Caspase-mediated Cleavage of Insulin Receptor Substrate.

Apoptosis is an important mechanism for maintaining tissue homeostasis. The efficient induction and execution of apoptosis are essential for cell clearance in specific developmental situations. Insulin-like growth factor (IGF)-I is a survival factor for epithelial cells in the mammary gland, and its withdrawal or inhibition leads to apoptosis. In this paper we describe a novel mechanism that may lead to suppression of an IGF-I-mediated signaling pathway through cleavage of insulin receptor substrate (IRS). During the process of forced weaning, when mammary epithelial cells rapidly enter apoptosis in vivo, IRS-1 and IRS-2 disappear. We have used cultured mammary epithelial cells to demonstrate that IRS removal can be mediated through the action of caspase 10. Caspase 10 activation and IRS-1 cleavage are regulated by a MKK1-signaling pathway but not by a phosphatidylinositol-3 kinase pathway nor by the extracellular proapoptotic ligands FasL, tumor necrosis factor-alpha-related apoptosis-inducing ligand (TRAIL), or transforming growth factor-beta3. In addition we show that the loss of IRS-1 after MKK1 inhibition prevents IGF-mediated phosphorylation of FKHRL1.

Apoptosis is an important mechanism for maintaining tissue homeostasis. The efficient induction and execution of apoptosis are essential for cell clearance in specific developmental situations. Insulin-like growth factor (IGF)-I is a survival factor for epithelial cells in the mammary gland, and its withdrawal or inhibition leads to apoptosis. In this paper we describe a novel mechanism that may lead to suppression of an IGF-I-mediated signaling pathway through cleavage of insulin receptor substrate (IRS). During the process of forced weaning, when mammary epithelial cells rapidly enter apoptosis in vivo, IRS-1 and IRS-2 disappear. We have used cultured mammary epithelial cells to demonstrate that IRS removal can be mediated through the action of caspase 10. Caspase 10 activation and IRS-1 cleavage are regulated by a MKK1-signaling pathway but not by a phosphatidylinositol-3 kinase pathway nor by the extracellular proapoptotic ligands FasL, tumor necrosis factor-␣-related apoptosis-inducing ligand (TRAIL), or transforming growth factor-␤3. In addition we show that the loss of IRS-1 after MKK1 inhibition prevents IGF-mediated phosphorylation of FKHRL1.
During development and in normal tissue homeostasis, unwanted or damaged cells are deleted by apoptosis. This process needs to be performed rapidly to ensure efficient cell removal. In some situations, an excessive activation of survival signaling can overcome commitment to apoptosis, thus preventing the removal of potentially deleterious cells. For example, following DNA damage and chromosomal aberrations in genetically unstable cells, apoptosis can be prevented by hyperactivation of survival signaling (1). In this study we have uncovered a cellular mechanism to minimize the activity of the signaling pathway driven by insulin-like growth factor-I (IGF-I) 1 in mammary epithelial cells.
IGF-I is a survival factor for many cell types including fibroblasts, neurons, epithelia, and cancer cells. In mammary epithelial cells, IGF signaling is important for survival as demonstrated through the use of both in vivo transgenic and cell culture models (2)(3)(4)(5)(6)(7). IGF-I binds the ␣-subunit of its cell surface tyrosine kinase receptor, inducing autophosphorylation of the intracellular ␤-subunit of the receptor. Signaling adaptor molecules, including insulin receptor substrates (IRS), are then recruited to the receptor via highly conserved amino-terminal pleckstrin homology and phosphotyrosine binding domains and activate downstream signals through interaction with Src homology-2 domain-containing proteins (8). IGF-I can trigger several downstream survival effector pathways in mammary epithelial cells, either through a protein kinase B pathway that leads to phosphorylation and therefore suppression of forkhead transcription factors and BAD or via activation of Ras, Raf, MAPK kinase 1 (MKK1, also called MEK), and Rsk-1 through either Shc or epidermal growth factor receptor transactivation (7, 9 -13).
Following weaning of mammalian offspring, the secretory alveolar epithelial cells that normally produce milk are no longer required and are removed by apoptosis. The induction of apoptosis within this cell population can be synchronized experimentally in the mouse by removing suckling pups at lactation day 9. The ensuing involution occurs in two phases. The initial step involves apoptosis of the secretory epithelial cells, whereas the second stage comprises remodeling of the mammary gland by proteases (14,15). Apoptosis within many tissue types in vivo occurs sporadically, and it is therefore difficult to determine how it is regulated. However, the ability to synchronize apoptosis induction in vivo in the mouse mammary gland provides an excellent method for biochemical analysis of the mechanisms involved.
The decisive trigger for apoptosis in the first phase of mammary gland involution is unknown; however, a number of molecules and stimuli may be involved. These include milk stasis, transforming growth factor-␤3, leukemia inhibitory factor, interleukin-6, tumor necrosis factor-␣-related apoptosis-inducing ligand (TRAIL), Fas ligand (FasL), and loss of integrin activation (16 -23). Because IGF is an essential survival factor for mammary epithelial cells in culture, a further apoptosis mechanism in vivo may occur via inhibition of IGF signaling. One route for this is through the induction, early in involution, of IGF-binding protein 5 (IGFBP-5), which blocks IGF signaling and thereby induces apoptosis (13,24,25). The possibility that the survival potential of IGF is suppressed in vivo led us to hypothesize that its intracellular signaling pathway may indeed be altered during involution.
