CD26 in inductive capacity was robust throughout the entire culture period from passage 14 to 47 and was independent of source of epithelial cells (Fig. 3). Secondly, we looked at the myoepithelial compartment, which is believed to contain the apical-most progenitors in the human breast hierarchy [41][42][43]. Here, we took advantage of the fact that ductal and lobular myoepithelial cells in situ differ in both their marker expression and their differentiation potential [11]. The question remains as to whether these properties to some extent rely on topographical conditions such as those determined by the adjacent broblasts. To address this, we isolated the entire complement of myoepithelial cells from three different biopsies by a CD271 FACS protocol.
These myoepithelial cells were plated directly on either iHBFC CD26 or iHBFC CD105 and cultured for one week followed by staining for keratin K17 (Fig. 4a) and CD271 (Fig. 4b). Notably, the readout for ductal-like myoepithelial differentiation was based on both high CD271 and high keratin K17 since these co-segregated in FACS and stainings (Additional le Fig. 3). Interestingly, ductal-like, high expression of both CD271 and K17 entirely relied on co-culture with iHBFC CD26 . That broblasts indeed in uence epithelial differentiation was further substantiated by passaging the cells to a second passage with switching of the feeders. Now, those myoepithelial cells that were initially ductal-like in phenotype with high CD271 expression became lobular-like with reduced CD271 expression and vice versa (Fig. 4c). This indicates that the myoepithelial phenotype is regulated by surrounding broblasts.
Whether this also applies to the next level of differentiation potential of myoepithelial cells, i. e. generation of luminal cells, was examined by measuring the pattern of induced luminal keratin K19 in myoepithelial progenitors under differentiating conditions. Whereas lobular-like luminal differentiation is characterized by emergence of scattered heterogeneous islets of K19-positive luminal cells, ductal-like luminal differentiation entails homogeneous islets reminiscent of their differentiation in vivo [11]. Accordingly, myoepithelial cells primed by co-culture with either iHBFC CD105 or iHBFC CD26 were plated at clonal density under identical luminal differentiation conditions without broblast feeders [11]. Based on experiments with 6 different biopsies we found that priming with either iHBFC CD105 or iHBFC CD26 impacted on the following luminal differentiation potential corresponding to preferentially scattered or homogeneous keratin K19 staining, respectively (Fig. 5a).

This observation was further validated in vivo. Myoepithelial cells primed in co-culture with either iHBFC CD105
or iHBFC CD26 orthotopically injected into NOG mice resulted in bilayered epithelial structures from both origins in 6/10 and 5/8 transplants, respectively. However, while iHBFC CD105 co-culture-derived myoepithelial cells gave rise to K14 -/low /K19 + cells, iHBFC CD26 co-culture-derived myoepithelial cells gave rise to K14 + /K19 + luminal cells (Fig. 5b). Taken together, these results imply that broblasts in uence epithelial progenitors and that lobular broblasts support the development of a more mature luminal phenotype characteristic of TDLU.

Interruption of a TGF-b signaling cascade in HBFC CD105 and control of epithelial progenitors
Since lobular broblasts exhibit a TGF-b signaling signature [10] and CD105 is a co-receptor for TGF-b (reviewed in [44]), we speculated whether the TGF-b signaling pathway plays a role in the crosstalk between broblasts and epithelial progenitors. To explore this, we used the quantitative morphogenesis assay described above and initially incubated luminal epithelial-broblast co-cultures directly with the small molecule TGF-b signaling inhibitor, SB431542, previously shown by others to impinge on CD105 signaling [45]. Indeed, in 2 out of 2 tests, the number of epithelial structures in HBFC CD105 , but not in HBFC CD26 cocultures, was reduced by treatment with SB431542 (Additional le Fig. 4). To exclusively target the broblasts, we then incubated con uent broblast feeders with SB431542 for three days prior to plating of the luminal cells on top. Disruption of TGF-b signaling signi cantly reduced epithelial structure formation in HBFC CD105 co-cultures, but not in HBFC CD26 co-cultures (Fig. 6). This result suggests that intact TGF-b signaling in lobular broblasts is instrumental in modulating parenchymal cells.

