Detection of the Cell Cycle-Regulated Negative Feedback Phosphorylation of Mitogen-Activated Protein Kinases in Breast Carcinoma using Nanofluidic Proteomics

Mitogen-activated protein kinases (MAPKs) play an important role in the regulation of cell proliferation, oncogenic transformation, and drug resistance. This study examined the capability of nanofluidic proteomics to identify aberrations in the MAPK signaling cascade, monitor its drug response, and guide the rational design of intervention strategies. Specifically, the protein post-translational modification (PTM) profiles of MEK1, MEK2, and ERK1/2 were measured in breast carcinoma and breast cancer cell lines. Nanofluidic proteomics revealed hyper-phosphorylation of MAPKs in breast carcinoma and breast cancer cells treated with kinase inhibitors that interfere with cell cycle regulation, such as dinaciclib, an inhibitor of cyclin-dependent kinases, and rigosertib, an inhibitor of polo-like kinase 1. A pMEK1 (Thr286) phosphor-isoform, which serves as a biomarker of cell cycle-regulated negative feedback phosphorylation in breast cancer cells, was detected in breast carcinoma. Inhibition of the MAPK pathway with dabrafenib, a B-Raf inhibitor, or trametinib, a MEK1/2 inhibitor, suppressed both the positively regulated phosphorylation of MAPKs and the negatively regulated phosphorylation of MEK1. Interestingly, the combinations of dabrafenib and rigosertib or trametinib and rigosertib permitted the suppression of positively regulated MAPK phosphorylation together with the promotion of negatively regulated MEK1 phosphorylation. The effectiveness of protein PTM-guided drug combinations for inhibition of the MAPK pathway remains to be experimentally tested. Via protein PTM profiling, nanofluidic proteomics provides a robust means to detect anomalies in the MAPK signaling cascade, monitor its drug response, and guide the possible design of drug combinations for MAPK pathway-focused targeting.

In the last several decades, cancer treatment has progressively evolved from non-specific cytotoxic chemotherapy toward selective mechanism-based therapeutics 1 . This therapeutic revolution is led by clinical success in cancer treatment via the use of small-molecule kinase inhibitors to target kinases whose mutations drive cancer growth and development 2 . The burgeoning library of molecular targeted drugs that interfere with specific oncogenic abnormalities ushers endless possibilities for cancer therapy 3,4 . However, the realization of molecular targeted cancer therapy is hindered by multiple challenges, such as the fact that only some human cancers have known kinase-domain mutations [5][6][7][8] and the rapid development of drug resistance due to intrinsic inter-and intra-tumor heterogeneity 9,10 .
To overcome such challenges, molecular targeted cancer therapy is being applied more broadly, extending beyond specific oncogenic lesions to encompass aberrant signaling pathways whose components are not necessarily mutated 5 . Furthermore, multi-component therapy with combinations of molecular targeted drugs is being

Results
Detection of protein phosphor-isoforms using nanofluidic proteomics. First, two methods of protein detection, Western blotting and cIEF immunoassay, were deployed to profile MEK1, MEK2, and ERK1/2 proteins in total cell extracts (TCEs) of a breast cancer cell line BT474. TCEs of BT474 were either untreated (−λ) or treated with λ phosphatase (+λ) to remove protein phosphor-isoforms. Western blots separated proteins based on their molecular weights, and changes in the protein migration patterns following λ phosphatase treatment were not detectable (Fig. 1a), a finding that was expected due to the very small change in protein molecular weight associated with a phosphorylation event. By contrast, cIEF immunoassays separated proteins based on charges and could detect changes in the protein migration patterns following λ phosphatase treatment ( Fig. 1b-d). To distinguish ERK1 and ERK2 from their phosphor-isoforms, primary antibodies specific to p-ERK1/2, ERK1, and ERK2 were used for analysis (Fig. 1d, bottom three panels). Alternatively, the cIEF immunoassay data were presented graphically as a function of the isoelectric points versus intensity ( Fig. 1e-j). The isoelectric points corresponding to the protein isoforms are summarized in Table 1. The pI range assignments for protein isoforms were experimentally determined by this study as well as by several previous studies from multiple independent research groups [17][18][19][20]23,31,32 . Compared with size separation by Western blots, charge separation by the cIEF immunoassay was highly sensitive to the analysis of protein isoforms of the MAPK pathway.