The onset of mammary epithelial apoptosis in vivo coincides with the removal of IRS, the adaptor protein for IGF signaling (26). The specific aim of this study was to use cultured mouse mammary epithelial cells to identify a mechanism for IRS disappearance. We show that this involves the activation of a caspase 10-related molecule with similar characteristics to human caspase 10. Caspase 10 has not been found previously to have substrates except other caspases, and we show that it cleaves IRS. Moreover, caspase 10 is activated by a novel mechanism that can be mimicked by blockade of a MKK1-signaling pathway.

EXPERIMENTAL PROCEDURES
Mammary Gland Analysis-Mammary tissue from female ICR mice (Harlan Olac) was collected from 8-week-old virgins and from mice pregnant for 9 and 15 days after detection of a vaginal plug. Litters were normalized to 10 pups/mouse and removed on lactation day 9. Tissue was also collected from mice on lactation day 9; 24 h after this on involution day 1; and at involution days 2, 3, 5, and 10. DNA from mammary gland tissue was analyzed as described previously (27).
Protein Analysis-Mammary gland tissue was homogenized in radioimmune precipitation assay buffer containing phosphatase and protease inhibitors. Primary mammary epithelial cells were prepared according to previously published methods (28). Briefly, mammary glands from 14.5-18.5-days-pregnant ICR mice were isolated, chopped, digested in a collagenase digestion medium, and washed to remove contaminating fibroblasts and other single cells. The resulting epithelial organoids were plated for 3 days on collagen I-coated tissue culture dishes in Ham's F12 medium containing 10% heat-inactivated fetal calf serum, 10 ng/ml epithelial growth factor, 1 mg/ml fetuin, 880 nM insulin, and 2.8 nM hydrocortisone. Primary cells and FSK-7 mouse mammary epithelial cells (29,30) were harvested by scraping into NET (50 mM Tris, 150 mM NaCl, 5 mM EDTA, 1% Nonidet P-40, pH 7.4) lysis buffer (7). 20-g samples (concentration determined by BCA assay) were loaded onto gels initially, but slight adjustments were subsequently made to ensure that the levels of E-cadherin (as a marker for epithelial cells) were equivalent in each lane. The samples were then immunoblotted with primary antibodies against IRS-1 or IRS-2 (Upstate Biotechnology, 1 g/ml); BAD (RϩD Systems, 1 g/ml); caspase 3 (Santa Cruz Biotechnology, 0.2 g/ml); caspase 10 (Cell Signaling Technology, 65 ng/ml; and Abcam Ltd., ab2012, 1:500); phospho-Erk (Santa Cruz Biotechnology, 50 ng/ml), and horseradish peroxidase-linked secondary antibody before visualization by enhanced chemiluminescence (Amersham Biosciences).
PCR-Total RNA was isolated from mammary gland tissue using Bio/RNA-Xcell TM (Biogene) and treated with RNase-free DNase prior to reverse transcription with random primers using Superscript preamplification (Invitrogen). Primers to GAPDH cDNA (forward, 5Ј-ACC-ACAGTCCATGCCATCAC-3Ј; reverse, 5Ј-TCCACCACCCTGTTGCTG-TA-3Ј) and IRS-1 cDNA (forward, 5Ј-AACTATGCCAGCATCAGCTT-3Ј; reverse, 5Ј-TCTACTGAAGAGGAAGACGT-3Ј) were utilized in PCR. Reverse transcription negative samples with no reverse transcriptase added were used to identify genomic DNA contamination. GAPDH primers showed that equal amounts of RNA had been added to the reaction. PCR products were electrophoresed on 1% agarose gels.
Caspase Cleavage Assay-Lysates from primary mammary epithelial cells were immunoprecipitated with 0.2 g of anti-IRS-1 or -2 antibody and 25 l of protein A-Sepharose beads (Zymed Laboratories Inc.) and washed 3 times with NET buffer with a final wash of 50 mM Tris, pH 7.2. Caspase reaction buffer (50 mM HEPES, 50 mM NaCl, 0.1% CHAPS, 10 mM EDTA, 5% glycerol, 10 mM dithiothreitol, pH 7.2) was added to the beads; and they were incubated with 1 unit of recombinant human active caspase 1, 2, 3, 6, 7, 8, 9, or 10 (37°C, 1 h). The supernatant was removed, and the beads were boiled in sample buffer. Caspases were supplied in an active form (Calbiochem, 218812), and we confirmed that they were actually active by using kinetic assays to measure the cleavage of specific fluorogenic substrates for each of the caspases (Calbiochem fluorogenic substrate set IV, 218809).
Transfections-Constructs containing myc-tagged human IRS-1 (Ken Siddle, University of Cambridge) were transfected into 50 -70% confluent FSK-7 cells using LipofectAMINE PLUS TM (Invitrogen) for 3 h; then the medium was changed, and the cells were cultured overnight before harvesting. Immunoprecipitations and caspase cleavage assays were performed as above.
