Postlactational involution biomarkers plasminogen and phospho-STAT3 are linked with active age-related lobular involution

Terminal duct lobular units (TDLUs) are microscopic structures comprising terminal ducts and acinar substructures that produce milk after birth and may develop cancer precursors among some women. With aging, TDLUs undergo age-related lobular involution (LI), such that the number and size of acini per TDLU are reduced and the intralobular stroma is replaced with collagen [1]. Analysis of the Mayo Clinic Benign Breast Disease (BBD) cohort of more than 13,000 women who underwent a surgical biopsy diagnosed as BBD showed that the timing of LI centers around the perimenopausal years, although there was considerable variation among women [2], and that incomplete LI among postmenopausal women or delays in the ongoing process of LI are associated with increased breast cancer (BC) risk [2, 3]. Such observations indicate that investigations to understand the biological processes underlying LI may reveal markers and mechanisms associated with lowering of breast cancer risk. However, very little is known about the signaling processes that control LI or why ongoing LI is associated with decreased BC risk.

Postlactational involution, which is triggered by weaning, is a distinct process from LI, involving a highly controlled collapse of alveolar structures, programmed removal of secretory epithelial cells, phagocytosis by macrophages, proteolytic degradation of basement membranes, and stromal remodeling [4]. As a result, most of the differentiated epithelial cells disappear and an adipocyte-rich stroma, in which the resting ductal system is embedded, reappears [5‐9]. During postlactational remodeling, the mammary gland shares striking similarities with pathologically induced wound-healing and tumorigenic microenvironments, and the postlactational state is associated with increased cancer progression [10‐12]. Moreover, experimental studies using animal models have shown that delays in postlactational lobular involution are associated with increased cancer incidence and progression [13].

Thus, since increased cancer incidence and progression have been associated with delays in age-related LI as well as delays in postlactational LI, we set out to determine whether mediators and effectors of postlactational remodeling are associated with the physiological process of age-related LI. We used tissue biopsies from the Mayo BBD multiple biopsy cohort, a group of women who had multiple biopsies with benign findings and for which LI status has been determined at initial and subsequent biopsy [3]. We compared two groups composed of women who showed LI progression between the initial to the subsequent biopsy, and those who did not. We stained sequential sections, using tissue from the initial benign biopsy, for a range of biomarkers that have been implicated in postlactational involution, including pSTAT3 and c-Myc, key modulators of apoptosis in the early stages of postlactational involution [7, 14, 15], the proliferation marker Ki67, which shows increased expression when postlactational involution is delayed [16]. We also evaluated expression of the matrix metalloprotease MMP9, which is activated during postlactational involution to induce remodeling of the extracellular matrix (ECM) [17‐19], and the preprotease plasminogen, which is required for postlactational involution [20, 21]. We measured expression of ECM component Tenascin C and ECM receptor CD44, expressed during postlactational involution [22, 23] and markers of luminal (CK19) and myoepithelial (CK14) cells, which can show altered abundance during and after postlactational involution [24]. We assessed staining of all markers in individual TDLUs using unbiased digital quantification methods, and analyzed relative expression of each marker and its association with LI progression. Our analyses of these results revealed for the first time specific molecules that correlated with age-related LI status, and implicated functional pathways that may underlie both postlactational and LI involution processes.

While the biologic mechanisms underlying age-related LI remain unknown, the processes that control postlactational involution have been extensively studied, primarily through investigations in animal models [4]. Postlactational involution proceeds through an initial, reversible stage in which there is widespread apoptotic cell death, followed by an irreversible second stage in which the mammary gland is remodeled to the pre-pregnant state [4]. This process is regulated by complex signaling pathways, including cytokine signaling, apoptosis, inflammatory process, and STAT proteins [14]. Much of the knowledge about the mediators of postlactational involution are derived from animal investigations; while mice do display regression of ductal complexity with age [13], there have been no studies to date to determine whether there are differences in the age-associated regression rate between different mice (or different strains of mice) and, if these differences are present, whether delays in age-associated mammary gland regression in mice may be associated with increased risk of spontaneous mammary cancer (or susceptibility to oncogene- or chemical-induced mammary cancers). In this study, we evaluated expression of a select set of these signaling molecules in TDLUs of clinical BBD biopsies that were judged as either progressive or non-progressive in their LI status to identify any shared pathways that might exist between age-related LI and postlactational involution. We assessed staining using a common digital analysis method that provided consistent application of the objectively defined quantification metrics across the whole sample set. Our analyses collapsed multiple per-TDLU expression values per woman into a single individual-specific summary measure based on the maximum of all values. Two main findings emerged: (1) progression was associated with lower expression of pSTAT3; and (2) progression was associated with higher expression of plasminogen. These findings present the first major advancement in understanding the biological mechanism of age-related LI.

