Introduction

Cells in the mammary gland undergo successive rounds of proliferation and cell death during pregnancy and involution [1]. During each estrus cycle an expansion of the epithelium generates tertiary branches and lobuloalveolar structures that then regress if pregnancy does not take place. Pregnancy results in rapid proliferation of epithelial cells mediated by progesterone stimulation of the TNFR-family member, Receptor Activator of NF-κB (RANK), to populate lobuloalveolar units that terminally differentiate near parturition to produce milk [2,3,4]. At weaning the gland undergoes the process of involution that takes place in two phases [5]. The first phase occurs within the first 2–3 days in mice and is thought to be a response to milk accumulation and resulting alveolar stretch [6]. This phase is reversible and involves extensive apoptosis combined with macroautophagy of the alveolar epithelial cells. In the second phase, lasting 6–8 days in mice, an irreversible structural remodeling of the mammary gland occurs [5, 6].

The first phase of involution is regulated by soluble factors belonging to the tumor necrosis factor (TNF) family including TNFα and the TNF-related weak inducer of apoptosis (Tweak) [7]. Leukemia inhibitory factor (LIF) is then secreted activating a Stat3-regulated cell death program [8]. The control of apoptosis and signaling through the TNFR-related receptors requires the cellular inhibitor of apoptosis proteins (cIAPs), cIAP1 and cIAP2, which possess E3 ubiquitin ligase activity. cIAP1 is often constitutively expressed and promotes the ubiquitin-mediated degradation of cIAP2 [9]. Thus, cIAP2 levels are typically low but can be induced by cellular stress signaling [10]. The cIAPs are required for the induction of canonical NF-κB (p65/RelA) following (TNFα) stimulation of the TNF receptor-1 (TNFR1) [9, 11, 12]. Consequently, loss of cIAP1/2 leads to the formation of a caspase 8-activating complex and induction of apoptosis by TNFα [11,12,13,14,15]. Tweak stimulation of its receptor, Fn14, results in the sequestration of the cIAPs and TNF receptor activating factors (TRAFs) at the plasma membrane, stabilizing the NF-κB inducing kinase thus activating p52/RelB NF-κB [16, 17]. Tweak also promotes the lysosomal degradation of cIAP1-TRAF2 resulting in both autocrine TNFα secretion (induced by NF-κB) and enhanced sensitivity to apoptosis in the presence of TNFα [16], thereby synergizing with TNFα in the induction of cell death. Ultimately Tweak induces the formation of a death-inducing signal complex (DISC) promoting the auto-activation of caspase 8 [17, 18].

Global knockouts for both cIAP1 and cIAP2 have been described. cIAP1 null mice are viable and are not overtly hypersensitive to proapoptotic stimuli, however cIAP1−/− cells express markedly elevated levels of cIAP2 protein due to reduced degradation [19]. Similarly, cIAP2−/− mice are viable and fertile with no obvious phenotype, although macrophages lacking cIAP2 undergo apoptosis following challenge with lipopolysaccharide, a stimulus which normally activates cIAP2 expression in these cells [20]. Compound mutants are embryonic lethal, consistent with critical roles for cIAP1/2 that can be reciprocally compensated by either protein in most aspects of development [19].

Aside from functions in normal physiology, the involvement of the cIAPs in cancer is underscored by the frequent induction of these genes in several tumors [21, 22]. Interestingly, cIAP1/cIAP2 loss is seen in a small percentage of multiple myelomas leading to the activation of the alternative NF-κB signaling pathway in B cells [23, 24]. Given the role of cIAPs in receptor-mediated activation of NF-κB and the fact that NF-κB plays a major role in numerous cancers [25, 26], loss of cIAP1/2 expression has the potential to be associated with either cell death or tumor promotion.

The mammary lactation/involution cycle and tumorigenesis are processes that involve significant cellular stress as well as cytokines that can alter the balance of cIAP proteins. Our results illustrate key cIAP1-independent functions of cIAP2 in mammary gland biology and pathogenesis.

Results

Absence of cIAP2 Results in Reduced Branching Morphogenesis and Early Onset Rapid Involution

To study the role of cIAP2 in mammary development we used the previously characterized cIAP2−/− mouse [20] in which cIAP2 is null in all cell subsets in the mammary gland. Studies have shown that XIAP and cIAP1/2 proteins are expressed at various times during pregnancy, lactation and involution [27, 28]. In agreement with others [27], using qRT-PCR analysis, we found that both cIAP1/2 transcripts were expressed in wild-type (WT) mouse virgin glands, increased during pregnancy (P) and declined during lactation (L) and after weaning (INV) (Fig. 1 a, b). Although levels were not statistically significant, cIAP1 mRNAs trended toward higher levels throughout development in cIAP2−/− glands. Whole mount analysis and H&E-stained sections showed that virgin cIAP2−/− mouse glands appeared to have reduced ductal structures compared with control 9-week old mice. (Fig. 1 c, d). No differences in apoptotic cells were detected in cIAP2−/− and WT virgin mammary tissues (Supplemental Fig. S1). Enumeration of ductal branch points in virgin 9-week old cIAP2−/− mice compared with WT confirmed this observation (Fig. 1e, Supplemental Fig. S2A). Consistent with images in Fig. 1c, lobular areas at d12P were greater in WT mice compared to cIAP2−/− mice (Supplemental Fig. S2B). As pregnancy and lactation progress, reduced branching complexity appears to be compensated by enlargement of alveoli evident at day 18 of pregnancy (d18P) and day 7.5 of lactation (d7.5 L) in cIAP2−/− glands.