Apoptosis Assay-Primary mammary epithelial cells were cultured on collagen I for 3 days and serum-starved for 2 h, and those for suspension studies were trypsinized onto a non-adherent substratum, PolyHEMA (Sigma). Cells were incubated for 5 or 24 h with the following: 16 pM SUPERFas Ligand TM or 18 nM Killer TRAIL TM (Alexis); 10 M U0124, 10 M U0126, 1 M LY294002, or 10 M staurosporine (Calbiochem); and 150 M caspase 10 inhibitor z-AEVD-fmk (RϩD Systems). Cells were harvested in sample buffer and immunoblotted as above. Activation of caspase 3 can be measured by immunoblotting with an antibody to the active form of caspase 3, which corresponds to induction of apoptosis as measured by changes in nuclear morphology (7). We were not able to identify an antibody that recognizes activated mouse caspase 10 and thus assessed its activation in cultured cells by the disappearance of the full-length 58-kDa form.
In Vitro Transcription and Translation-myc-tagged IRS-1 was in vitro transcribed and translated incorporating [ 35 S]methionine (Amersham Biosciences) using TNT® coupled reticulocyte lysate (Promega). The resulting protein was electrophoresed and Coomassie Blue-stained, and then it was treated with Amplify and exposed to autoradiography film or a phosphorimaging plate.

Apoptosis in Vivo Coincides with the Disappearance of IRS
Protein and Dephosphorylation of BAD-The mechanism for the regulation of apoptosis during mammary gland involution is not fully understood; however, it has been proposed that the actions of the survival factor IGF-I are suppressed. Because primary cultures of mammary epithelial cells depend on IGF-I for their survival, we sought biochemical evidence for altered IGF signaling during the transition from lactation to involution (5).
First we examined components of the IGF-I-signaling pathway in vivo and focused on IRS, the adaptor for IGF signaling. When we analyzed IRS levels, we found a dramatic reduction in the protein levels of both IRS-1 and -2 on involution day 1 (Fig.  1a). The epithelial cell content of the mammary gland changes during the pregnancy cycle, and we were careful to use quantities of extract that had been normalized according to an epithelial cell marker, E-cadherin. Apoptosis is induced extensively within 24 h of weaning in ICR mice and remains maximal for 3 days (Fig. 1b) (27). IRS disappearance therefore coincides with the apoptosis that characterizes involution.
To examine whether the loss of IRS correlates with altered activity of any components of the IGF-signaling pathway that are known to be involved directly with IGF-mediated survival, we examined the phosphorylation status of the proapoptotic protein BAD. This protein is expressed in the mammary gland, and its proapoptotic activity is inhibited by phosphorylation (27,31,32). Moreover, the phosphorylation status of BAD is regulated by IGF-I in mammary epithelial cell cultures (7). Multiple phosphoforms of BAD can be detected using high percentage SDS-polyacrylamide gels, and we therefore asked whether its phosphorylation is altered during the transition from lactation to involution (33). BAD was phosphorylated during lactation, indicating its inactivation. Twenty-four h after removing pups, when apoptosis is initiated in vivo (Fig. 1b), the phosphorylated forms of BAD were not detected (Fig. 1c). This suggests that a survival-signaling pathway leading to BAD phosphorylation and inactivation during lactation is inhibited at involution. In this study, we have not attempted to identify specifically which signaling pathway is responsible for the altered BAD phosphorylation in vivo, but our results are suggestive that the inhibition of IGF signaling through loss of IRS might be involved.
Multiple mechanisms are required for mammary gland apoptosis in vivo, and we now suggest that the loss of IRS-1 and -2 might provide an important contribution. Others have noted the disappearance of IRS in the mammary gland (26), but the mechanism of its loss has not been addressed previously. Therefore the subsequent aim of this study was specifically to determine a possible mechanism for IRS disappearance by using cultured mammary epithelial cells.
The Decrease in the Level of IRS-1 Protein Is Not Mediated through a Transcriptional Mechanism-IRS-1 is the major IRS isoform involved in IGF-I survival signaling (34). One possible mechanism to explain the loss of IRS-1 during involution is a reduction in its mRNA level. Semiquantitative reverse transcription PCR with primers designed to GAPDH and IRS-1 was therefore performed to identify whether the decrease in IRS-1 protein levels is due to a corresponding decrease in IRS-1 mRNA levels. There were no large changes in IRS-1 mRNA levels during the transition from lactation to involution compared with GAPDH levels (Fig. 1d). This indicates that an alternative, most likely posttranscriptional, mechanism leads to the decrease in IRS-1 protein during involution.
Caspase 10 Can Cleave IRS-1 and -2 in Vitro-An alternate mechanism for the loss of IRS is that it is removed by caspases following an initial apoptotic trigger. Caspases are activated during apoptosis and are responsible for destroying numerous cellular proteins (35,36). To determine whether caspases are responsible for IRS cleavage, IRS-1 and -2 protein was isolated from primary mammary epithelial cell lysates by immunoprecipitation and incubated with a panel of recombinant active human caspases comprising caspases 1, 2, 3, 6, 7, 8, 9, and 10. Caspase 10, but none of the other caspases, cleaved both IRS-1 and -2 proteins (Fig. 2, a and b). The specificity of the reaction was confirmed by including the unique caspase 10 inhibitor z-AEVD-fmk in the reaction (Fig. 2c). IRS proteins have not been shown previously to be a target for caspases, and there are very few, if any, known caspase 10 targets.