Factors identified as regulators of the two-phase process of postlactational mammary gland involution include transforming growth factor-β3 during the first phase of involution [31] and the IL-6 cytokine family members leukemia inhibitory factor (LIF) in the first phase [32, 33] and oncostatin M (OSM) during the second phase [34]. All of these cytokines activate STAT3, which is essential for the initiation of apoptosis and remodeling following forced weaning [35]; the absence of STAT3 leads to delayed postlactational involution [35, 36]. Conversely, deletion of suppressor of cytokine signaling 3 (SOCS3), which suppresses STAT3 expression, results in premature involution, and increased activation of c-MYC and its pro-apoptotic effectors E2F1, BAX, and p53 [36, 37]. Thus, there is ample evidence for the involvement of the apoptotic cell death machinery downstream of STAT3 activation for postlactational involution. This presents a contrast with our observations for pSTAT3 in age-related LI, as we found that levels of phosphorylated STAT3 are decreased in the tissue of women who will undergo LI relative to the tissue of women with non-progressive LI. One possibility is that the STAT3-activated apoptotic signaling pathway is disrupted in women who do not progress in LI, and levels of phosphorylated STAT3 accumulate as the factors regulating STAT3 continue to signal activation; additional analysis of downstream signaling effectors could provide insight into this possibility. It is also possible that pSTAT3 may regulate other non-apoptotic pathways, which may be present in the aging mammary gland and which may impact LI progression.

The plasminogen cascade of serine proteases is involved in many different processes in many different tissues, primarily due to the ability of this system to regulate pericellular proteolytic activity [38], and dysregulation of this system results in tumor growth and metastasis formation [20]. The involvement of plasminogen in postlactational involution became clear when plasminogen-deficient mice were found to be defective in postlactational involution [21]. In plasminogen knockout mice, the reduction in overall mammary gland size after 5 days of involution was less than one-third the reduction seen in wild-type mice, and the reduction in secretory alveolar volume was almost one-thirtieth the reduction seen in wild-type mice [21]. Although the effect of plasminogen deficiency on postlactational involution is evident, the potential mechanism by which plasminogen loss leads to abnormal postlactational involution is unknown. One possibility that has been advanced is that plasminogen loss results in a failure to sense apoptotic signals [21], which is consistent with the role for plasmin, the activated form of plasminogen, in a variety of apoptotic mechanisms [39], as well as the decreased apoptosis observed in involuting mammary glands in plasminogen knockout mice. It may be that plasminogen is also needed for apoptosis in age-related LI.

An interesting finding in our study is the fact that associations of pSTAT3 and plasminogen with progressive LI strengthen after adjustment for each other in logistic regression models. This is due to the fact that pSTAT3 and plasminogen are positively associated with each other (r = 0.37, Fig. 3), and that each is associated with progressive LI (pSTAT3 negatively and plasminogen positively). Because of this complex relationship, failure to account for the collinearity of the two biomarkers in regression models will result in a spurious dilution of association of each with LI progression. This underscores the importance of accounting for inter-relationships between biomarkers when assessing their independent effects on clinical outcomes.

Strengths of the study include the unique tissue resource, which allowed us to define LI progression status in individual women through evaluations of sequential separate biopsies, and to assess correlations of these rates with biomarker expression. Limitations of our findings include evaluation of only a limited number of the many potential pathways implicated in postlactational involution. Immune cells have been implicated as playing critical roles in the postnatal mammary gland development [40]; in particular, pro-inflammatory macrophages have been implicated in the increased breast cancer risk during postlactational involution [41‐44], and assessment of immune markers expression and localization vs LI progression status is an important future area of investigation. Additional limitations include potential losses of information and a resulting decrease in statistical power from using a single individual-specific summary measure based on the maximum of all values for a given patient. We had initially intended to fit models that included all TDLU-specific expression values and account for intra-individual collinearity using a generalized estimating equation (GEE) approach within the logistic regression framework. We attempted this using a number of different GEE correlation structures, ranging from a simple exchangeable pattern to an unstructured pattern. Model likelihoods and individual parameter estimates failed to converge in each instance, requiring us to resort to the summary values described herein. Also, the use of the maximum expression biomarker value across all lobules could have introduced a bias, such that women with more lobules would have a greater chance to have a higher maximum value than those with fewer lobules. We mitigated this somewhat by examining no more than 15 lobules per sample. Also, associations using median values, which would not introduce any such bias, were similar in direction to those with maximum values. Finally, all biomarkers are subject to intra-lesion heterogeneity, much like histology. Using the maximum biomarker value across the lesion is similar in rationale to summarizing the benign histology based on the most extreme histologic characteristic, be it non-proliferative disease, proliferative disease without atypia, or atypical hyperplasia. Other limitations include the use of overall staining intensity as our metric, as this may mask critical location-dependent information for some biomarkers; as a consequence, the fact that we did not find significant associations for some of the assessed biomarkers with LI progression does not indicate that no association exists. Finally, none of the p values presented in the manuscript account for multiple testing. It is possible, even though the biomarkers are correlated and thus the number of independent tests is smaller than the total number of tests, that accounting for multiple testing would have resulted in non-significant associations. Because of our modest sample size and the fact that this study was somewhat discovery in nature, we chose to present uncorrected p values. Certainly, our associations will need to be replicated by others to confirm results.

In summary, we have identified for the first time molecules that may play a role in mediating age-related LI. Further studies of each molecule, both in human tissue and mouse models, will evaluate their validity and begin to address the biological mechanism of age-related LI. Understanding the biological mechanism of age-related LI is likely to provide important insights into why the process of LI is delayed in approximately 40% of postmenopausal women [2], a group with increased risk for breast cancer compared with women who do not have delayed LI.