Profiling protein isoforms of the MAPK pathway in breast cancer and non-cancerous cell lines.
Next, cIEF immunoassays were deployed to measure the protein isoform profiles of MEK1, MEK2, and ERK1/2 in seven cell lines (Table 2). These cell lines included five breast cancer cell lines (MCF-7, MDA-MB-231, MDA-MB-453, MDA-MB-468, and BT474) and two non-tumorigenic epithelial cell lines of breast tissues (MCF-10A and MCF-12A). Although there was significant variability among the cell lines, MEK1 isoforms were generally distributed among three major populations, MEK1, pMEK1, and ppMEK1 (Fig. 2a). Similarly, MEK2 isoforms were distributed among three major populations, MEK2, pMEK2, and ppMEK2, with pMEK2 being the dominant isoform in all cell lines (Fig. 2b). Expectedly, ERK1 and ERK2 and their phosphor-isoforms pERK1 and ppERK2 were present in all cell lines (Fig. 2c).
Profiling protein isoforms of the MAPK pathway in breast carcinoma. Using tissue biopsies from nine patients with breast carcinoma (Table 3), the protein isoform profiles of MEK1, MEK2, and ERK1/2 were measured using cIEF immunoassays. Interestingly, the protein PTM profiles of the MAPK pathway in breast carcinoma were significantly elevated compared with those in cell lines. In addition to the three MEK1 isoforms of MEK1, pMEK1, and ppMEK1 observed in cell lines, the pppMEK1 isoform was clearly observable in all breast carcinomas (Fig. 2d). On the other hand, MEK2 isoforms were present in two populations of MEK2 and ppMEK2, with ppMEK2 having a prevalence greater than 90% (Fig. 2e). Moreover, the presence of ppERK1, which was absent in the breast cancer cell lines, was detected in breast carcinoma (Fig. 2f). While the ERK2 and ppERK2 isoforms were evenly distributed, only pERK1 and ppERK1, but not ERK1, were observed.
The small-molecule kinase inhibitors induced perturbations in MAPK signaling activity that could be characterized by three distinctive groups. The first group included lapatinib and dabrafenib, which suppressed phosphor-isoforms and promoted the unphosphorylated isoforms of MEK1, MEK2, and ERK1/2 ( Fig. 3b-d). The second group included trametinib, which caused shifts toward lower pI values for MEK1 and MEK2 isoforms and suppressed phosphor-isoforms of ERK1/2 ( Fig. 3e-g). It is important to note that shifts toward lower pI values for MEK1 and MEK2 induced by trametinib were resistant to λ phosphatase treatment (Supplemental Fig. S1a,b), suggesting that these shifts were not due to increased phosphorylation. It is plausible that the changes in the surface charges of MEK1 and MEK2 could be consequences of conformational changes in protein structures due to direct binding to trametinib. Finally, the third group included dinaciclib and rigosertib, which promoted the phosphorylation of MEK1, MEK2 and ERK1/2 and induced the appearance of pppMEK1, ppMEK2, and ppERK1 ( Fig. 3h-j). Interestingly, the protein phosphorylation profiles of the MAPK pathway induced by dinaciclib and rigosertib in MDA-MB-231 cells resembled those of breast carcinoma.
Measuring the effects of drug combinations on the distribution of MAPK pathway protein isoforms.