To identify any cleavage products produced by incubation of IRS-1 with caspase 10, lysates from FSK-7 mouse mammary epithelial cells transfected with a plasmid expressing carboxylterminal myc-tagged human IRS-1 were immunoprecipitated with an anti-myc antibody. The precipitated IRS-1-myc was then treated with recombinant caspase 10, and immunoblots were probed with an antibody that recognizes the amino terminus of IRS-1. An ϳ16-kDa-sized fragment of IRS-1 was produced by incubation with caspase 10 (Fig. 2d). To confirm that this amino-terminal fragment of IRS-1 was generated by the direct action of caspase 10, IRS-1 was translated in vitro in the presence of [ 35 S]methionine and then treated with recombinant caspase 10. The amino-terminal anti-IRS antibody immunoprecipitated full-length IRS-1 from the untreated in vitro translation reaction, but IRS-1 was cleaved by caspase 10 to yield an amino-terminal fragment of ϳ16 kDa (Fig. 2e).
Together these results demonstrate that IRS-1 is subject to caspase-mediated cleavage, specifically by caspase 10, resulting in the disappearance of full-length IRS-1 and the generation of an amino-terminal fragment of ϳ16 kDa.
A Caspase 10-related Protein Is Present in the Mammary Gland in Vivo-Because IRS-1 disappears during mammary gland involution in vivo, we asked whether caspase 10 is present in the tissue, in particular within epithelial cells, as these are the cells that normally undergo apoptosis in vivo. Caspases 1, 3, 7, 8, and 9 have been identified in the mammary gland, but caspase 10 has not been studied previously (37,38).
The expression of caspase 10 was determined by immunoblotting with an antibody recognizing all human caspase 10 splice variants (raised to a peptide corresponding to residues around Asp-219 within the amino-terminal prodomain). Caspase 10-related proteins were present both in the gland of the mouse and also in cultured mammary epithelia (Fig. 3a). Two isoforms of caspase 10 were identified, a 58-kDa protein found both in vivo and in primary cultures and a 40-kDa isoform present only in the gland. The detection of both proteins by the antibody was inhibited when it was preincubated with its immunizing peptide, confirming the specificity of the anticaspase 10 antibody interaction. The 58-kDa isoform seen in primary cells was also present in the mouse mammary FSK-7 cell line (Fig. 3b).
To confirm that the 58-and 40-kDa proteins represent caspase 10, we performed two further experiments. First, the size of the 58-kDa isoform in mouse mammary cells was compared with that of caspase 10 in human Jurkat T cells (39) and was found to be identical (Fig. 3b). Caspase 10 has largely been studied in immune cells, and we identified a similarly sized protein in the mouse thymus. Second, we used an additional antibody to caspase 10 with a recognition epitope contained within its carboxyl-terminal domain. This is the only other currently available antibody to mouse caspase 10, and it recognizes the FLICE2 splice variant (caspase 10b) but not Mch4 (caspase 10a). This antibody also detects the 40-kDa but not the 58-kDa protein in mammary tissue (Fig. 3c).
Together these results indicate that the proteins present in mammary cells and tissue correspond to mouse caspase 10. Furthermore they suggest that two splice variants are present, a 58-kDa protein that is present in epithelial cells and is likely to be the Mch4 form of caspase 10 and a 40-kDa protein that is FIG. 1. Loss of IRS during mammary apoptosis in vivo. a, involution coincides with the disappearance of IRS. Immunoblots (IB) of the mouse mammary gland of virgin (V) mice and tissue from pregnancy days 9 and 15 (P9 and P15), lactation day 9 (L9), and involution days 1, 2, 3, 5, and 10 (I1-10) were probed with antibodies to IRS-1 or -2. An E-cadherin antibody was used to confirm that the immunoblotting analysis was performed on the basis of equivalent epithelial cell content. Note that IRS-1 and -2 completely disappear early in involution at times corresponding to apoptosis induction (27). In this and all subsequent experiments, the data shown are representative of at least three repeat experiments. b and c, apoptosis coincides with the dephosphorylation of BAD. b, DNA from mammary glands of the mice used for protein analysis was separated by agarose gel electrophoresis. The only samples demonstrating apoptosis by internucleosomal DNA fragmentation were those of early involuting mammary gland (I1, -2, and -3). c, immunoblots probed with an antibody to BAD. Note that under the conditions used, BAD separates into three bands, the upper two of which are phosphorylated (*), and they disappear on involution day 1 (7,33). d, disappearance of IRS-1 in involution is not due to loss of IRS-1 mRNA. Total RNA from mammary tissue from the same mice used in a, b, and c was reverse transcribed and PCR-amplified with primers for IRS-1 and GAPDH for the number of cycles shown. Control reverse transcriptions without reverse transcriptase (Ϫ) confirmed that the samples were not contaminated with genomic DNA. in a cellular compartment other than epithelia and may be a short splice variant of FLICE2.