Fig. 1: Expression of cIAP1/2 and histological comparison between cIAP2 null and wild-type mammary glands during development.
figure 1

RNA was harvested from mammary glands at different stages of development. A, qRT-PCR of cIAP1 RNA from WT and cIAP2−/− mammary glands at 9 weeks of age (virgin), 12 and 18 days of pregnancy (d12P, d18P), 7.5 days of lactation (d7.5 L) and 24 h post-weaning (24hINV). Data are means from three mouse glands performed in triplicate. Bars are S.E.M. B, qRT-PCR for cIAP2 in WT mammary glands at the indicated times. Data are representative of the triplicate means from three glands. Bars are S.E.M. Transcripts in A and B are normalized to β-actin mRNA. C, Whole mount analysis of 9-week old virgin cIAP2−/− and WT mammary glands at d12P, d18P and d7.5 L. Mammary glands were dissected and stained with hematoxylin as described in Materials and Methods. 10X objective. D, Hematoxylin and eosin-stained paraffin sections of mammary glands from mice at developmental stages as in C. E, Graph showing enumeration of branch points derived from analysis of whole mount 9 week old virgin mammary glands from cIAP2−/− and WT mice (n = 3 mice per genotype). F, Graph depicting the average mass of WT and cIAP2−/− pups at the indicated post-natal days. At least 5 pups were weighed per genotype for each time point. The arrow indicates day of weaning. Bars are S.E.M.

Full size image

Despite these defects, cIAP2 null mice are able to lactate sufficiently to support litters based on similar mean pup weights compared with controls up until weaning (Fig. 1f). Pup weights thereafter were lower in cIAP2−/− mice indicating that cIAP2 may have a role in cell survival on a variety of tissues during rapid post-weaning growth.

Evidence for Premature Cell Detachment during Lactation, Dysfunctional Myoepithelium and Accelerated Involution in cIAP2 −/− Mammary Glands

Our histological findings strongly suggest that as natural weaning approaches, the process of involution initiates prior to cessation of suckling in cIAP2−/− mammary tissue. WT mammary glands showed evidence of phase I involution with reduced alveolar size and luminal cell detachment at 36 h after weaning (36hINV). In contrast cIAP2−/− glands displayed more advanced involution at this time point where ducts and alveoli were devoid of detached cells and alveoli were collapsed (Fig. 2a). Both cIAP2−/− and WT mice produced litter sizes between 5 and 11 pups where 3/5 litter sizes were identical in WT and cIAP2−/− mice. Since natural weaning can occur slightly earlier with smaller litters, we also examined mammary gland histology at 24 h after forced weaning at d10L. A similar acceleration of involution was observed when weaning occurred mid-lactation in cIAP2−/− mice compared to WT mice (Fig. 2b). As early as d7.5 L regions of discontinuous and bulging myoepithelial cells were evident in cIAP2−/− glands detected by smooth muscle actin IHC. This disrupted basal layer was typically associated with detached luminal cells and provides further confirmation that early and rapid involution was not a function of early natural weaning in cIAP2−/− mice. Cell death in early involution occurs through lysosomal cell death independent of classical apoptosis [29]. At 24 h after forced weaning cIAP2−/− glands contained high numbers of detached intraluminal, often binucleated death at 24 h cells which is a feature of lysosomal cell and consistent with the rapid onset of involution. Notably, the latter glands appeared highly similar to cIAP2−/− glands at d19L (Supplemental Fig. S3).

Fig. 2: Histological evidence of precocious involution in cIAP2−/− mammary glands.
figure 2

a, H&E sections taken from WT and cIAP2−/− mammary glands at 36 h after weaning (36hINV). Images were acquired using a 10X and 20X objective as indicated. Note the presence of detached cells in both alveolar and ductal lumens in WT but not cIAP2−/− glands. Images are representative of mammary glands from 3 mice per group. b, Mammary gland sections stained with H&E 24 h after forced weaning of WT and cIAP2−/− dams at d10Lwhich depict the presence of detached cells in the lumen of cIAP2−/− mammary glands and loss of cellularity. Images are representative of 3 mammary glands of each genotype. c, IHC to detect smooth muscle actin (SMA) expressed by myoepithelial cells in paraffin sections from d7.5 L and d19L mice. Sections were counterstained with hematoxylin. Red circles indicate dying cells typical of lysosomal cell death in phase I involution (swollen cells with two hypercondensed nuclei without membrane blebbing). Green arrows indicate luminal epithelial cells and red arrows indicate SMA-positive myoepithelial cells in WT and cIAP2−/− glands. Images are representative of 3 mammary glands at each time point per genotype

Full size image

By d19L cIAP2−/− glands showed increased spacing between nuclei of stretched luminal alveolar cells, numerous detached cells and reduced cellularity compared with WT glands (Fig. 2c and Supplemental Fig. S3). Alveoli consisted of highly elongated spindle-like SMA-positive cells. In some alveolar peripheries both luminal epithelium and SMA-positive cells were undetectable. In contrast the underlying myoepithelial cells were present at both days in WT glands as a nearly continuous layer.

Milk stasis induces mechanical stretch on alveolar cells to trigger early events in involution [8]. Milk accumulation during lactation in cIAP2−/− mice would increase milk content in the gland. Supplemental Fig. S4 shows only a slightly increased content of β-casein in mammary glands lacking cIAP2 relative to control glands at d18P. This suggests that the early failure of myoepithelium and luminal cell detachment in cIAP2−/− glands may be due to stress associated with reduced ductal branching and volume capacitance.

