Positional Variations in Mammary Gland Development and Cancer

What determines the number and position of the individual pairs of mammary glands? The 2:1 correlation between the number of mammary glands and average litter size in most species [6] and co-incidence of supernumerary nipples with more frequent twinning as reported for isolated groups of sheep and macaques [7, 8] suggests considerable genetic plasticity that is molded by evolutionary pressure. Indeed, supernumerary breasts (polymastia) or nipples (polythelia) are an inherited trait in some humans [9], and pigs and sheep can be selectively bred for increased teat number [7, 10], indicating the involvement of a genetic component. Accordingly, several genetically modified mice have been identified with mammary induction defects leading to numerical aberrancies (Table 1). They can be organized into four categories depending on the subset of MRs affected, and loss or gain of MRs.

The various mesenchymal inductive mechanisms seem to converge on shared epithelial regulators, whose extent of activation appears to regulate the number of induced MRs. It is conceivable then that the variable inductive signals bestow the MR cells with unique properties. Indeed, preliminary gene profiling data suggest that each of the five paired mammary placode epithelia express unique subsets of genes (Sun and Veltmaat, http://​www.​veltmaatlab.​net/​research.​html). While the functional relevance of these genes remains to be determined, mammary development proceeds differently in all MRs, as evidenced by the different growth rate and morphogenesis of the MRs [13], culminating in rudimentary glandular trees before birth that differ in size and number of branches (Fig. 2). These differences may very well underlie some of the differential attributes of adult MGs (discussed below).

Why do mesodermal genes act differently at different positions along the mammary line? The answer may be found in their expression pattern, which is in part reflected in the temporal development of the murine mammary line. This line is actually composed of three initially unconnected streaks of Wnt10b expression [3]. As concluded from the phenotypes summarized in Table 1, these streaks form independently of each other.

The first streak forms on the flank, between the limbs. Its initially fragmented Wnt10b expression pattern in wild-type embryos (Fig. 1) and dorsalized position upon hypaxial truncation of the somites, as in Pax3 null mutants, indicate somitic involvement in the formation of this streak [14]. Fgf10 is expressed highly in the hypaxial dermomyotomal tips of somites #12–18, which underlie the region where MR2/7 and MR3/8 form [14, 15]. FGF10 activates its main receptor FGFR2-IIIb, expressed throughout the overlying surface ectoderm. Partial reduction of somitic Fgf10 expression, as in Gli3 ^Xt-J/Xt-J null or hypomorphic Fgf10 ^mlcv/- mutants, allows formation of a mammary line and all MRs except MR3/8, indicating that at least this pair requires somitic Fgf10 and, contrary to MR2/7, is susceptible to reduced levels of Fgf10 [14]. Fgf10 is not expressed in somites 19–24, and neither Wnt10b expression in this region nor induction of MR4/9 (above #24 somites) requires FGF10/FGFR2b signaling. These observations suggest that the anterior and posterior regions of this streak are two separate entities which develop by independent mechanisms, and that another yet to be identified somitic molecule(s) may be responsible for the initiation of mammogenesis in the posterior region, at the posterior edge of which arises MR4 or MR9. Interestingly, the boundary between the Fgf10-dependent and Fgf10-independent region at the level of #18 somites is home to a small mesenchymal Nrg3 expression domain, which seems a prerequisite for normal formation of MR3/8 [16].

The two other mammary streaks are located in the axilla and inguen, where they give rise to MR1/6 and MR5/10, respectively. They gradually elongate dorsally until they connect to the first streak on the flank [3]. Defects in these streaks or their MRs often coincide with defects in limb induction or specification in the absence of limb mesenchymal factors like Tbx2, Tbx3, Gli3 and Fgf10 [13, 14, 17, 18]. It remains to be determined whether MR1/6 and MR5/10 require such factors or the limb mesenchyme directly, or indirectly via formation of the limbs as new signaling centers. Perhaps MR2/7, that form at the junction of the first and axillary streak, and depend less than MR3/8 on somitic signals [14], similarly benefit from signals that originate in the limb. In conclusion, MRs develop due to in part variable mesodermal inductive signals along the AP-axis that converge on common ectodermal executive pathways. Thus, the different glands experience unique developmental histories and, furthermore, are embedded and surrounded by tissue that exhibits region-specific properties.