An Intracellular Mechanism for Caspase 10 Activation and IRS-1 Cleavage-Because our results indicate that IRS-1 protein might be cleaved by caspase 10, we addressed the mechanism of its activation using cultured primary mammary epithelial cells. Adhesion to the extracellular matrix (ECM) is essential for survival in culture, and integrins, which mediate cell attachment, become deactivated on day 1 of involution in vivo (20,40,41). Moreover, growth factors are necessary for mammary cell survival and are likely to have an important role in vivo (4, 5, 7). We therefore asked whether primary cells cultured on ECM or on a non-adhesive substratum (Poly-HEMA), with or without growth factors, showed evidence of IRS-1 cleavage. Although apoptosis occurred under these conditions, as seen by cleavage of full-length caspase 3 into a smaller, active form, there were no changes in IRS-1 protein levels (Fig. 4a). Moreover, the 58-kDa caspase 10 was expressed in the cells, but neither the loss of cell-matrix interactions nor the removal of growth factors caused a disappearance of full-length enzyme, indicating that it was not activated.
Transforming growth factor-␤3 is required for efficient mammary apoptosis in vivo, and we therefore tested its involvement (17). Although this factor induced apoptosis in cultured cells, we found no evidence for caspase 10 activation or IRS cleavage (data not shown). Moreover, leukemia-inhibitory factor, which  a and b,  respectively). Individual enzymes, a mixture of all of the caspases (All), controls with no caspases (None), and controls with no primary antibody in the immunoprecipitation (*) were used. In each case a sample of the cell lysate was analyzed by immunoblotting with anti-IRS-1 or -2 antibody to confirm that equal amounts of IRS-1 or -2 were present in each sample prior to immunoprecipitation (lower panels of a and b, respectively). c, IRS-1 and -2 immunoprecipitations, prepared as in a, were treated with recombinant caspase 10 in the presence or absence of its specific inhibitor, z-AEVD-fmk. Note that z-AEVD-fmk prevented the reduction in IRS-1/2 protein levels induced by caspase 10. d, caspase 10 cleaves recombinant human IRS-1 to generate a 16-kDa amino-terminal fragment. FSK-7 mammary cells were transfected with pRC.cmv, encoding myc-tagged IRS-1; the lysates were immunoprecipitated with an anti-myc antibody; and the immunoprecipitations were treated with recombinant caspase 10 before separation and immunoblotting with an amino terminus-specific anti-IRS-1 antibody. Controls with no primary antibody or no DNA transfected were also used. Caspase 10 caused the disappearance of full-length myc-tagged IRS-1 and instead yielded a 16-kDa aminoterminal fragment. e, caspase 10 cleaves in vitro translated IRS-1. 35 S-Labeled in vitro translated IRS-1 was incubated with recombinant caspase 10. The resulting protein was immunoprecipitated with an amino-terminal IRS-1 antibody and analyzed by SDS-PAGE and autoradiography. Note that full-length IRS-1 disappears following caspase 10 treatment, yielding a 16-kDa amino-terminal fragment. The extra bands present in the caspase 10-treated and untreated lysates are likely to represent nonspecific 35 S-labeled reticulocyte lysate proteins or IRS-1 fragments for which correct transcription and translation have not been accomplished. also regulates mammary apoptosis in vivo (22), did not induce caspase 10 cleavage (data not shown).
Two death receptor ligands, FasL and TRAIL, have been implicated in mammary gland involution (18,19). Although the mechanism of caspase 10 activation has not been defined clearly, caspase 10 is most similar to caspase 8 and has therefore been considered an initiator caspase, activated by the ligation of death receptors (42,43). To attempt to activate caspase 10 in primary cell culture, we therefore treated cells in suspension with FasL or TRAIL. Both ligands induced caspase 3 activation (Fig. 4b) and apoptosis as measured by examining nuclear morphology (data not shown) but did not alter the amount of full-length caspase 10 or IRS-1. Thus, death ligands do not activate caspase 10, nor do they result in the cleavage of IRS-1.
Our data further suggest that caspase 10 is not a substrate of either the apoptosome, which cleaves caspase 3, or caspase 3 itself because caspase 3, but not caspase 10, is activated by death ligands and by growth factor/ECM withdrawal. We therefore examined whether intracellular enzymes promote IRS cleavage. Initially, we studied apoptosis induced by the FIG. 3. A caspase 10-related protein is present in the mammary gland and in cultured mammary epithelial cells. a, blots (IB) of protein from the mammary gland and primary cultures were probed with antibodies to human caspase 10 and to Erk1/2 as a confirmation of equal sample loading. Note that isoforms at 40 kDa and 58 kDa were present, but only the 58-kDa form was expressed in the epithelial cells isolated from the mammary tissue. Immunodetection of both the 40and 58-kDa forms of caspase 10 was inhibited by preincubating the antibody with its immunizing peptide. b, protein extracts of primary mammary epithelial cells and FSK-7 cells were separated by SDS-PAGE adjacent to an extract from human T cells (Jurkat) and mouse thymus tissue, and the blots were probed with antibodies to full-length caspase 10 (upper panel) and caspase 10 antibody preincubated with the immunizing peptide (lower panel). The mobility of the mouse protein is identical to that of human caspase 10. c, protein extracts of mouse mammary gland tissue from pregnancy day 15 were separated by SDS-PAGE and transferred to nylon, and strips were probed with antibodies to (ii) all splice variants of caspase 10 and (iii) the carboxylterminal domain of caspase 10b. Lane i was probed with IgG as a control. Note that the antibody recognizing the carboxyl terminus only recognizes the 40-kDa protein.

