Cell culture media (DMEM, DMEM Glutamax, EMEM, Opti-MEM) and antibiotics (penicillin and streptomycin) were purchased from Invitrogen Co. (Carlsbad, CA). DMSO and Fetal bovine serum (FBS) were purchased from Fisher Scientific (Fair Lawn, NJ) and Denville Scientific Inc. (Metuchen, NJ), respectively. Protein Assay Kit was purchased from Bio-Rad Laboratories (Hercules, CA) while Chemi-luminescence Reagent was purchased from Amersham Biosciences (Piscataway, NJ). Structure and synthesis of experimental compounds CARP-1 Functional Mimetic (CFM)-4, -5, and -16 have been described previously (14). Clinical grade Adriamycin, Cisplatin, 5-fluouracil, Gemcitabine, and Paclitaxel were obtained from the Karmanos Cancer Institute Pharmacy, Wayne State University, Detroit, MI. 5-dimethyltiazol-2-yl-2.5-diphenyl-tetrazolium bromide (MTT) was purchased from Sigma Chemical Co, St. Louis, MO. The affinity purified, anti-CARP-1 (α1 and α2) polyclonal antibodies have been described (9). The affinity purified, anti-phospho-CARP-1 T⁶²⁷ rabbit polyclonal antibodies were generated and characterized on a fee-for-service basis by Kaneka Eurogentec, Seraing-Belgium. Additional, commercially available antibodies utilized in the study are listed in Supplementary Table S1 .

The plasmids for expression of various proteins were either procured from Addgene or generated in this report and are listed in Supplementary Table S2 . All the recombinant plasmids were sequenced to confirm the accuracy and validity of various inserts/epitopes.

Routine maintenance and culture of different cell lines was carried out as described in reports listed for respective cell line in Supplementary Table S3 . MDA-MB-468 cells having CRISPR-based p38γ knock-out were generated and characterized on a fee-for-service basis by Biocytogen Corp., Wakefield, MA. All the cell culture media were also supplemented with 10%FBS, 100units/ml of penicillin, and 100μg/ml of streptomycin, and the cells were maintained at 37°C and 5% CO₂. Generation, characterization, and culture of drug (ADR or Cisplatin)-resistant human TNBC MDA-MB-468 and MDA-MB-231 cells, ADR-resistant murine 4T1 cells, as well as MDA-MB-468 cells expressing reduced CARP-1 have been detailed before (9, 14). For cell growth and MTT studies, the cells were cultured in fresh media with 5%-10% FBS prior to their treatments with various agents.

The stable sublines are listed in Supplementary Table S4 , and were generated by transfecting various plasmids into the MDA-MB-468 and Hela cells followed by selection in the presence of 800μg/ml neomycin using described methods (8, 9, 14). The cell lysates from wild-type, untransfected cells, neomycin-resistant pools, or individual sublines were then subjected to IP and WB analyses as below.

Logarithmically growing cells were seeded in 96 well plate (500-1000 cells/well) and then were either untreated (control) or treated with various agents for noted times. At the end of treatment, cells were incubated with MTT reagent (0.5mg/ml) for 2-4 hours at 37°C, followed by addition of DMSO to solubilize formazan. The plate was read at 570nm in a plate reader, and histograms indicating levels of cell viability were generated by plotting the net absorbances as described (8, 9, 14). For western blot and immunoprecipitation experiments, logarithmically growing cells were either untreated or treated with different agents for various time periods followed by lysis of cells to prepare protein extracts. Western blotting was carried out by electrophoresis of 50-100µg of respective protein lysate on 8-12% SDS-PAGE. For immuno-precipitation (IP), the cell lysates (1mg/sample) were first incubated with appropriate antibodies, and the protein complexes were analyzed by SDS-PAGE as above. The proteins on gels were then transferred to supported nitrocellulose membranes followed by hybridization of membranes with primary antibodies essentially as described (8, 9, 14).