The embryonic flanks develop in the trunk as two broad lateral regions where the dorsal paraxial mesoderm intermingles with the ventral lateral plate mesoderm (LPM). They establish new, ventro-lateral boundaries with the ventrum, which serve as bilateral platforms for mammary line formation. Externally, these presumptive mammary forming regions look left-right (LR) symmetric. But looks can be deceiving! LR asymmetries during mammogenesis become evident at the level of individual mammary gland pairs: The size and branching pattern differ between contralateral glands; one counterpart of a normal or supernumerary pair may be missing or shifted from its symmetric position, shown for mouse in Figs. 2 and 3. In humans with Poland syndrome, one breast (most often the right) may remain hypoplastic [19]. Interestingly, formation of the mammary line and MRs is slightly advanced on the left side (Fig. 3) and the left side seems to have a greater mammogenic potential. For example, unilateral polythelia or polymastia occurs most frequently on the left side in humans [20] and mice [21].

The basis for LR asymmetric mammary gland development is likely rooted in embryonic LR patterning: During vertebrate embryogenesis, the LR axis establishes the positions of the heart and visceral organs and promotes structural asymmetries that are necessary for their functions. The driving force for LR patterning is unilateral activity of molecules such as the secreted factor nodal, which is expressed by the left half of the node, left-sided LPM and overlying ectoderm [22]. In the left-sided LPM, nodal initiates the expression of the transcription factor Pitx2, an essential “interpreter” of LR axial patterning during organogenesis [reviewed by 23]. Thus, as ectodermal appendages that are induced at the ventro-lateral boundaries, i.e. at the fronts of somitic penetration into two molecularly different regions of left and right LPM [14], the left and right mammary glands have distinct molecular “histories” from their inception!

Moreover, and despite their morphological bilateral symmetry, the left and right somites also differ from each other with significant LR differences in gene expression [24, 25]. Of interest is perhaps heparin-binding EGF-like growth factor [25], which is transiently elevated on the left side, temporally overlapping with formation of MR3/8. Furthermore, retinoic acid (RA) signaling promotes LR asymmetry in the somites, as demonstrated by one-sided delay in somite formation in chick and zebrafish embryos, caused by pharmacological inhibition of RA production [26, 27]. Several modulators of RA signaling and synthesis, including Raldh2, are expressed in the somites proper, and the RA receptor RARβ is expressed in the flank mesenchyme underlying MR2/7, 3/8, 4/9 [28]. Notably, perturbation of RA-signaling affects MR3/8 formation ex vivo [28]. Unfortunately, laterality in effects of RA signaling on mammary gland development was not tested. This will be difficult to do, due to lethality of RA-deprived mouse models prior to mammary gland formation, and due to artifacts introduced during MR development ex vivo under otherwise normal conditions, but especially under experimental conditions such as pharmological modulation of RA-signaling. However, it is of interest that RA-deficient Raldh2 ^−/− mouse embryos exhibit transiently delayed right-sided development and segmentation of the paraxial mesoderm into somites #8–15 [29]. With MR2/7 and MR3/8 forming above the #12 and #15 somites respectively, it is thus conceivable that RA signaling may be a pivotal link between somitic laterality and asynchrony and asymmetry in development of LR counterparts of particularly these MR pairs. Moreover, the already known requirement of somitic signals for induction of MR3/8 [14], provides strong support for the hypothesis that this mammary gland pair—despite its seemingly morphological symmetry - may be lateralized, either directly or indirectly via inductive interaction with lateralized aspects of somites, including effects of RA-signaling.

Finally, the curvature of the AP-body axis may also be involved in promoting subtle asymmetry of mammogenesis: Most often the AP-axis is curved to the right (tail over right shoulder), leading to a ‘compressed’ right flank and ‘stretched’ left flank. If molecular gradients are involved in early mammary gland formation, which is the case at least for somitic Fgf10 [14], their shape may be different between left and right flanks, which could possibly explain the slight LR asynchrony in length of the mammary lines and mis-alignment between paired glands that is sometimes observed even in wild-type embryos (Fig. 3).

A striking but still unexplained feature of breast cancer is a higher tumor incidence in the left breast. Although modest with a ~1.10 left/right incidence ratio, such laterality was consistently found in several epidemiological studies [30‐36] regardless of gender and stage of disease. Even in bilateral breast cancer cases, the left breast is more frequently affected first, or with a larger tumor [37].

Survival rates also appear to reflect LR differences: A breast cancer study among 590 Russian patients reported a statistically significant higher survival rate of patients with right-sided, advanced stage disease [38]. In contrast, a study of 163 Norwegian patients associated longer survival with left-sided disease, but did not relate this finding to disease stage [39]. In a third study of 10,702 Israeli women, no laterality was detected for overall patient survival from years 1–15 [40], but trends of a higher 5–10 year survival rate of left-sided later stage disease, and a higher 10–15 year survival rate of right-sided early stage disease, call for statistical analysis of the data within such smaller time periods, which had not been done. The most recent study of 384 Pakistani breast cancer patients indicates that although left-sided tumors are larger and more prevalent, when standardized for tumor size, right-sided tumors are more prone to metastasis and are associated with lower survival rates [36]. Although the results of these studies appear discordant with each other, they certainly attest to the need for well-defined parameters to better assess laterality. Furthermore, these available data contest, at least in part, the assumption that asymmetry in survival may be due to radiation-induced heart damage, which is greater with left-side targeted therapy. Instead, the data suggest that the LR differences may be due to lateralized tumor biology, similar to LR asymmetry in incidence and patient survival observed with cancers of other paired organs, including kidney cancer, which does not entail cardiac complications following side-specific radiation therapy [33, 41].