Tweak Expression Is Dysregulated in cIAP2 −/− Mammary Glands Early in Involution

Stat5 phosphorylation occurs in response to prolactin [1] and was normal in cIAP2−/− mammary glands during pregnancy and lactation (Supplemental Fig. S5). The first phase of involution is induced by cytokines including TNFα and Tweak resulting in NF-κB-mediated transcription of LIF which then activates Stat3 [8]. qRT-PCR showed that virgin cIAP2 mice trended toward production of higher levels of TNFα mRNA relative to WT mice (Fig. 3a). However, as indicated in the Methods section, virgin mice were not estrus matched, therefore we cannot rule out that differences in some gene and protein expression across different virgin mice within this study could be influenced by estrus stage. Tweak expression peaks at 12 h following forced involution in normal mice [7] which coincides with milk stasis. Consistent with this we observed an approximate 2-fold increase in Tweak mRNA at 12hINV in WT mice. In contrast, cIAP2−/− mice displayed a significant (3.5-fold) increase in Tweak mRNA at 12hINV that was further elevated at 24hINV (Fig. 3b).

Fig. 3: Absence of cIAP2 results in hyperactivation of Tweak expression and amplified death signaling in vivo.
figure 3

qRT-PCR for a, Tnfα and b, Tweak mRNA transcripts in virgin mice, 48 h prior to weaning (d19L) and 1 day post weaning (24hINV). Values shown are normalized to β-actin and are representative of triplicate values from 3 mice per time point. Bars are S.E.M. * p < .05, Student’s t-test. C and D, Representative immunoblots for cIAP1/2 in virgin and lactating mammary glands. Fifteen μg of protein was immunoblotted with an antibody which detects both cIAP1 and cIAP2 mouse proteins. Vinculin was used as a loading control. Bar graphs beside the blots indicate means ± S.E.M. of densitometric values for cIAP2 relative to actin for 3 mice per time point. * p < .05, Student’s t-test. E and F Immunoblots detecting activation of non-canonical NF-κB indicated by processing of p100 to p52 at the indicated stages of development and comparing virgin and d19L in WT and cIAP2−/− mammary glands respectively. The asterisk below p100 indicates a non-specific immunoreactive band. Numbers below are densitometry for p52/actin and are representative of determinations from 3 mice per time point

Full size image

Activation of the Tweak receptor, Fn14, results in internalization and degradation of cIAP1 within a multi-vesicular body [17]. We analyzed cIAP1/2 protein levels using a commercial antibody that reacts with both mouse cIAP1 and cIAP2. The amino acid sequences of murine cIAP1 and cIAP2 exhibit 69% identity and 81% similarity and have predicted molecular masses of 69.7 and 67.3 kDa respectively but can migrate inversely or be indistinguishable on SDS-PAGE [30] (and supplemental references). Notably, we detected increasing expression of cIAPs in WT glands through pregnancy and lactation which persisted until 12hINV but were then profoundly reduced by 36hINV. While cIAP1 was also increased up to d18P in cIAP2−/− mammary glands tissues, unlike WT mice, the levels declined by d7.5 L and did not increase until 36hINV (Fig. 3c). The reduction in cIAP protein beginning at d7.5 L in cIAP2−/− mammary glands combined with the morphological evidence of engorgement at this time point is suggestive of stress-induced reduction of cIAP1. The continued expression of cIAP1/2 until 12hINV in WT glands contends that cIAP1/2 protein expression is normally maintained at least at early stages of cellular stress associated with milk stasis.

The onset of involution results in the upregulation of cell death receptor components [7]. Traf2 is an integral component of TNFR superfamily signaling. Tweak also induces cathepsin-mediated lysosomal proteolysis of Traf2 [17]. Traf2 expression increased at d19L in cIAP2−/− mice (Supplemental Fig. S6). Consistent with high Tweak levels, by 24hINV Traf2 protein was undetectable.

Both cIAP1 and cIAP2 are critical components of TNFR family signaling controlling the activation of both canonical and alternative NF-κB activation (p52) by subsets of this receptor family involved in lobuloalveolar expansion [31] and involution [32]. Analysis of p52 expression in developing mammary glands during pregnancy and early lactation showed no differences between cIAP2−/− and WT glands (Fig. 3e) although cIAP2−/− virgin glands expressed a constitutive level of p52. During involution, activation of Fn14 by Tweak is predicted to activate IKKα and processing of p100 to p52 [16]. Indeed, p52 was present in d19L cIAP2−/− mammary glands but absent in WT glands at this time point (Fig. 3f). Thus cIAP2−/− mammary glands undergo accelerated involution in conjunction with persistent elevation of Tweak expression and a reduction in cIAP1 protein levels beginning at mid-lactation.

cIAP2 −/− Mammary Glands Demonstrate Persistent Autophagic Flux during Lactation which Precedes Early Anoikis