FIG. 4. Physiological disappearance of IRS-1 in mammary epithelial cells is regulated by the MAPK pathway.
a, loss of adhesion induces apoptosis but not IRS cleavage. Primary mammary epithelial cells were cultured on collagen with (control) or without growth factors (no GF) or plated on PolyHEMA for 5 h with growth factors. Proteins from cell lysates were immunoblotted (IB) with antibodies to cleaved caspase 3 or full-length caspase 10. Apoptosis was increased slightly in the absence of growth factors and strongly by loss of adhesion to ECM, but full-length IRS-1 and caspase 10 remained intact. b, FasL and TRAIL induce apoptosis but not IRS cleavage. Mammary cells were suspended by plating on PolyHEMA (control) and treated with FasL or TRAIL for 5 h in the absence of growth factors. Apoptosis was demonstrated by immunoblotting lysates with an antibody specific for cleaved active caspase 3. No changes in the levels of full-length caspase 10 or IRS-1 were observed. c, STS-induced apoptosis is accompanied by caspase 10-dependent loss of IRS-1. Primary cells were suspended by plating on PolyHEMA (control) and treated with STS alone or STS together with the caspase 10 inhibitor z-AEVD-fmk for 24 h. A low level of caspase 3 activation was observed in control samples cultured on PolyHEMA, but this increased after STS treatment. Full-length caspase 10 disappeared in the presence of STS, indicating its activation, which correlated with the loss of IRS-1, confirming a link between caspase 10 activation and IRS cleavage within cells. z-AEVD-fmk blocked the IRS-1 cleavage. d, inhibition of MKK1 induces caspase 10-dependent cleavage of IRS-1. U0126 and its non-functional homologue, U0124, were added to serum-starved cells on collagen for 24 h. Lysates were immunoblotted for phosphorylated Erk to confirm inhibition of signaling through MKK1 and for Erk to ensure equal loading. Serum starvation resulted in some caspase 3 cleavage, but this increased with U0126. MKK1 inhibition induced IRS-1 loss and caspase 10 cleavage, which were both prevented by addition of z-AEVD-fmk, even though Erk remained unphosphorylated. generic protein kinase inhibitor staurosporine (STS). Cell death is triggered by STS in many cell types (44,45). This caused a reduction in IRS-1 protein levels, and moreover caspase 10 became activated, as indicated by loss of its fulllength form (Fig. 4c). The caspase 10 inhibitor z-AEVD-fmk prevented the STS-mediated decrease in IRS-1 protein levels, confirming that IRS loss is mediated through the action of a caspase. This was accompanied by a corresponding inhibition of caspase 10 activation.
Although STS is a fairly nonspecific apoptosis activator, this experiment suggests that protein kinase inhibition might be responsible for triggering the activation of caspase 10 and IRS cleavage. We therefore asked whether specific signaling pathways previously shown to regulate survival are involved. First, we prevented phosphatidylinositol 3-kinase signaling using LY294002, which induces apoptosis in mammary epithelial cells (5). LY294002 activated caspase 3 but had no effect on caspase 10 or IRS-1 (data not shown). We then inhibited signaling through the MAPK pathway with U0126, which is largely specific for MKK1 at 10 M (46). U0126 increased apoptosis in primary mammary epithelial cells, as observed by caspase 3 cleavage (Fig. 4d). Importantly U0126, but not its non-functional homologue U0124, caused a decrease in the levels of full-length IRS-1 and caspase 10. Furthermore, the cleavage of these two proteins was inhibited by addition of z-AEVD-fmk, confirming caspase involvement. The caspase 10 inhibitor did not reverse the dephosphorylation of Erk1/2 by U0126, indicating that activation of caspase 10 occurs downstream of MKK1.
Together these results demonstrate that the MAPK pathway is involved in the regulation of survival through a novel mechanism. Inhibition of MKK1 leads to the activation of caspase 10 and thereby the cleavage of IRS.
Cleavage of IRS-1 Prevents IGF-mediated Intracellular Signaling-Removal of IRS-1 protein may prevent, or at least suppress, IGF-mediated survival signaling. To test this hypothesis, cells were treated with U0124 or U0126 for 24 h, and IGF-I was then added for 15 min. The forkhead transcription factor FKHRL1 is a downstream target in the IGF-I pathway and is phosphorylated rapidly after IGF stimulation (13). Phosphorylated FKHRL1 electrophoreses slightly behind non-phosphorylated FKHRL1 on SDS-polyacrylamide gels, and both forms can be detected with an anti-FKHRL1 antibody. A decrease in the levels of IRS-1 after U0126 treatment resulted in a reduced ability of IGF-I to trigger FKHRL1 phosphorylation (Fig. 5). Thus, the cleavage of IRS-1 prevents efficient signal transduction by IGF to one of its cellular targets. Because dephosphorylation of FKHRL1 leads to its activation as a proapoptotic factor (9), we suggest that the cellular pathway leading to IRS cleavage alters the ability of IGF-I to act as a survival factor in mammary cells.