Discussion
A number of contextual signals have been implicated in the maintenance of tissue homeostasis in the human breast some of which originating from neighboring broblasts and impacting on stem cell behavior (reviewed in [46]). Also, it has long been suspected that broblasts exhibit functional specialization according to their anatomical location [9,47,48], but it still remains an open question which cells identify the stromal microenvironment, and how they are speci ed for the production of proliferation and differentiation cues [46].
Our research in adult breast tissue has revealed the existence of two distinct lineages -a lobular and an interlobular which remain inherently functionally distinct [10]. Here, using hTERT expression vectors we have been able to generate two populations of cells that reside stably in the lobular-like and interlobular-like states, respectively, as de ned by a number of properties including the CD105 and CD26 expression. The resulting broblast cell lines are faithful to their identity corresponding to their anatomical site of origin, and speci cally, lobular-like broblasts, relying on a TGF-b signaling pathway, govern epithelial morphogenesis and differentiation typical of the TDLU.

Page 14/28
The above observations leave several questions unanswered about the role of broblasts in the human breast.
We have previously shown that lobular-derived and ductal-derived epithelial cells maintain their distinct properties either in the absence of broblasts, that is in three-dimensional culture within a reconstituted basement membrane, or in co-culture on mouse-derived broblasts (3T3-cells) suggesting that epithelial cells are not submitted to modulation by microenvironmental cues [11,12]. In the present study, however, we show that early myoepithelial progenitors are susceptible to cues from lobular-and interlobular-like broblasts in terms of luminal differentiation repertoire. Thus, if primed on lobular-like broblasts luminal differentiation is more elaborate, that is reminiscent of the luminal lineage in TDLUs in situ. On the other hand, if myoepithelial progenitors are primed with interlobular-like broblasts, the luminal differentiation is limited to K14 and K19 double positive progenitors both in culture and in vivo. It is possible that human breast epithelial progenitors for appropriate interaction with the surrounding stroma rely on species-speci c crosstalk. This notion is supported by an experimental paradigm described more than a decade ago, when it was shown that normal morphogenesis and differentiation of human breast epithelial cells transplanted into mice required coimplantation with human broblasts [49]. Our present ndings extend this observation to include plasticity of prospectively isolated human breast progenitors as determined by positional information from resident broblasts.
Lobular-like human breast broblasts generated either by prospective FACS isolation from primary tissue or through hTERT immortalization exhibit a strong expression of CD105. A number of studies have indicated that CD105 modulates TGF-b signaling through ALK5 and responds to bone morphogenic proteins (BMPs) (reviewed in [50]). BMPs also play an important role in maintenance and speci cation of human breast stem cells [51]. Consistent with this, we found that inhibition of the TGF-b signaling cascade by pre-incubation with SB431542 speci cally in the lobular-like broblasts led to attenuated interaction with epithelial progenitors in the subsequent co-culture experiment. Although still not completely elucidated, it appears that such TGF-b dependent epithelial-stromal interaction is crucial also for cancer development. While TGF-b1 converts the majority of normal breast broblasts to alpha-smooth muscle actin-positive myo broblasts [15], disrupted TGF-b-signaling attenuates CAF-induced cancer cell growth [52].
We show here that while lobular-like broblasts in many respects are similar to human bone marrow-derived MSCs, they fail in an ultimate in vivo test gauging for bone formation. Thus, as far as the human breast is concerned we can now distinguish resident broblasts from bona de MSCs. This is important because the latter has been implicated in reactive stroma formation such as that occurring in cancer. Thus, it has been speculated that MSCs are recruited to the breast as a source of myo broblasts or CAFs responsible for important aspects of tumor cell-stroma interaction including promotion of metastasis ( [53][54][55], reviewed in [56]). With the in vivo bone formation assay employed here, the question of a "third" immigrant mesenchymal lineage in breast pathology can be addressed also in a human context. Such investigations are ongoing in our laboratory.
The broblast heterogeneity described herein is likely to be in operation in a wider variety of tissues and organs. In the present study, we demonstrate by genome wide gene expression pro ling that lobular-like and interlobular-like broblasts differ by entire lineage programs with characteristics and functions in common with previously reported papillary and reticular broblasts, respectively, in mice and humans [1,14,39]. In this regard, it is interesting that CD26broblasts in mice segregate into mature CD26 + papillary broblasts [1] and that in both mice and humans such broblasts are responsible for ECM production and in turn brosis [57][58][59].
We propose that the CD26 + interlobular-like broblasts are responsible for the dense brous tissue of the breast and further responsible for the differences in breast density between individuals -a known risk factor for development of breast cancer. This would concur with the observation that another marker, CD36, expressed by lobular broblasts, is repressed in high density breast stroma [33]. CD105 + lobular-like broblasts on the other hand have properties in common with bone marrow-derived MSCs and reticular broblast progenitors, which participate in wound healing and myo broblast generation [1]. Previous results from our laboratory have shown that lobular broblasts readily generate a-smooth muscle actin-positive myo broblasts [10] and that interlobular broblasts exhibit an immune related gene expression pro le [10]. Whether the breast cancer repertoire of CAFs is a caricature and maybe even a reminiscence of the normal stromal cell heterogeneity remains an open question. Interestingly, however, recent single-cell RNA sequencing of breast carcinomas has resolved stromal cell diversity to include both myo broblastic and in ammatory CAFs and not least perivascular cells [60,61]. While this concurs with our early studies, which suggested diverse cellular origins of CAFs, including resident broblasts and perivascular cells [62,63], our present ndings suggest that lineage heterogeneity within the resident broblast compartment adds to the complexity. If indeed this is the case and myo broblast and in ammatory classi cation operate among both CAFs and normal resident broblasts, it is tempting to speculate on lineage interrelationships and how these may be taken advantage of in a clinical setting [64]. The cell lines established in the present study may prove valuable in determining such lineage relationships. It is also a possibility that phenotypic and functional CAF heterogeneity re ects plasticity in a broader sense, and in general may be governed more or less by the tumor genotype as suggested in mouse models of pancreatic cancer (reviewed in [65]). In this context, the cell lines may serve to decipher whether a speci c tumor genotype instructs development of a particular stromal response independent of recipient initial stromal cell type.
For these reasons, it is likely that both lobular and interlobular-like broblasts play important albeit different roles in normal breast as well as in breast cancer.