Why do tumors form with unequal probability and perhaps outcome in left versus right members of organ pairs? A variety of epidemiological correlates have been investigated, including age, race, ethnicity, sex, breast size, stage of disease at diagnosis, histological tumor type, and even left- versus right-handedness. None of these factors emerged as a consistently significant variable for the increased left-sided incidence [30‐32, 37]. By contrast, laterality in outcome may in part be explained by laterality of axillary lymph node metastasis. A higher mean number of lymph nodes with metastases in breast cancer patients has been reported for right-sided tumors, consistent with more lymph nodes present on the right side [43, 44]. However, when axillary lymph node involvement was analyzed with respect to the total number of affected nodes, the right-sided increase was present only in patients with 1–3 positive nodes [45]. For patients with >3 positive nodes (a characteristic of locally advanced or higher stage disease) left-sided tumors were associated with more affected nodes despite the LR asymmetry in overall node number [45]. Thus, metastasis to lymph nodes seems to be a lateralized disease feature, and may indicate LR differences in tumor biology, e.g. greater metastatic potential of left-sided tumors, or lateralized features of the lymph nodes themselves. In general, attempts to identify lateralized features may be confounded by the profound heterogeneity of breast cancer and could possibly be improved by inclusion of more detailed cancer subtype-specific information such as genomic classifiers. Indeed, an interesting correlate in laterality of breast tumors is the expression of Her-2 and perhaps also hormone receptors, although the sidedness may co-vary with other parameters: Among 166 Kuwaiti patients spanning all disease stages at diagnosis, right-sided tumors are significantly associated with amplified Her-2 expression, without lateralized differences for hormone receptor expression [46], whereas among 384 Pakistani breast cancer patients with bone metastasis, right-sided tumors were markedly less prone to express Her-2, estrogen receptor, or progesterone receptor, either individually or in combination [36].

It is tempting to speculate that the LR differences in tumor incidence and patient survival may be related to expression of genes that govern LR axial patterning during embryonic development, like Nodal and Pitx2 [reviewed by 47]. Increased expression of Nodal and methylation of Pitx2 DNA correlates with poor prognostic outcome in breast cancer patients [48‐50]. Overexpression of another nodal pathway component, the co-receptor Cripto-1, is also associated with decreased patient survival [reviewed by 49]. Consistent with these findings, nodal pathway signaling promotes an invasive phenotype in breast cancer cells and induces tumor vascularization, which has implications for disease progression and relapse [51]. Whether nodal or nodal pathway members are expressed asymmetrically during mammary gland development remains to be investigated, as does the question of whether such putative LR differences are recapitulated in breast tumors.

The mouse is widely used as a breast cancer model. The axial position of the human breast approximates the 5th thoracic somite or its derived rib. Taking into account the differences in somite number between mouse and human, this area corresponds best to the area between MG2/7 and 3/8 in the mouse. However, the inguinal mouse MG4/9 are generally used for molecular and histological analyses. Does it matter? As described above, the five pairs of mammary glands in the mouse are each induced by unique inductive signaling networks. Yet surprisingly little information exists on differences between the five pairs of glands in adult mice, both at the level of normal physiology and with respect to response to tumor-initiating events. However, several studies report significantly differential tumor incidence or biology depending on the position along the AP axis. So the answer is almost certainly: Yes!