In response to cell detachment, activation of autophagy normally occurs within 24–48 h of involution and contributes to the reversibility of this phase [33]. Numerous detached cells are found in the lumen of cIAP2−/− mouse alveoli and ducts as early as d7.5 L (see Fig. 2b). Loss of attachment can transiently induce autophagy as a survival mechanism. A gradual increase in the lipidated form of LC3 (LC3-II) in cIAP2−/− tissue detected from d7.5 L to 24hINV but only evident in WT glands at 24hINV (Fig. 4a). Upon detachment or changes in cell shape, epithelial cells lose integrin-mediated survival signaling resulting in a form of programmed cell death termed anoikis [34, 35]. Cell detachment from the basement membrane downregulates EGFR expression thereby reducing Erk1/2 activity [35]. Analysis of P-Erk1/2 in cIAP2−/− and WT mice showed similar activation in pregnant and lactating glands at the days shown with the exception of 24hINV where a strong reduction in P-Erk1/2 can be seen in cIAP2−/− mammary glands that was not a consequence of reduced levels of Erk1/2 (Fig. 4b). Loss of cell adhesion can result in phosphorylation of p38 which can contribute to inhibition of Erk1/2 activity [36]. Consistent with anoikis, immunoblots showed strong induction of P-p38 at d19L in both cIAP2−/− and WT glands which remained expressed at 24hINV in cIAP2−/− tissue (Fig. 4d). Interestingly, a second induction of P-p38 at 36hINVW in both mouse glands that was much stronger in cIAP2−/− tissue consistent with the massive changes in the architecture of the glands.

Fig. 4: Evidence for autophagy and anoikis-associated cell death signaling prior to weaning in cIAP2−/− mammary tissue.
figure 4

a, Detection of LC3-I and lipidated LC3-II in immunoblots of mammary gland lysates during lactation. The presence of both forms after the first week of lactation is apparent in cIAP2−/− glands. b, Representative immunoblots to detect phosphorylated Erk1/2 kinases in virgin glands and at the indicated time points in early and late lactation and after weaning. Twenty μg of mammary gland lysate from WT and cIAP2−/− mammary glands harvested on the indicated days were immunoblotted with anti-P-Erk1/2 and total Erk1/2. The bar graph indicates mean densitometric values from 3 mice per time point for phosphorylated proteins relative to total Erk1/2 protein and gapdh ± S.E.M. c, Immunoblots of P-Akt-1 and total Akt-1 in the late stage of lactation and early involution. The asterisk denotes a high mol wt non-specific band in the cIAP2−/− lane. d, Activation of p38 Mapk (P-p38) relative to total p38 and Bim-EL proteins detected by immunoblot at the indicated time points. e, Bim-EL and P-p38 expression in virgin and 36hINV mouse glands. Representative blots are shown. Numbers below are the means of densitometric values from blots of 3 mice per time point normalized as indicated

Full size image

Another hallmark of anoikis is increased Bim-EL protein expression [37]. In matrix-attached cells activated Erk1/2 phosphorylates Bim, targeting it for proteasomal degradation [38]. Bim is weakly expressed in WT glands at d19L and increases markedly at 24hINV. In contrast, Bim begins to increase at d7.5 L and levels are profoundly increased in cIAP2−/− glands at d19L which only decline slightly by 24hINV (Fig. 4d). Of note, in addition to cell detachment, Bim can be also be regulated through NF-κB-mediated Egr-1 expression [39]. Therefore induction of alternative NF-κB (see Fig. 3g) could also contribute to the large increase in Bim by d19L. Bim is expressed in both virgin and 36hINV cIAP2−/− glands and in WT mice as previously reported [40]. Interestingly, WT and cIAP2−/− glands showed a second increase in P-p38 at 36hINV although this was greater in cIAP2−/− compared to WT tissue (Fig. 4e). Taken together, these results suggest that cIAP2 contributes to the maintenance of the mammary epithelium in the first stages of involution in response to cellular stress and Tweak signaling.

Apoptotic Cell Death in cIAP2 −/− Mammary Glands

Phosphorylated-Stat3 (P-Stat3) is a key regulator of cell death during the first phase of involution where it promotes the uptake of milk fat globules resulting in cell death due to lysosomal membrane permeabilization [41]. P-Stat3 expression in cIAP2−/− and WT mammary glands followed a similar pattern of activation where activity was strongly reduced at d7.5 L relative to pregnancy then robustly increased by 24hINV (Fig. 5a). In 2/4 WT mice we detected P-Stat3 at d19L while all cIAP2−/− mouse tissue contained P-Stat3 at this time (Fig. 5b). Most strikingly, levels of P-Stat3 were consistently higher in cIAP2−/− mice compared to controls. Stat3 activation can block the inhibitory phosphorylation of FOXO3a [42], thereby permitting the transcriptional activation of target genes such as Tweak [42]. Thus the strong activation of P-Stat3 may contribute to the enhanced level of Tweak transcripts detected in cIAP2−/− mammary glands at 24hINV (see Fig. 4b).