Conclusion
Collectively, our study shows that we have established two physiologically relevant, phenotypically distinct human breast broblast cell lines, which exhibit specialized functions in maintenance of region-speci c characteristics and regulation of neighboring epithelial cells. In the longer perspective, the present developments may provide a basis for the experimentation in cell-based assays to elucidate the earliest events in human breast cancer evolution.

Declarations
Ethics approval and consent to participate     (grey)). Note that the myoepithelial phenotype shifts as a consequence of a switch between broblasts.

Figure 5
Page 27/28 The luminal differentiation repertoire of myoepithelial progenitors is directed by interaction with specialized broblasts. (a) Comparison of capacity of broblasts to direct epithelial progenitor capacity. Myoepithelial cells co-cultured with iHBFCCD105 or iHBFCCD26 were passaged and subjected to luminal differentiation conditions at clonal density and peroxidase stained for K19. While the induced K19 appeared mainly scattered when derived from iHBFCCD105 co-culture (left), additional rather homogenous islets presented from iHBFCCD26 co-cultures (right). The distinct phenotypes were observed in ve out of seven tests with absence of homogeneous islets from iHBFCCD26 in two tests, (bar = 500 μm). (b) Representative multicolor confocal images (K19, red; K14, green; Nuclei, blue) of cryostat sections of xenografted NOG mice 8 weeks after orthotopic injection of myoepithelial cells from primary co-culture with iHBFCCD105 or iHBFCCD26. Bilayered epithelial structures were obtained in 6/10 and 5/8 injections from iHBFCCD105 and iHBFCCD26, respectively, although at limited numbers, down to a few per transplant. Whereas iHBFCCD105 co-culture derived myoepithelial cells readily differentiated into luminal K14-/low/K19+ cells, co-culture with iHBFCCD26 resulted mainly in K14+/K19+ luminal cells. (bar = 50 μm).