Human breast tumor TMAs were obtained from Dr. George Sandusky, Indiana University Simon Cancer Center Tissue Bank Biorepository, Indianapolis, IN. TMAs were derived from 14 blocks with ~45 patients/block. A total of 504 tumor specimens with non-identifiable patient treatment and survival outcome were analyzed. Immunostaining of the TMA slides was carried out with anti-phospho-CARP-1 antibodies (1:1,500 dilution) as well as with total CARP-1 (α2) antibodies essentially following methods described by us before (8, 9, 13, 14, 16). Staining for phospho-CARP was scored by Dr. Edi Levi, Staff Pathologist and Associate Chief of Staff (Research), John D. Dingell VA Medical Center, Detroit, MI. Nuclear staining was arbitrarily assigned as 1 vs 0 with minimum 20% staining as cut off. Images were taken using Zeiss LSM 510 Meta NLO (63X) essentially as described before (11).

To identify the phosphorylating kinases, we utilized two experimental strategies. In the first approach, a fee-for-service kinase profiling was carried out by SignalChem Lifesciences Corporation, Richmond, BC, Canada. Various protein kinases and their respective substrates were generated in-house at SignalChem using proprietary methods followed by their routine quality control testing of each kinase and its substrate to ensure compliance to acceptable standards. The assay condition for each of the protein kinase were optimized to yield acceptable enzymatic activity and high signal-to-noise ratio. ³³P-ATP was purchased from PerkinElmer. All other materials were of standard grade. The wild-type CARP-1 peptide (EQDEE EKDDG EAKEI STPTH WSKLD PKTMK KK), and its mutant versions (EQDEE EKDDG EAKEI AAPTH WSKLD PKTMK KK) and (EQDEE EKDDG EAKEI AAPAH WSKLD PKTMK KK) were synthesized and reconstituted in dH₂O and used in phosphorylation experiments. Protein kinase assays (in duplicate) were performed at 30°C for 30 min in a final volume of 25µl containing 5µl of diluted active protein kinase, 5µl of stock solution of regular substrate or respective CARP-1 peptide, 5µl of assay buffer, 5µl of protein kinase activator or assay buffer, and 5µl of ³³P-ATP (250µM stock solution, 0.8µCi). The reaction was terminated by spotting 20µl of the reaction mixture onto phosphocellulose P81 plate, followed by washing (x3) for ~15 minutes each in a 1% phosphoric acid solution. The radioactivity on the plate was measured in a scintillation counter. Blank control, which included all the assay components except the addition of the appropriate substrate (replace with equal volume of assay dilution buffer), was set up for each protein kinase. The corrected activity for each protein kinase was determined by removing the blank control value. The corrected activity values for CARP-1 peptides were calculated relative to activity values for normal substrate of respective kinase that were set at 100%.

K-CLASP (Kinase- catalyzed Cross linking and streptavidin purification) method to identify the phosphorylating kinases of a known phosphosite was also used. HeLa cell lysate (500 µg total protein) was combined with protein kinase buffer (10% v/v, New England Biolabs, B6022S). Kinase-catalyzed crosslinking was initiated by adding ATP-ArN₃ (5 mM; synthesized as previously described; 23) and either N-biotin wild type CARP-1 (614–638) (biotin-EEEKDDGEAKEISTPTHWSKLDPKT, synthesized by ProImmune) or N-biotin mutant CARP-1 (614-638; S⁶²⁶,T⁶²⁷/AA) (biotin-EEEKDDGEAKEIAAPTHWSKLDPKT, synthesized by ProImmune) peptides (1 mM) in a total volume of 100 µL. The crosslinking reactions were incubated at 31°C for 2 hours with shaking at 300 rpm with or without UV irradiation at 365nm. Excess N-biotin peptide and endogenous biotin were removed using Amicon Ultra-0.5 centrifugal filters (0.5 mL, Millipore, UFC500396). The filtered samples (200 µL) were then incubated with prewashed (200µL of phosphate-binding buffer three times) streptavidin resin (400 µL bead slurry, Genscript) for 10 minutes at room temperature with rotation. The bound resin was collected by centrifuging (500 rcf, 1 min, RT) with subsequent ten washes with phosphate-binding buffer (200µL) and then four washes with water (200 µL). Proteins were eluted by boiling in 2% SDS in water (200 µL) for 8 minutes and desalted using 10% SDS-PAGE in preparation for LC-MS/MS analysis.