Examples for AP variations in tumor-incidence and biology come from endogenous tumor agents, chemical carcinogenesis, various treatments, and transgenic approaches. Integration of mouse mammary tumor viruses (MMTV) in the C3H mouse strain causes mammary tumors most often in thoracic glands. While tumor incidences are equal among the gland pairs after correction for total wet tissue weight [52], the relative ratios of epithelial to stromal/adipose tissue may not correspond to wet weights and could confound the analysis. Indeed, increased incidence of DMBA-induced tumors in the thoracic mammary glands of rats was correlated to asynchronous development of the glands. Differences in the number of terminal end buds (TEBs) and alveolar buds per gland results in variable abundance of target cells for chemical carcinogens depending on treatment age [53]. Likewise, murine MG2/7 and MG3/8 exhibit a greater number and larger TEBs than MG4/9 (see also Fig. 2); and MG2/7 specifically undergoes a reduction in TEBs from 5 to 6 weeks of age [54]. The observed asynchrony in postnatal development can also result in differential responses to hormones [53]. These AP-specific developmental features are a plausible basis for increased tumor incidence in thoracic glands. Interestingly, exposure to power frequency magnetic fields (MF) increases DMBA-induced mammary tumorigenesis in SD1 rats, particularly of MG1/6 and 2/7 in correlation with more significant MF-induced increases in ornithine decarboxylase activity—a possible indicator for stem cell proliferation [55]. However, in F433 Fisher rats, MF exposure increased the incidence of DMBA-induced adenocarcinoma specifically in the 5th of the six pairs of glands in rats, but latency was reduced most in the 3rd pair [56]. Thus, region-specific effects on tumorigenesis may vary even between strains of the same species.

Transgenic MMTV-Polyoma middle T antigen (PyMT) causes the largest tumor burden with shortest latency in MG1/6 of C57BL/6 mice [57]. Tumor latency is more similar among all other glands, with MG4/9 harboring the second largest tumor burden. Introduction of an iNOS (Nos2) null mutation increases tumor latency and reduces tumor burden most in MG4/9 and least in MG1/6 [57]. Overexpressed Erbb2 in MMTV-cNeu mice [58] induces tumors preferentially in MG2/7 + 3/8 (Table 2) and tumor incidence is comparatively low in MG4/9/5/10. Transgenic C3(1)-SV40-T/t-antigen causes tumor formation first in MG1/6 and MG2/7 with the highest total tumor burden in MG2/7 [59]. Differential activity of transgene promoters is one possible explanation for variations in transgene-initiated mammary tumorigenesis. Even then, such mechanism(s) would serve to highlight the significance of biological differences between pairs of glands. What is clear from these data, though, is that these transgenic models generate most tumors in specific anterior mammary glands, raising the question if other transgenic models exhibit similar or other unique gland-specific incidences along the AP-axis.

Transplantation experiments also provide clues to the existence of AP-axial effects both in the tumor cells and environment. PyMT tumor cells isolated from anterior tumors proliferate faster than cells from posterior tumors, independent of implantation site. Significant differences in gene expression exist between anterior and posterior tumors despite equivalent histology, and the difference in proliferative potential was maintained over transplantation passages [60]. Moreover, the higher take rate of teratocarcinoma cell grafts at lumbar subcutaneous sites despite their faster growth rate at thoracic sites [61] provides evidence for unique positional features. Likewise, MDA-MB-435 tumor cell xenografts in thoracic MFPs grow better in response to high fat diet but not when in inguinal MFPs [62]. Thus, along the AP-axis, both the mammary epithelium and stroma harbor positional variations that influence tumor growth.

Finally, modality of disease progression may also depend on positional differences: Tumors of C3HxVY-Avy mice are more likely to metastasize from thoracic glands than from abdominal glands [63]. MDA-MB-435 cell implants into MFPs of athymic nude mice showed the same tendency and moreover, thoracic implants spread via the vascular system, while inguinal implants spread via the lymphatic system [62].

What could be the cause of these positional variations? A comprehensive review article on wound healing, immune responsiveness, and mammary tumorigenesis in mice and rats, concluded that not only tumor growth, but also wound healing responses are more pronounced in anterior regions [64]. The prominent role of inflammatory signaling in breast cancer suggests that this AP-gradient may influence murine mammary tumor biology. The AP-gradient of the nervous system and vascularization, resulting in reduced blood flow in the posterior body, provides perhaps the likeliest explanation for the region-specific variations in tumor biology [64]. Such a gradient has also been confirmed for blood flow in pectoral versus inguinal mammary glands [65]. Consequently, one can hypothesize that posterior glands exhibit reduced oxygen tension and lower concentrations of systemic factors, which may modulate tumor incidence. Alternatively, varying relative positions of lymph nodes may alter lymph drainage of the tissue and consequently tumor biology.

The role of embryonic signaling pathways in promoting cancer cell stemness and metastasis [66] may suggest that differential intrinsic features of tumor cells and their microenvironment could possibly be traced back to the differential developmental signals along the mammary line of the mouse. For example, by interpretation of regional cues, Hox transcription factors lay down positional memory that results in region-of-origin specific cell biology, as shown for dermal fibroblasts [67]. By analogy, mammary stroma may secrete unique factors depending on their position along the AP axis. Aberrant Hox gene activity has also been implicated in breast cancer [68] and in altered Wnt signaling [69, 70]. Given the prominent role of the Wnt pathway in embryonic MG development and in breast tumor biology, it is conceivable that the Hox and Wnt pathways affect tumors differentially along the AP axis.