To understand how distinct regulatory mechanisms interact, MDA-MB-231 cells were treated with combinations   Table 3. Human primary breast carcinoma. of dabrafenib and rigosertib or trametinib and rigosertib. Both cIEF immunoassays and Western blotting were performed to measure the protein phosphor-isoform profiles of the MAPK pathway. cIEF immunoassays revealed that treatment with a combination of dabrafenib and rigosertib reduced the ppMEK1 isoform, increased the ppMEK2 isoform, and suppressed the ERK1/2 phosphor-isoforms (Fig. 5a-c). Treatment with a combination of trametinib and rigosertib caused shifts to lower pI values for both MEK1 and MEK2 and suppressed ERK1/2 phosphor-isoforms (Fig. 5d-f). On the other hand, Western blotting revealed that treatment with a β-actin served as a loading control. Cropped Western blots were from different gels loaded with the same amount of TCEs. All gels were run on the same day and subjected to the same experimental procedures, including the same exposure duration during detection. Representative immunoblots of β-actin are presented to highlight comparable TCE loading between lanes.
The cell cycle-regulated negative feedback phosphorylation of MEK1 in breast carcinoma. cIEF immunoassays and Western blotting were performed to identify specific MEK1 phosphor-isoforms that contribute to the presence of pppMEK1 in selected breast carcinoma samples. Using primary antibodies that recognize MEK1 phosphor-isoforms at pSer217/221, pThr286, pThr292, and pThr386, cIEF immunoassays were performed in selected breast carcinoma samples. A representative set of cIEF immunoassay data is presented in Fig. 6a-d. Briefly, four MEK1 phosphor-isoforms from pMEK1 to ppppMEK1 were detected with a primary antibody specific to pSer217/221 (Fig. 6a). Two MEK1 phosphor-isoforms, pppMEK1 and ppppMEK1, were detected with a primary antibody specific to pThr286 (Fig. 6b). Three MEK1 phosphor-isoforms, ppMEK1, pppMEK1, and ppppMEK1, were detected with a primary antibody specific to pThr292, where the ppMEK1 population was less than 10% of all phosphor-isoforms (Fig. 6c). Three MEK1 phosphor-isoforms, ppMEK1, pppMEK1, and pppMEK1, were detected with a primary antibody specific to pThr386, where the ppMEK1 population comprised up to 40% of all the phosphor-isoforms (Fig. 6d). cIEF immunoassay data indicated that MEK1 phosphor-isoforms at pThr286 and pThr292 were present as pppMEK1 and ppppMEK1. Indeed, analyses of all nine breast carcinoma samples with a primary antibody specific to pThr286 revealed the exclusive presence of pThr286 as pppMEK1 and ppppMEK1 isoforms. Furthermore, Western blotting was performed using primary antibodies specific to pThr286 and pThr292 in four selective breast carcinoma samples. Breast carcinoma samples were selected based on the sufficiency of TCEs for Western blot analyses. Consistent with the cIEF immunoassay data, MEK1 phosphor-isoforms at pThr286 and pThr292 were detected (Fig. 6d).

Discussion
Using cIEF immunoassays, this study reported hyper-phosphorylation of the MAPK pathway in breast carcinoma compared with that in breast cancer cell lines. Specifically, the triple-phosphorylated pppMEK1 isoform and double-phosphorylated ppERK1 isoform were present in breast carcinoma but absent in breast cancer cell lines. In addition, double-phosphorylated ppMEK2 was the dominant isoform in breast carcinoma, whereas mono-phosphorylated pMEK2 was the dominant isoform in breast cancer cell lines. Interestingly, treatment with small-molecule kinase inhibitors that cause cell cycle arrest, such as rigosertib and dinaciclib, promoted hyper-phosphorylation of the MAPK pathway in breast cancer cell lines. Western blot analyses revealed that the hyper-phosphorylation of MEK1 induced by rigosertib treatment was due to increased phosphorylation at negative regulatory sites at amino acid residues Thr286, Thr292, and Thr386. Both cIEF immunoassays and Western blotting confirmed the presence of the pMEK1 (Thr286) phosphor-isoform in selected breast carcinoma samples, indicating cell