For LC-MS/MS, digested peptides were separated by reverse-phase chromatography under acidic conditions (0.1% formic acid) using an EASY nLC-1000 UHPLC system (Thermo). Peptides were next analyzed on a Q-Exactive mass spectrometer (Thermo). MS1 profiling was carried out over a 375-1600 m/z range at a resolution of 70,000. MS2 fragmentation was performed using higher energy collision-induced dissociation (HCD) on the top 15 ions using a 1.6 m/z window and normalized collision energy of 29. Dynamic exclusion was turned on (15 s). MS raw data were processed using MaxQuant (version 1.5.2.8) against a human protein database from UniProt (downloaded 2016.04.07, 20159 entries). Searches included up to 2 missed tryptic cleavages. Mass tolerances for parent ions were 20 ppm for the first search and 4.5 ppm for the second search and 20 ppm for fragment ions. The S5 iodoacetamide derivative of cysteine was specified as a fixed modification. Oxidation of methionine and acetylation of protein N-termini were set as variable modifications. Minimum protein and peptide identification probabilities were set at ≤1% false discovery rate (FDR) as determined by a reversed database search, and proteins required just 1 unique peptide. All other parameters were left at their default settings. A total of 737 proteins were observed ( Supplementary Table S5 ). Nonspecifically bead bound proteins were removed by requiring higher intensities in the CARP-1 (614–638) peptide sample with UV light sample compared to without UV light. Among the remaining proteins, possible kinases and associated proteins were selected by generating an enrichment value by dividing the protein intensities from the CARP-1 (614–638) peptide with ATP-ArN₃ and UV light sample by the CARP-1 (614-638; S⁶²⁶,T⁶²⁷/AA) with ATP-ArN₃ and UV light sample from each trial. If the enrichment values of a protein were ≥ 1.2 in at least 3 out of 4 trials, that protein was considered a hit ( Supplementary Table S6 ).

The statistical analyses were performed using Prism 6.0 software. The data were expressed as mean ± SEM and analyzed using two-tailed student t-test or one-way ANOVA followed by a post hoc test. A p value of <0.05 was considered statistically significant.

Our prior and on-going studies have indicated that DNA damage-inducing genotoxic agents such as ADR, Cisplatin, or CFM compounds promote apoptosis in part by stimulating levels of CARP-1, phosphorylated SAPKs p38α/β and JNK1/2, cleaved PARP, and activated caspases-3, and -8 (7–16). Knockdown of CARP-1 or pharmacological inhibition of caspase-3 and -8 abrogated apoptosis by ADR, CFM compounds, or EGFR TKIs (7, 9, 11, 14, 16). Moreover, expression of CARP-1 with in-frame deletion of amino acids 600-650 (CARP-1 Δ600-650) also abrogated apoptosis by ADR and CFM compound (11). CARP-1 600-650 region also contains highly conserved, putative consensus TP epitope (T⁶²⁷P⁶²⁸) for the proline-directed MAPK/SAP kinases ( Figure 1A ), and a high number of hits for potential phosphorylation of T⁶²⁷ are also indicated at the phosphosite plus platform (22). CARP-1 S⁶²⁶ and T⁶²⁹ sites also had a few hits for phosphorylation, albeit these amino acids were also highly conserved in vertebrates with the exception of S⁶²⁶ in the rodent (mouse and rat) CARP-1 proteins. Therefore, we first investigated whether CARP-1 S⁶²⁶, T⁶²⁷, and/or T⁶²⁹ residues regulated signaling by DNA damage-inducing agents. We generated and characterized stable, neomycin-resistant human breast cancer (HBC) MDA-MB-468 and cervical cancer Hela cells that express myc-His-tagged CARP-1 protein where S⁶²⁶, T⁶²⁷, and/or T⁶²⁹ were substituted to alanines (CARP-1 S⁶²⁶T⁶²⁷/AA and CARP-1 S⁶²⁶T⁶²⁷T⁶²⁹/AAA) as detailed in Methods ( Figure 1B ; Supplementary Figures S1A, C ). ADR or CFM-4 treatments resulted in generally higher viabilities of cells expressing myc-His-CARP-1 S⁶²⁶T⁶²⁷/AA or myc-His-CARP-1 S⁶²⁶T⁶²⁷T⁶²⁹/AAA when compared with their myc-His-CARP-1 expressing counterparts ( Figures 1C, D ; Supplementary Figures S1B, D ). Increased viabilities of ADR or CFM-4-treated myc-His-CARP-1 S⁶²⁶T⁶²⁷T⁶²⁹/AAA cells was due in part to diminished expression of caspases-2, -3, -8, and -9 ( Figure 1E ) as well as diminished activation of SAPK JNK1/2 ( Supplementary Figures S2A, B ).