Fig. 5: Absence of cIAP2 results in accelerated mammary involution associated with precocious apoptosis.
figure 5

a and b, Representative P-Stat3 and total Stat3 immunoblots of 20 μg of mammary gland lysate from WT and cIAP2−/− mammary glands collected at the indicated times. The bar graphs show mean densitometric quantification of phosphorylated P-Stat3 relative to Stat3 protein and vinculin from 3 mice per time point ± S.E.M. * p < .05 Student’s t-test. Protein extracts from virgin, lactating and weaned glands immunoblotted with anti-caspase8 (c) and anti-caspase-3 (d). The 24 kDa band in C represents activated cleaved (Cl-Caspase8). The lower band in D (Cl-caspase3) indicates activated caspase3. Values below blots in C and D are means of densitometric values from blots of proteins from 3 mice per time point. e, Apoptag labeling of 3’ DNA ends detected using alkaline peroxidase and chromogen to identify apoptotic cells in WT and cIAP2−/− mammary glands at d19L. Full images from a 10X objective and cropped 40X images are shown. The 40X image of a WT gland shows an example of the few apoptotic detached cells detected while the cIAP2−/− gland shows an example of apoptosis in the wall of cells surrounding stretched alveoli (arrows). Images are typical of Apoptag labelling in glands from 3 mice per genotype. f, Graph depicting the percentage of apoptotic cells in d19L WT and cIAP2−/− glands. Mean percentages were calculated based on enumeration of total labelled nuclei from 8 random 40X fields (approximately 600 cells) from 3 mice per genotype. Bars are standard deviation of apoptotic cell counts for each field. ** p < .005, unpaired t-test

Full size image

During normal phase I involution, caspase activation is initially limited to detached cells and only later does it become activated in the luminal structures [32]. Stimulation of cell death by receptors in the TNFR superfamily activates caspase 8. Immunoblots for the pro-form of caspase 8 and cleaved caspase 8 in Fig. 5c show that the activated p24 cleavage product [43] was strongly detected in virgin and d19L cIAP2−/− mouse glands suggestive of constitutive death receptor-activation. Previous reports have shown that caspase 3 is not activated until day 3 of involution in WT mice [44] which is consistent with our results showing the absence of the 24 kDa cleavage product at the stages analyzed. Immunoblot to detect caspase 3 in Fig. 5d shows that cleaved caspase 3 was prematurely present in cIAP2−/− mouse glands at 24hINV compared with WT glands, consistent with large scale involution. Paraffin sections from d19L mice labeled using Apoptag to detect apoptotic cells confirmed the presence of numerous cIAP2−/− basal and luminal apoptotic cells while control glands showed apoptosis in only a subset of detached cells (Fig. 5e). Enumeration of apoptotic cells showed that a significantly greater percentage of cells were apoptotic in cIAP2−/− glands compared to control glands (Fig. 5f).

Thus, in the absence of cIAP2, early onset involution progresses rapidly to apoptotic cell death.

Absence of cIAP2 Increases the Latency of Dimethyl-Benzanthracene (DMBA)-Induced Tumor Formation

The cIAP1/2 proteins are thought to have pro-survival roles in cancer cells however the role of individual cIAPs in vivo in the process of mammary tumor formation has not been studied. Figure 6a depicts the time to tumor detection following the DMBA mammary carcinogenesis protocol [45] in WT and cIAP2−/− mice. The mean latency for cIAP2+/+ mice was 93 days while that for cIAP2−/− mice was significantly longer at 127 days. Three healthy cIAP2−/− DMBA-treated mice (12.5%) never developed any palpable lesions.

Fig. 6: Mice lacking cIAP2 develop tumors with long relative latency.
figure 6

a Kaplan-Meier analysis of tumor-free period in DMBA treated cIAP2−/− and WT mice. Animals were palpated three times weekly for lesions and latency is calculated from the final DMBA gavage. The analysis shows significantly longer latency to tumor palpation in cIAP2−/− mice compared with WT mice (p = .006, Log rank test), with 12% of cIAP2−/− animals remaining lesion-free. Only carcinomas were included in the analysis. b, TUNEL assay using Apoptag on sections of tumors derived from WT and cIAP2−/− mice. The top panel shows typical staining in a section from a cIAP2+/+ adenocarcinoma. The two panels below are representative of TUNEL staining in all adenocarcinomas obtained. Bar represents 100 μm. c, Enumeration of apoptotic cells in tumor sections from DMBA-treated cIAP2−/− and WT mice. Three random high power fields of 300 cells were counted from each tumor section (excluding necrotic areas) and subjected to a two-tailed student’s t-test (* p < 0.05 for both sets, Student’s t-test). Tumors: DMBA-induced tumors: cIAP2−/− n = 7; WT n = 6. d, Graph depicting the number of lesions representing each level of tumor differentiation in WT and cIAP2−/− mice. Adenocarcinomas were classified as: WDA, well-differentiated; MDA, moderately-differentiated; PDA, poorly-differentiated; ANA, anaplastic (undifferentiated). e, 30μg of protein extracts from adenocarcinomas, a squamous hyperplasia (SH) from WT and cIAP2−/− mice as indicated were immunoblotted to detect β-galactosidase. Both squamous hyperplasias used for blotting contained significant inflammation

Full size image

High rates of cell death in the absence of cIAP2 could delay the formation of tumors. TUNEL assays on tumor sections are shown in Fig. 6b. Percentages of apoptotic cells were quantified in tumor sections from both groups of mice. Figure 6c shows that cIAP2−/− tumors contained approximately an 8-fold greater percentage of apoptotic cells compared with WT tumors. Tumor histopathologies ranged from well to poorly-differentiated adenocarcinomas. Examples are shown in Supplemental Fig. S7A. Figure 6d depicts tumor pathologies from cIAP2−/− and WT mice. Some tumors contained squamous cell regions with varying keratinization and were categorized as benign pseudoepitheliomatous hyperplasia.

cIAP2 null mice have an insertion of the β-galactosidase gene downstream of the cIAP2 promoter [20]. Immunoblot for β-galactosidase showed the cIAP2 promoter is activated in cIAP2−/− tumors consistent with a role for cIAP2 in tumor survival and/or signaling (Fig. 6e). Squamous hyperplasia contained almost no detectable cIAP1 or cIAP2 mRNA while cIAP2 mRNA was expressed in WT tumors (Supplemental Fig. S7B). cIAP1/2 protein levels were higher in WT tumors compared to cIAP2−/− tumors with similar histopathology which likely correlates with the combined expression of cIAP1/2 in WT tumors (Supplemental Fig. S7C). No compensatory difference in the expression of XIAP protein in WT and cIAP2−/− tumors was detected (Supplemental Fig. S8). Collectively these results indicate that cIAP2 plays a significant anti-apoptotic role in the establishment of mammary adenocarcinomas.