Regulation of Caspase 10 during Mammary Gland Development-Because the 58-kDa splice variant of caspase 10 is present in mammary epithelial cells and is likely to be the isoform responsible for IRS cleavage, we asked whether it becomes activated during mammary gland involution in vivo. Tissue samples were examined for the disappearance of full-length caspase 10 as a marker for its activation (Fig. 6). Caspase 10 was present in mammary glands of non-pregnant and pregnant mice, was reduced during lactation, and became almost completely absent from the start of involution, reappearing again on involution day 5. These results demonstrate that the loss of IRS-1 and IRS-2 in vivo, occurring at the beginning of involution when apoptosis is maximal, is accompanied by activation of caspase 10. This enzyme may therefore act as a control point for IGF survival signaling in vivo. DISCUSSION IRS-signaling intermediates in the IGF-I survival-signaling pathway are dramatically decreased in the mammary gland at involution with kinetics that reflect the induction of apoptosis in vivo (26). We have now elucidated a possible mechanism for IRS disappearance, which is through the activation of caspase 10, and have identified an intracellular signaling pathway that leads to its activation and IRS cleavage. We do not suggest that IRS loss triggers apoptosis but rather that it represents a novel enhancement step for inhibiting IGF-mediated signals. Future studies will be necessary to determine whether IRS removal in vivo is actually mediated by this mechanism. However, given that caspase 10 is expressed in mammary epithelial cells and that Erk signaling is rapidly deactivated at the onset of involution in vivo (22), this seems a likely possibility.
Mechanism for Mammary Gland Involution in Vivo-The signals that control the initiation of mammary gland involution are not fully understood. It is known that a variety of triggers is required in combination to achieve full epithelial cell apoptosis, but importantly, deletion of individual components does not fully suppress involution (47). Factors that are currently known to be involved include the following: (i) induction of transforming growth factor-␤3 synthesis, which occurs after weaning and is required for rapid apoptosis (17); (ii) induction of leukemia-inhibitory factor, which activates Stat3 and is similarly required for efficient apoptosis in vivo (22,23); (iii) in-  Fig. 1 were immunoblotted (IB) with the antibody that recognizes the 58-kDa form of caspase 10. Blots were also probed with an antibody to E-cadherin. Note that lower levels of caspase 10 are present during lactation but that caspase 10 has completely disappeared at the beginning of involution. V, virgin mice; P9 and P15, tissue from pregnancy days 9 and 15; L9, tissue from lactation day 9; I1-10, tissue from involution days 1, 2, 3, 5, and 10. duction of interleukin-6 expression (21); (iv) disruption of Wnt signaling through the expression of a secreted frizzled-related protein-4 (48); (v) up-regulation of Fas, FasL, and TRAIL at the onset of involution (18,19); and (vi) alteration of cell-ECM interactions (15,20). Data to support a role for IGFs in mammary epithelial cell survival comes both from in vivo studies, where overexpression of IGF-I or des(1-3)IGF-I in the mammary gland suppresses apoptosis and delays the involution process, and from culture models, where exogenous IGF is a potent survival factor (3)(4)(5)7). One mechanism that is utilized to alter IGF signaling and thereby contribute to apoptosis is the up-regulation of IGFBP-5 (13,25).
Apoptosis Enhancement through Caspase-mediated Cleavage of IRS-We now describe a second mechanism that may contribute to inhibition of IGF survival signaling. This mechanism involves the cleavage and inactivation of IRS-1/2 by caspase 10. IGF-I induces the phosphorylation of IRS, subsequently leading to phosphatidylinositol 3-kinase-mediated survival signaling. IGF can also induce the MAPK pathway through Shc in some cell types (49), but in mammary cells it appears to do so through EGF receptor transactivation (7). Although loss of IRS is unlikely to induce apoptosis dominantly, as for example its deletion in Drosophila does not compromise viability (50), we suggest that its cleavage provides an apoptosis enhancement mechanism by blocking intracellular survival signals induced by IGF-I. Indeed following IRS cleavage, IGF-I is no longer able to induce the phosphorylation (and thereby suppress activity) of one of its apoptosis targets, FKHRL1. In an organ where tissue remodeling following an aborted lactation is required as quickly as possible, efficient apoptosis induction is paramount for successful and rapid subsequent pregnancies, and our data suggest that caspase 10-mediated IRS cleavage may contribute to this process.
Several other intracellular proteins are cleaved by caspases during apoptosis. Many are architectural, and their cleavage is important for collapsing structural elements and packaging the apoptotic cell for phagocytosis. Others are involved with nuclear changes and cleaving DNA (51). In addition, some caspase-cleaved proteins have inhibitory effects on their signaling pathways such as p130 cas , myocyte enhancer factor-2, and desmin (52)(53)(54). IRS may fall into this category, as its cleavage suggests a functional inhibition of the IGF survival pathway. We have not yet characterized all of the caspase 10-generated cleavage products in our in vitro analysis, although an ϳ16-kDa amino-terminal fragment is produced that contains the pleckstrin homology domain of IRS-1 (55). This domain is required for signal-dependent membrane localization of IRS-1 and its tyrosine phosphorylation (56,57). The pleckstrin homology domain is also necessary for the downstream functions of IRS-1 (58). Thus, following cleavage, any residual IRS-1 may exhibit a reduced ability to localize to the cell membrane and transduce survival signals. A further possibility is that once cleaved, the pleckstrin homology domain might act in a dominant negative fashion toward the IGF-I receptor to prevent any additional signals being transduced through alternative signaling intermediates such as Shc, Grb2, Sos, and Ras.