We next investigated the kinases that phosphorylate CARP-1 S⁶²⁶, T⁶²⁷, and/or T⁶²⁹ amino acid(s). In the first instance, we conducted in vitro kinase assays utilizing CARP-1 (611–640) wild-type or CARP-1 (611–640) peptides with S⁶²⁶T⁶²⁷/AA or S⁶²⁶T⁶²⁷T⁶²⁹/AAA substitutions in conjunction with various active serine/threonine protein kinases as detailed in methods. The initial profiling experiment indicated possible phosphorylation of the wild-type but not mutant peptides by ERK1, JNK1, CDK1/Cyclin B1, and CDK1/Cyclin A1 ( Supplementary Table S7 ). As CARP-1 T⁶²⁷ is followed by a proline, and the TP motif is preferred site of protein phosphorylation by MAPK/SAPKs, a subsequent kinase profiling experiment was performed by utilizing the above peptide substrates in conjunction with various active kinases that included respective MAPK/SAPK isoforms. The data revealed phosphorylation of the wild-type but not mutant peptides by ERK1, ERK2, JNK1, JNK2, JNK3, P38γ, and p38δ but not by p38α or p38β ( Table 1 ). Of note is a rather robust phosphorylation of CARP-1 (611–640) wild-type peptide by p38γ. We then investigated whether and to the extent different MAPK/SAPK isoforms function to regulate signaling by DNA damage-inducing agents. We first generated and characterized multiple, stable neomycin-resistant MDA-MB-468 HBC and HeLa cervical cancer sublines that express catalytically-inactive MAPK/SAPK isoforms (ERK1 AEF; JNK1a1 APF; JNK2a2 APF; p38δ AGF; p38γ APF; and p38γ AGF) as described in methods ( Supplementary Figures S3A-C, F, G, K, L ). In addition, we generated and characterized MDA-MB-468 HBC and HeLa cervical cancer sublines that express anti-sense cDNAs of p38δ and p38γ to knock-down the respective kinase proteins ( Supplementary Figures S3D, E, H, I ). MDA-MB-468 HBC and HeLa cervical cancer cells (wild-type, vector transfected, or transfected with catalytically-inactive and antisense cDNAs of above MAPK/SAPK isoforms) were then treated with DNA damage inducing agents ADR, Etoposide, and CFM-4, -5, and -4.16 compounds, followed by determination of cell viabilities by MTT assays as noted in methods. The cells expressing catalytically-inactive JNK1a1, JNK2a2, p38δ and p38γ as well as the sublines with knockdown of p38δ and p38γ were resistant to inhibitory effects of DNA damage-inducing compounds when compared with their wild-type or vector expressing counterparts ( Table 2 ; Supplementary Figures S4B-F , S5B-F ). However, the cells expressing catalytically inactive ERK1 were found to be more sensitive to inhibitory effects of DNA damage-inducing compounds when compared with their wild-type or vector expressing counterparts ( Table 2 ; Supplementary Figures S4A , S5A ). These data indicate that DNA damage signaling likely activates ERKs to promote cell survival and consequent resistance. The genotoxic agents activate JNKs and p38 SAPKs to promote cell growth inhibition.

To clarify the MAPK/SAPK that functions to phosphorylate CARP-1 in the presence of genotoxic stress, we first generated and characterized rabbit polyclonal antibodies utilizing T⁶²⁷-phosphorylated CARP-1 peptide as antigen as described in methods. The serum from rabbit with high titer was affinity purified, and utilized in subsequent studies. The WB analyses showed a robust CARP-1 T⁶²⁷ phosphorylation in cell lysates derived from MDA-MB-468 HBC cells expressing myc-His-CARP-1 (WT) that were treated with ADR, Etoposide, CFM-4, or CFM-4.16 compounds when compared with untreated control cells ( Figures 2A, B ). No CARP-1 phosphorylation was detectable in the protein lysates that were derived from MDA-MB-468 HBC cells expressing myc-His-CARP-1 (S⁶²⁶T⁶²⁷/AA) that were treated with either ADR, Etoposide, or CFM compounds ( Figures 2A, B ). Moreover, a time-dependent increase in phosphorylation of MKK4, JNK1/2, and ERK1/2 kinases as well as STAT3 protein occurred in ADR-treated MDA-MB-468 HBC cells expressing myc-His-CARP-1 (WT), while phosphorylation of JNK1/2 and STAT3, but not ERK1/2 or MKK4, was diminished in MDA-MB-468 HBC cells expressing myc-His-CARP-1 (S⁶²⁶T⁶²⁷/AA) that were treated with ADR ( Figure 2C ). These data suggest that ADR-induced cell growth and survival signaling by JNK1/2 and STAT3 likely involves CARP-1 phosphorylation.