Discussion

Both cIAP1 and cIAP2 have similar roles in signaling through TNFR-and related family member receptors where they function as E3 ligases to regulate caspases and the activation of NF-κB and MAPK [46]. Although cIAP1 and cIAP2 are considered to be redundant in most normal cells, our analysis showed that cIAP2 has a discrete role in maintaining mammary gland integrity during lactation such that absence results in the rapid onset of involution that was nearly complete by 36hINV. The reduced ductal branching in virgin cIAP2−/− mice likely contributed to engorgement by mid-lactation. The underlying reasons for the branching deficit are not clear since the combined levels of cIAP1/2 in virgins of both genotypes appeared to be similar until mid-lactation. It is possible that the global nature of the knockout may have affected factors involved earlier in mammary development. Indeed both Tweak [47] and another TNFR-related receptor for ectodysplasin [48] participate in mammary branching. Of note, cIAP2 mRNA is also expressed in the normal human mammary epithelial stem and immature progenitor cell hierarchy (Supplemental Fig. S9) and its absence may reduce viability and or affect cytokine signaling during branching morphogenesis. The only documented phenotype in cIAP2 mice is a defect in the macrophage response to lipopolysaccharide challenge [20] which appears to be unrelated to the mammary gland phenotype described here. The presence of constitutive alternative NF-κB activity and activated caspase 8 in 9 week-old cIAP2−/− virgin mice suggests that some aspects of TNFR-family signaling may be aberrant in the absence of cIAP2 during mammary gland development. However, activation of large scale apoptosis was not evident as indicated by the lack of caspase 3 cleavage and no apparent increase in apoptotic cells in glands of cIAP2−/− virgin mice. Notably, cIAP2−/− mammary tissues demonstrated a reduction in cIAP1 protein at mid-lactation coinciding with the onset of ductal distension. Moreover, cIAP1 levels remained low in cIAP2−/− mice until 36hINV while a reduction in total cIAP proteins was only detected in WT mice after 12hINV. Together these results suggest that cIAP2 is either very low or not expressed in the normal mammary gland until milk accumulation-induced cell stress activates Tweak reducing cIAP1 protein [17] and stabilizing cIAP2 protein [9]. Thus, cIAP2 contributes to maintaining the gland during phase I involution and, in its absence, involution proceeds rapidly. This thesis is further supported by results in other tissues such as intestinal epithelial cells where detachment induces the specific upregulation of cIAP2 which then functions to delay anoikis [49]. Importantly, the translation of cIAP2 mRNA is also promoted by stress [10, 50]. Applying this paradigm to mammary epithelium, cIAP2−/− cells would lack the response to weaning induced cellular stress and the restraint on cell death. Both Tweak signaling through Fn14 [17] and TNFα signaling through the TNFR-II [46] result in the degradation of cIAP1 and Traf2, ostensibly without affecting cIAP2. Baxter et al. [42] concluded that alternative NF-κB (p52) is a prosurvival transcription factor during mammary gland development. Thus, in the mammary gland, NF-κB (p52)-regulated cIAP2 may protect cells from rapid cell death induced by TNFα and Tweak thereby contributing to the reversibility of early stage involution.

The strong induction of P-Stat3 could contribute directly to cell death as described [29] and also indirectly induce Tweak [42]. Interestingly, recent evidence suggests that activation of Fn14 by Tweak can also activate Stat3 signaling [51] which could establish a feedforward loop. The ensuing rapid cell death would be amplified by the absence of both cIAP1 and cIAP2 in the Tweak-stimulated cIAP2−/− mammary glands as described above.

Autophagy is thought to function as a short term survival mechanism during early involution [33, 42]. cIAP2−/− glands showed increased autophagic flux by d7.5 L coincident with the detection of numerous detached cells in the lumen of alveoli and ducts. Interestingly, Tweak has been demonstrated to promote anoikis of detached vascular smooth muscle cells [52]. The reduction in P-Erk1/2 may result from loss of attachment to the basement membrane while the expression of Bim at d19L is consistent with the early onset of anoikis since Bim is normally detected at late-stage involution [53]. The diminished level of P-Erk1/2 in cIAP2−/− mammary glands at 24hINV could also potentiate the pro-apoptotic effects of Stat3 in mammary glands [8]. Figure 7a summarizes the proposed role of cIAP2 in the mammary gland where cIAP2 contributes to prevention of mechanical stress-induced cell death during early involution.