cycle-regulated feedback phosphorylation. Furthermore, cIEF immunoassays revealed that the pMEK1 (Thr286) phosphor-isoform contributed to the presence of the pppMEK1 isoform in all breast carcinoma samples examined. This study revealed that the protein PTM profile could be used to identify aberrations in a signaling pathway and assist in the design of pathway-focused therapy. For example, the hyper-phosphorylation of MEK1, MEK2, and ERK1/2 in breast carcinoma samples indicated hyper-activity of the MAPK pathway. In addition, the presence of the pMEK1 (Thr286) phosphor-isoform in breast carcinoma indicated negative feedback phosphorylation that could stem from cell cycle dysregulation. When breast cancer cells were treated with dabrafenib or trametinib, both the positively regulated phosphorylation at Ser217/221 and negatively regulated phosphorylation at Thr292 and Thr386 of MEK1 were suppressed. This was likely a consequence of the reduced positively regulated phosphorylation of ERK1/2 at Thr202/Tyr204, leading to inhibition of the negative feedback phosphorylation of MEK1 at Thr292 and Thr386 by ERK1/2. By contrast, treatment of breast cancer cells with rigosertib promoted the negative feedback phosphorylation of MEK1 at Thr286, Thr292, and Thr386 without affecting the positively regulated phosphorylation of MEK1 at Ser217/221 or ERK1/2 at Thr202/Tyr204. Theoretically, inhibition of a pathway could be achieved most effectively by the suppression of positive regulation coupled with the promotion of negative regulation. Treatment of breast cancer cells with the combinations of both dabrafenib and rigosertib and trametinib and rigosertib achieved this design objective. The combination of dabrafenib and rigosertib suppressed the positively regulated phosphorylation of MEK1 at Ser217/221 and ERK1/2 at Thr202/ Tyr204, promoted the negatively regulated phosphorylation of MEK1 at Thr286, and maintained the negatively regulated phosphorylation of MEK1 at Thr292 and Thr386. On the other hand, treatment of breast cancer cells with a combination of trametinib and rigosertib suppressed the positively regulated phosphorylation of ERK1/2 at Thr202/Tyr204 and maintained the negatively regulated phosphorylation of MEK1 at Thr292 and Thr386. Future studies will experimentally investigate the effectiveness of drug combinations guided by the protein PTM profiles of a specific signaling protein for pathway-focused targeting.
Consistent with many previous studies, cIEF immunoassays were highly suitable for the analysis of protein PTM profiles in finite tissue biopsies 45 . Typically, only 40 ng of TCE were used per cIEF immunoassay 46 . By comparison, a 1D Western blot required approximately 10 µg for molecular weight analysis, and 2D Western blotting required approximately 500 µg for both charges and molecular weight analyses 32 . cIEF immunoassays were highly sensitive to the detection of various modes of protein PTM, including phosphorylation, acetylation, and O-linked glycosylation 17,21,[47][48][49][50] . This capability allows cIEF immunoassays to detect changes in the signaling or enzymatic activities of a protein in diseased states or in response to drug treatment. A drawback of the cIEF immunoassay is its inability to identify the specific sites of PTM that provide critical information on the regulation of protein activity. This deficiency is somewhat mitigated by the use of selective antibodies that recognize the specific phosphorylation sites of a protein. However, the availability of these selective antibodies is highly limited. Due to the lack of antibodies that recognize specific phosphorylation sites, which amino acid residues on ppMEK2 or ppERK1 received phosphorylation in breast carcinoma or breast cancer cells following treatment with small-molecule kinase inhibitors that target cell cycle regulation remain unclear. An emerging proteomic technology that couples cIEF with mass spectrometry has the potential to overcome the current limitation of cIEF immunoassays and advance protein PTM profiling capability 51 .