Involvement of Caspase 10 in IRS Cleavage-Our culture analysis suggests that the mechanism for IRS loss is through a specific cleavage by caspase 10. This suggestion is supported by in vivo experiments indicating that the loss of IRS-1 and IRS-2 is accompanied by a decrease in detectable full-length caspase 10. This enzyme has at least four splice variants, caspases 10a/ Mch4, 10b/FLICE2, 10c, and 10d (42, 43, 59, 60). Caspase 10 is conserved from amphibia to mammals (61), and three previous reports have identified caspase 10 protein in rodent cells (62)(63)(64). Using the antibody that recognizes the aminoterminal prodomain contained in all isoforms of human caspase 10, we detected a 58-kDa isoform of caspase 10 in the mouse mammary gland. This isoform is also present in purified mammary epithelial cells, suggesting that it is responsible for IRS-1 cleavage. It has the same mobility as both the previously identified caspase 10 in human Jurkat T cells and a protein in mouse thymus (39). The 58-kDa isoform is likely to be caspase 10a/Mch4 because it contains the prodomain but is not recognized by the FLICE2-specific antibody. We also found a 40-kDa isoform in the mouse mammary gland in vivo but not in epithelial cells. It is possible that the 40-kDa form of caspase 10 is expressed within other mammary cell types that are not isolated during primary epithelial cell preparation, such as adipocytes, fibroblasts, or within the lymph nodes. The 40-kDa protein is likely to be a splice variant of caspase 10 containing the amino-terminal prodomain linked to the carboxyl-terminal region of FLICE2, which is encoded by exon 11 (59).
Caspase 10 has been proposed as both an initiator caspase and, in one instance, an effector caspase; however, little is known of its activation stimulus or cellular substrates (39). It can activate caspases 2, 3, 6, and 7 (42), but no substrates other than caspases have been reported. Since recombinant caspase 10 but not other caspases can cleave IRS-1, our study is the first account of a non-caspase target for caspase 10. It may therefore have a wider role in apoptosis regulation than previously thought.
Caspase 10 Activation via the Inhibition of a Component in the MAPK-signaling Pathway-The mechanism for caspase 10 activation in mammary epithelial cells appears to be different from that in immune cells. We found no evidence for caspase 10 activation, or indeed IRS-1 cleavage, after FasL or TRAIL treatment, even though these death ligands did promote caspase 3 activation and apoptosis. Caspase 10 is therefore not activated through the "extrinsic" pathway in these cells. Caspase 10 and IRS cleavage do not occur via the "intrinsic" mitochondrial pathway either, as caspase 3 is activated following the removal of ECM or growth factors; however, their removal does not activate caspase 10 or induce IRS cleavage. This contrasts with caspase 8, which can be cleaved by caspase 3 under certain apoptotic situations (65). Thus in mammary epithelial cells, caspase 10 is activated through an alternative mechanism. This may involve the MAPK pathway, as inhibition of MKK1 with U0126 leads to cleavage of both caspase 10 and IRS-1. MKK1 is upstream of caspase 10 since z-AEVD-fmk blocks IRS-1 cleavage in the presence of U0126 but does not reverse Erk dephosphorylation. Interestingly, Erk is rapidly dephosphorylated during mammary involution in vivo, although the mechanism of altered Erk signaling has not been established (22).
We do not yet know the nature of the mechanistic link between MKK1 and caspase 10. It may be through a MKK1 substrate that is separate from Erk. Alternatively Erks and some of their downstream effectors might be involved. Preliminary evidence from studies where mammary cells have been cultured in the presence of cycloheximide suggests that IRS-1 cleavage following U0126 treatment requires the expression of new proteins. 2 One possibility currently under investigation is that unidentified caspase-activating proteins or alternatively antagonists of "inhibitors-of-apoptosis proteins" (IAPs), such as Diablo or Omi, may be involved and that an Erk-controlled pathway regulates their levels (66).
It is of interest that the levels of full-length caspase 10 are significantly reduced at lactation, yet IRS-1 is still present at this time. One possibility to explain this is that the epithelial cells become primed to undergo rapid apoptosis during lactation, yet the apoptosis machinery is held in check by inhibitory factors until weaning occurs. Candidates for such inhibitory factors are the inhibitors-of-apoptosis proteins (IAPs) (66), and we are currently examining whether any of these are expressed in the mammary gland and whether their levels or activities are altered at the lactation-involution switch. In our previous analyses of mammary involution mechanisms, we have argued similarly that apoptosis is primed during lactation to allow rapid activation at weaning (27).
Caspase 10 and IGF Signaling in Breast Cancer-Our study has implications for understanding mechanisms of avoidance of apoptosis in cancer and suggests possible novel therapeutic strategies for treatment. Certain carcinomas, including breast cancer, have activated IGF signaling, which is involved with their unscheduled ability to proliferate and survive at metastatic sites (67). Caspase 10 is posttranscriptionally down-regulated in many breast cancers (60), and this may contribute to the maintenance of IGF signaling and therefore the avoidance of apoptosis. Currently the IGF pathway is viewed as an attractive candidate pathway for therapeutic intervention in cancer (68,69). Our results suggest a possible alternative target through the development of agents that either activate or maintain the levels of caspase 10 and that thereby cleave IRS and inhibit IGF-mediated survival.