Fig. 7: Proposed roles of cIAP2 in the mammary gland.
figure 7

a, Alveolar and ductal stretch after weaning (or due to inadequate mammary branching complexity) results in myoepithelial detachment followed by luminal cell detachment or simultaneous separation of both cell layers from the basement membrane. Loss of attachment initiates autophagy which, when prolonged, leads to anoikis. The concerted cell detachment signals a proportionate induction of Tweak expression. In cIAP2−/− glands this results in accelerated involution, at least in part, due to the absence of cIAP2 which is stabilized in WT glands by reduced cIAP1 following its degradation induced by Tweak/Fn14. b, In the DMBA/MPA model of tumorigenesis MPA induces NF-κB through stimulation of RANKL thus promoting proliferation of carcinogen-initiated cells. NF-κB induces both TNFα and cIAP2 which promotes tumor cell survival in WT but not cIAP2−/− mouse mammary tumor cells where TNFα may contribute to apoptotic cell death

Full size image

IAP proteins are attractive targets for facilitating apoptosis in cancer cells, however, it is not clear whether cIAP1 and cIAP2 have redundant or distinct functions in tumor cell survival [54]. Human and mouse mammary tumors express NF-κB and inhibition prevents tumorigenesis [55, 56]. cIAP2 is robustly activated by NF-κB [57] and it is likely that cIAP2 survival signaling induced by NF-κB has a critical role in preventing cell death in transformed cells. In the DMBA/medroxyprogesterone (MPA) tumor model [45], MPA induces the RANKL/RANK axis in mammary epithelial cells and activation of NF-κB which then promotes proliferation, including DMBA-initiated cells [58]. The absence of cIAP2 in mammary tumors correlated with long latency associated with a high apoptotic rate. Nevertheless, tumors formed in cIAP2−/− mice at a similar frequency as in WT mice suggesting that proliferation eventually outpaces apoptosis in these tumors.

SMAC mimetics bind to cIAP1/2 proteins enhancing their E3 ligase activity to promote autoubiquitination resulting in the proteasomal degradation of cIAP1/2 proteins [59]. Since cIAPs restrain NF-κB activity, the degradation of cIAP1/2 induces cell death through NF-κB-mediated induction of TNFα [60, 61]. The extensive apoptosis in mammary tumors in the absence of cIAP2 indicates that cIAP1 does not fully compensate for this deficit which could be indicative of a general insufficiency in cIAP1/2 proteins and/or a discrete function of cIAP2 that is absent in the null mice. Indeed cIAP2 is more highly expressed in cancer cells than in normal tissue cells, due in part to activation by NF-κB [54]. Similar to the effects of SMAC mimetics, the reduction in overall levels of cIAP proteins in cIAP2−/− tunour cells could result in activation of the alternative NF-κB pathway, transcription of TNFα and induction of apoptosis pursuant to the deficit of cIAPs (Fig. 7b).

Results of the studies presented here demonstrate that cIAP2 has discreet cIAP1-independent roles in preventing rapid involution in the mammary gland and promoting survival during tumorigenesis.

Materials and Methods

Mice

cIAP2−/− mice have been previously described [20]. Chimeric males were originally mated with 129/SvJ females then the heterozygous progeny were backcrossed to C57BL/6 mice for more than 10 generations. Genomic tail DNA was used for genotyping. Virgin female mice (C57BL/6 cIAP2−/− and WT controls) were sacrificed at 9 weeks of age. Virgin mice were communally caged but were not estrus matched based on vaginal smears. Pregnant and lactating mice (between the ages of 10–12 weeks) were sacrificed at different days after the appearance of a vaginal plug (days of pregnancy/dP) or after pups were born (days of lactation; dL) or after pups were removed (hours post-weaning/involution; hINV). Weaning took place on the morning of d21L or at d10L as indicated. Male mice were removed from the cages one day after appearance of the vaginal plug. Litters were monitored for suckling intermittently. The 4th inguinal mammary glands were removed and used for analysis. Three virgins/dams were analyzed in each experiment per time point.

DMBA/MPA-Induced Mammary Tumorigenesis

cIAP2−/− females and WT littermates at 10 weeks of age were subcutaneously implanted dorsally with a 60-day medroxyprogesterone release pellet (Innovative Research, Hialeah, FL) and 5 days later began a course of gavages with 50 mg/kg DMBA in cottonseed oil [45]. Mice were monitored twice weekly for tumor formation by palpation following the final gavage of DMBA. All tumors were allowed to reach endpoint (maximum of 1cm3). Three cIAP2−/− and two WT mice died without a palpable tumor within 60 days of final gavage. Tumors were removed immediately after euthanization of mice. Kaplan-Meier curves and statistics were generated based on time from last gavage to tumor endpoint using GraphPad Prism 4.0.

Animal housing and experimentation was conducted in accordance with the Canadian Council on Animal Care guidelines with the approval of the University of Ottawa animal care committee.

Mammary Gland Whole Mounts and Analysis

Fourth inguinal mammary glands were excised from mice and spread onto glass microscope slides. Glands were fixed overnight in Carnoy’s fixative (10% glacial acetic acid, 30% chloroform, 60% ethanol). Fixed glands were washed in 70%, 50%, 25% ethanol then distilled water for 15 min each then fat was extracted in acetone following 3–20 min incubations. Glands were rehydrated in 100% ethanol then 95% ethanol for 20 min each then stained with hematoxylin for one hour at room temperature. After rinsing clear in water, specimens were de-stained by submerging in acid alcohol (50% ethanol, 0.2% HCl) twice for 30 min and then in fresh acid alcohol overnight. Finally glands were dehydrated in 70%, 95 and 100% ethanol for 20 min each then stored in toluene. Samples were mounted using Permount (Fisher). Branch points were enumerated on the same duct relative to the nipple and lymph node on mammary glands from 3 different mice per genotype.