Since our kinase assays ( Table 1 ) indicated phosphorylation of CARP-1 (611–640) peptide by JNK1/2/3, ERK1/2, and p38γ, p38δ but not p38α/β kinases, we next determined the kinase(s) that target CARP-1 phosphorylation in the presence of ADR. We first conducted a high-throughput analyses of proteins interacting with CARP-1 epitope surrounding T⁶²⁷. We carried out Kinase- catalyzed Cross Linking and Streptavidin Purification (K-CLASP) (23–25) with an aim to identify the kinases and associated proteins that interact with CARP-1 T⁶²⁷ phosphosite. ADR-treated HeLa cell lysates were incubated with biotin-labelled CARP-1 (614–638) or CARP-1 (614-638; S⁶²⁶, T⁶²⁷/AA) and ATP-arylazide (ATP-ArN3) as a kinase co-substrate to crosslink the phosphopeptides with kinases and interacting proteins covalently with UV light as detailed in Methods. Peptide-bound complexes were then purified using streptavidin resin and interacting proteins were identified using liquid chromatography-tandem mass spectrometry (LC-MS/MS). We found 252 possible kinases and associated proteins that were enriched at least 1.2-fold in their interaction with CARP-1 (614–638) versus CARP-1 (614-638; S⁶²⁶,T⁶²⁷/AA) peptide in 3 out of 4 trials. A list of selected target proteins that interacted with CARP-1 (614–638) peptide is shown in Table 3 . Of note is the fact that CCAR2 (DBC1; a CARP-1 paralog), Filamin C, and BRAF have been demonstrated to interact with CARP-1 in prior studies by us and others (7, 16, 26, 27). CARP-1 (614–638) peptide also interacted with MAPK12/p38γ suggesting that MAPK12/p38γ could be the kinase among the MAPKs noted in Table 1 above that functions to transduce CARP-1 T⁶²⁷ phosphorylation by ADR.

We then validated binding of various signal transducers that interacted with and/or phosphorylated CARP-1 T⁶²⁷. We conducted co-immunoprecipitation-Western blot analyses utilizing lysates from HBC cells that stably expressed myc-His-tagged CARP-1, CARP-1 (S⁶²⁶T⁶²⁷/AA), or CARP-1 (Δ600-650) proteins as detailed in Methods. As shown in Supplementary Figures S6A-C , CARP-1 interacted with JNK1/2, p38γ, p38δ, ERK1/2, MEK1/2, H2AX, and STAT3 proteins. Moreover, as summarized in Table 4 , CARP-1 (WT) and CARP-1 (S⁶²⁶T⁶²⁷/AA) but not CARP-1 (Δ600-650) interacted with JNK1/2, H2AX, and Filamin C. However, only CARP-1 (WT) protein interacted with ERK1/2, RIPK1, CDK4/6, MEK1/2, STAT3, p38γ, and p38δ. Neither CARP-1 (WT) nor its mutant proteins interacted with MKK4, MSK2, and p38α/β ( Table 4 ). As also expected, ADR treatment caused a robust phosphorylation of CARP-1 but not its S⁶²⁶T⁶²⁷/AA mutant protein ( Supplementary Figure S6D ). Of note is that CARP-1 interactions with STAT3 and ERK1/2 were appreciably diminished in ADR-treated cells when compared with their interactions in the untreated control cells ( Supplementary Figure S6D ). Since ADR-induced CARP-1 phosphorylation at T⁶²⁷ while substitution of CARP-1 S⁶²⁶T⁶²⁷ to alanines interferes with CARP-1 interactions with STAT3, RIPK1, and ERK1/2 ( Supplementary Figures S6D, E ), it remains to be determined whether phosphorylation of CARP-1 T⁶²⁷ or substitution of CARP-1 T⁶²⁷ to alanine induces a conformational change that results in diminished interaction of CARP-1 with STAT3, RIPK1, and ERK1/2 proteins. Moreover, given that ADR-treated cells expressing CARP-1 (S⁶²⁶T⁶²⁷/AA) or CARP-1 (S⁶²⁶T⁶²⁷T⁶²⁹/AAA) have increased viabilities when compared with their counterparts expressing wild-type CARP-1 ( Figure 1D , Supplementary Figures S1B, D ), and ADR treatment also caused robust phosphorylation of MAPKs ERK1/2 ( Figure 3A ), it is possible that ADR-induced cell growth and survival signaling by MAPK pathway is regulated in part by P38γ-mediated CARP-1 phosphorylation. Whether and to the extent, ADR-induced, p38γ-mediated CARP-1 phosphorylation is also involved in activation of CDK4/6 to regulate cell proliferation remain to be clarified.