The MAPK signaling cascade plays an important role in the regulation of cell proliferation, oncogenic transformation, and drug resistance 52,53 . Several small-molecule kinase inhibitors that target components of the MAPK signaling cascade have been approved by the US Food and Drug Administration for the treatment of BRAF-mutant melanoma, including trametinib, dabrafenib, and vemurafenib, or the treatment of kidney and liver cancers with sorafenib 1 . However, the efficacy of these kinase inhibitors on other tumors have been somewhat limited 28,54 . Numerous clinical trials are ongoing using combinations of MEK1/2 inhibitors with BRAF inhibitors for the same pathway inhibition or MEK1/2 inhibitors with PI3K inhibitors for dual pathway inhibition 28,30,54 . Nonetheless, there is currently a lack of predictive biomarkers of sensitivity or resistance to kinase inhibitors, consequently hindering the development of a pre-clinical model that can identify the best combinations and predict clinical success 12,13,15,28,30,54 . Nanofluidic proteomics could complement other emerging technology platforms [55][56][57] and model systems 16 that seek to identify drug resistance biomarkers and design effective drug combinations that target MAPK or other oncogenic signaling pathways for personalized anti-cancer therapy.

Methods
Cell lines and drug treatment. All cell lines were obtained from American Type Culture Collection (ATCC, Manassas, VA) and cultured according to a published guideline 58,59 . All small-molecule kinase inhibitors were obtained from Selleckchem (Houston, TX). Cells were treated with kinase inhibitors for 24 hours prior to collection and analysis. The concentrations of kinase inhibitors included 5 µM lapatinib, 10 µM dabrafenib, 0.5 µM trametinib, 0.1 µM dinaciclib, and 1 µM rigosertib, which were chosen based on the EC 50 values determined by our own cytotoxic screening assays.
Human primary breast carcinoma biopsies. De 6 seconds. Homogenates were incubated on ice for 10 minutes, sonicated 4 times for 5 seconds each, rotated at 4 °C for 2 hours, and centrifuged at 12000 rpm in an Eppendorf 5430R microfuge for 20 minutes at 4 °C. Supernatants were collected as TCEs for both Western blotting and cIEF immunoassays. For Western blotting, the protein concentrations of TCEs were estimated using the CB-X protein assay kit (cat. no. 786-12, G-Biosciences, St. Louis, MO), adjusted with 4× SDS sample buffer containing 2-mercaptoethanol, boiled for 5 minutes and used for analysis. For cIEF immunoassays, the TCEs were prepared in a Premix G2 pH 5-8 separation gradient containing pI standards (Protein Simple) and added to 384-well plates for analysis. Treatment with λ phosphatase. Approximately 1 µl of phosphatase (cat. no. 14-405; Merck Millipore, Billerica, MA) was added to 1 µl of reaction buffer (final concentrations of 5 mM DDT, 50 mM Hepes, 100 µM EDTA, 2 mM MnCl 2 ) and 8 µl of TCEs (2 mg/ml). The mixture was incubated at 37 °C for 30 minutes, chilled on ice to stop the reaction, and used for Western blotting or cIEF immunoassays.
Western blotting. TCEs were separated on 10% SDS-PAGE gels, transferred to nitrocellulose membranes, incubated first with the primary antibodies against proteins or protein phosphor-isoforms of interest (Supplemental Table S1) and then with IRDye 680RD secondary antibodies (cat. no. 92668070, LI-COR, Lincoln, NE). Immunoblots were detected using the LI-COR's Odyssey CLx imaging system. Membranes were stripped and re-incubated with antibodies against β-actin, which served as a loading control. cIEF immunoassays. cIEF immunoassays were performed using the NanoPro 1000 system (Protein Simple). Samples of 400-nanoliter volume were separated by isoelectric focusing using the 96-capillary system, followed by immobilization of the proteins onto the inner capillary walls with ultraviolet irradiation. Primary antibodies (Supplemental Table S1) and horseradish peroxidase-conjugated secondary antibodies (cat. no. 7074, Cell Signaling, Danvers, MA) were sequentially introduced into the capillaries, followed by chemiluminescence detection reagents. The incubation times were 110 and 55 minutes for the primary and secondary antibodies, respectively. The separation time was 50 minutes at 15,000 microwatts. On average, 40 ng of TCE was loaded into each capillary, and the standard exposure time during signal detection was 240 seconds. All the cIEF immunoassays were performed with a minimum of four repeats. High fidelity between repeated measurements was consistent with published reports with coefficient of variation values ≤ 0.1 45,46 . Data availability. The authors declare that data supporting the findings of this study are available within the paper and its supplementary information files.