Immunoblots and Antibodies

Total mammary gland and tumor protein extracts were obtained from snap frozen tissue that was pulverized under liquid N2. After the addition of RIPA buffer (50 mM Tris-HCl, 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS, pH 7.5, 10 μg/ml aprotinin, 1 μM pepstatin, 1 μM leupeptin, 1 mM phenylmethlysulfonyl fluoride (PMSF), 1 mM Na2VO4), the samples were sonicated then incubated on ice for 30 min. Insoluble material was removed by centrifugation at 9000 g for 20 min at 4 °C. Protein samples were aliquoted and frozen at -80C. Figures are representative of immunoblots of protein derived from at least two sets of mice of each genotype for each time point. Tumor samples were prepared in RIPA buffer and immunoblotting was performed as previously described [56]. The following primary antibodies were used for western blot analysis: anti-β-actin (Sigma #A-2066); anti-p38 #9212, anti-P-p38 #9211, anti-p100/p52 #4882, anti-Erk1/2 #4695, anti-P-Erk1/2 #9106, anti-STAT5 #9358, anti-P-STAT3 #9134, anti-STAT3 #9139, anti-caspase-3 #9661, anti-P-Akt ser473 #9271, anti-Akt #9272 (all from Cell Signaling), anti-P-STAT5 #562077 and anti-TRAF2 #558890 (BD Bioscience,); anti-vinculin (abcam, #ab129003); anti-LC3 (Novus Biologicals, #NB100–2220), anti-GAPDH (Biolegend, #MMS-5805); anti-BimEL (Millipore, #MAB17001), anti-β-casein (Santa Cruz Biotechnology #sc-166,520); anti-cIAP2 (Abcam #ab-23,423). Anti-rat IAP1 (RIAP) was obtained from Dr. R. Korneluk. Both anti-IAP antibodies detect mouse cIAP1 and cIAP2 as previously described [9] (and supplemental references) and as per the Abcam datasheet. Where indicated, immunoblots were quantified using Image J2.

Mammary Gland and Tumor Sections

Mouse tissues were paraffin-embedded and sections were prepared and stained with hematoxylin and eosin (H&E) using standard procedures by the University of Ottawa Department of Pathology and Laboratory Medicine. Briefly, slides were deparaffinized in toluene, hydrated (100%, 95%, 70% ethanol and water, 10 min each). Dehydration was performed in 70%, 95%, and 100% ethanol followed by toluene. H&E-stained sections were mounted with coverslips using permanent mounting media (Permount, Fisher). For immunohistochemistry, antigen retrieval was performed in the presence of 10 mM sodium citrate pH 6.0. Slides were blocked for 1 h in 2% normal goat serum, 1% BSA, 0.3% Triton-X-100. Anti-SMA or anti-p100/p52 antibody was diluted in antibody dilution buffer (DAKO #S0809) and incubated O/N at 4C followed by washing and the addition of DAKO Envision+peroxidase (DAKO K4002) and developed using DAB chromogen (DAKO) as per the manufacturer’s instructions. Counter staining was performed using filtered hematoxylin followed by rinsing in distilled water. Dehydration was performed in 70%, 95 and 100% ethanol then toluene. Slides were mounted using Permount.

Densitometry and Statistics

Densitometry was performed using open source Fiji/ImageJ2. Results were expressed as mean ± S.E.M. of three or more separate experiments. Student’s t-test was used for statistical analysis. A p-value <0.05 was considered statistically significant.

qRT-PCR

Mammary glands were stored in RNAlater (Thermo-Fisher Scientific). Total RNA was isolated using Tripure Isolation Reagent (Roche) or Qiazol (Qiagen) according to the manufacturer’s protocol and was resuspended in 20 ul of nuclease-free water. cDNA was generated by reverse transcribing five μg of RNA with the RevertAid H Minus reverse transcriptase (Fermentas/ThermoFisher) or Superscript III (Invitrogen/ThermoFisher). The resulting cDNA was diluted 1:20 and used in a qRT-PCR reaction with FASTStart Universal SYBR Green Master Mix (Roche). All qRT-PCR reactions were performed with an ABI 7500 Real-Time PCR System (ABI/ThermoFisher) and relative gene expression was quantitated using the delta CT method. Primers were designed online using the Roche Universal Probe Library Assay Design Center (https://lifescience.roche.com/shop/CategoryDisplay?catalogId=10001&tab=&identifier=Universal+Probe+Library) and manufactured by Invitrogen/ThermoFisher. Murine β-actin (Actnb) was used as the endogenous control for all experiments. Data shown is representative of 2 independent experiments performed in triplicate or the mean of values obtained from 3 mice for each time point as indicated in the figure legends.

Mouse Target Genes

Forward Primer

Reverse Primer

Actnb (beta-actin)

Ctaaggccaaccgtgaaaag

Accagaggcatacagggaca

Birc2 (cIAP1)

Gaagaaaatgctgaccctacaga

Aatgacgacatcttccgaac

Birc3 (cIAP2)

Tcgatgcagaagacgagatg

Tttgttcttccggattagtgc

Tnf

Ctgtagcccacgtcgtagc

Ttgagatccatgccgttg

Tweak

Caggatggagcacaagcag

Gctggagctgttgattttg

TUNEL Assays

TUNEL assays were performed on paraffin sections of mammary tissue and tumors using the Millipore/Chemicon (Temecula, CA) Apoptag ® in situ apoptosis detection kit according to the manufacturer’s directions.