Next, we treated wild-type MDA-MB-468 HBC cells with ADR, pan-p38 inhibitor Doramapimod (28), JNK inhibitor JNK-1N8 (11, 29), or ERK inhibitor Ulixertinib (16, 30) as single agents or ADR in combination with respective MAPK/SAPK inhibitor as detailed in Methods. The cell lysates were analyzed by WB for presence of T⁶²⁷ phosphorylated CARP-1. This experiment revealed that Doramapimod, but not JNK-IN8 or Ulixertinib, abrogated ADR-induced CARP-1 phosphorylation ( Figure 3A ). Since p38α and β isoforms failed to phosphorylate CARP-1 (611–640) peptide in kinase assays, we then determined whether ADR-dependent CARP-1 phosphorylation involved p38γ and/or p38δ isoforms. We utilized MDA-MB-468 HBC cells that have stable expression of vector, catalytically inactive p38γ (p38γ AGF), or catalytically inactive p38δ (p38δ AGF) that are described above ( Supplementary Figures S3F, K ). Although ADR provoked a robust increase in CARP-1 phosphorylation in vector-expressing HBC cells, ADR-dependent CARP-1 T⁶²⁷ phosphorylation was diminished in HBC cells expressing catalytically inactive p38γ but not in HBC cells expressing catalytically inactive p38δ ( Figures 3B, C ). These data suggest that p38γ transduces ADR-dependent CARP-1 phosphorylation, and would be consistent with a robust phosphorylation of CARP-1 (611–640) peptide by p38γ that was noted in the kinase assays summarized in Table 1 . We next investigated whether p38γ was required for genotoxic chemotherapy-induced CARP-1 phosphorylation. For this purpose, we utilized p38γ-null mouse embryonic fibroblasts (MEFs, 31) that were generously provided by Dr. Ana Cuenda, (Department of Immunology and Oncology, Centro Nacional de Biotecnología (CSIC), Madrid, Spain). In addition, we generated and characterized p38γ-null MDA-MB-468 HBC cells as described in methods. As expected, ADR induced CARP-1 phosphorylation in MDA-MB-468, MDA-MB-231, and 4T1 Triple-negative breast cancer cells ( Figures 4A, C ). Absence of p38γ in MDA-MB-468 HBC cells abrogated genotoxic therapy-induced CARP-1 phosphorylation, demonstrating p38γ requirement for CARP-1 T⁶²⁷ phosphorylation by genotoxic chemotherapy. Moreover, genotoxic chemotherapies induced CARP-1 phosphorylation in additional HBC cells, human mammary epithelial cells (HMECs), human renal epithelial HK2 cells (13), as well as murine C2C12 myoblast cells (32), HC11 murine breast epithelial cells (33), and W0069 BRCA-deficient mouse mammary tumor cells (34) but failed to promote CARP-1 phosphorylation in wild-type murine MEFs as well as murine cardiomyocytes ( Figures 4B, C ). The reason(s) for this intriguing absence of ADR-induced CARP-1 phosphorylation in murine MEFs and cardiomyocytes remain to be clarified. Nevertheless, our data in Figures 1 - 4 collectively suggest that SAPK p38γ-mediated CARP-1 T⁶²⁷ phosphorylation functions to transduce apoptosis signaling by genotoxic chemotherapy.