Sphingosine Kinase 1 Isoform-Specific Interactions in Breast Cancer

Abstract

Sphingosine kinase 1 (SK1) is a signaling enzyme that catalyzes the formation of sphingosine-1-phosphate. Overexpression of SK1 is causally associated with breast cancer progression and resistance to therapy. SK1 inhibitors are currently being investigated as promising breast cancer therapies. Two major transcriptional isoforms, SK143kDa and SK151kDa, have been identified; however, the 51kDa variant is predominant in breast cancer cells. No studies have investigated the protein-protein interactions of the 51kDa isoform and whether the two SK1 isoforms differ significantly in their interactions. Seeking an understanding of the regulation and role of SK1, we used a triple-labeling stable isotope labeling by amino acids in cell culture-based approach to identify SK1-interacting proteins common and unique to both isoforms. Of approximately 850 quantified proteins in SK1 immunoprecipitates, a high-confidence list of 30 protein interactions with each SK1 isoform was generated via a meta-analysis of multiple experimental replicates. Many of the novel identified SK1 interaction partners such as supervillin, drebrin, and the myristoylated alanine-rich C-kinase substrate-related protein supported and highlighted previously implicated roles of SK1 in breast cancer cell migration, adhesion, and cytoskeletal remodeling. Of these interactions, several were found to be exclusive to the 43kDa isoform of SK1, including the protein phosphatase 2A, a previously identified SK1-interacting protein. Other proteins such as allograft inflammatory factor 1-like protein, the latent-transforming growth factor β-binding protein, and dipeptidyl peptidase 2 were found to associate exclusively with the 51kDa isoform of SK1. In this report, we have identified common and isoform-specific SK1-interacting partners that provide insight into the molecular mechanisms that drive SK1-mediated oncogenicity.

Sphingosine kinase 1 (SK1) is an important signaling enzyme that catalyzes the phosphorylation of the lipid sphingosine into sphingosine-1-phosphate (S1P). The relative levels of these two signaling lipids have a significant role in a number of cellular processes. Cell proliferation, survival, apoptosis, adhesion, and migration (1–3) have all been implicated as subject to influence by this signaling pathway. SK1 is associated with many diseases such as inflammatory diseases (asthma), atherosclerosis, hypertension, and cancer (4–6).ve

The sphingosine kinase system is intrinsically involved in modulating growth factor signaling in normal breast epithelial cells (7). Of particular interest is the tumorigenic potential of SK1, supported by the observations that elevated levels of SK1 have been found in a variety of human cancer cells (8–13), and tumor growth can be inhibited by SK1 inhibitors (14). In the past decade, SK1 has been found to have tumorigenic effects in breast cancer, specifically in promoting the growth and estrogen responsiveness of breast cancer cells (15–17). SK1 acts as an estrogen signaling regulator in breast cancer (18), coupling hormone receptor action with downstream functional actuators (19). Not only is SK1 involved in enhancing breast cancer cell proliferation, recent evidence suggests that endocrine resistance is causally associated with the overexpression of SK1 (20) and that the mechanism may be an overreliance on aberrant SK1 signaling. Furthermore, SK1 may be a predictive factor for neoadjuvant breast cancer treatment (21). Recently the spatial organization of SK1 and the assessment of clinical outcome in estrogen receptor-positive breast tumors have revealed an association between SK1 subcellular localization and prognostic outcome (22). Thus, sphingosine kinase inhibitors are currently being investigated as promising new breast cancer therapies (23, 24).

Two major isoforms of human SK1 have been identified, both transcribed from the same gene. These are 384 and 470 amino acids long, with predicted masses of approximately 43 kDa (SK1a) and 51 kDa (SK1b) (Uniprot database accessions Q9NYA1 and Q9NYA1–2). The first human SK1 isoform, 43 kDa, was cloned and characterized in 2000 by Pitson et al (25). In 2004 a second transcriptional isoform of SK1 was identified and found to be 86 amino acids longer at the N terminus compared with the conventional sequence (26). The longer 51-kDa isoform was also found to be present in breast cancer. The question of the different interaction capacities of the two SK1 isoforms has yet to be addressed because studies to date have exclusively considered the conventional 43-kDa isoform interactions (27–32), albeit these two variants have different intrinsic properties, which influences their interaction with SK inhibitors (33).

There is evidence in the literature demonstrating significant differences in the mechanism by which SK1a and SK1b are processed by the cell in response to SK1 inhibitors which demonstrates that SK1b evades the proteasome in response to sphingosine kinase inhibitor in androgen-independent prostate cancer cells, whereas SK1a is still susceptible (34). Numerous studies have attempted to identify regulatory or signaling components of SK1 activity in cancer development and progression. These studies have chiefly focused on identifying protein interaction partners of SK1 through screening with the yeast two-hybrid system (29, 30). Although several SK1 interaction candidates have been identified via this approach, with validation of the physiological interaction being performed in a variety of human and mouse cell lines, none of this work has been performed in breast epithelial cells. Because this is where most breast cancers occur, it presents a significant confounding factor in attempts to resolve the mechanisms of SK1's functional effects in this cancer.

Stable isotope labeling by amino acids in cell culture (SILAC) is a mass spectrometry (MS)-based technique, which, in conjunction with traditional immunoprecipitation methods, offers a powerful means for identifying and quantifying specific protein-protein interactions (35). Critically, SILAC is amenable to multiplexing (36), which presents the possibility of studying multiple protein variants simultaneously.

In this study a comprehensive analysis of SK1 interaction partners has been undertaken in the estrogen receptor-positive MCF-7 epithelial breast cancer cell line. To identify the interaction partners of SK1 and better understand the functional role of this protein in breast cancer, a FLAG-based affinity purification approach was used in conjunction with a SILAC quantitative mass spectrometric analysis. This enabled the identification of high-confidence SK1 interaction candidates while controlling for nonspecific binding proteins and the simultaneous direct quantitative comparison of each SK1 isoform's protein interaction partners.

Materials and Methods

Materials

Lysine-4 (²H_4-lysine 2HCl), arginine-6 (¹³C₆-L-arginine HCl), lysine-8 (¹³C₆¹⁵N₂-L-lysine HCl), and arginine-10 (¹³C₆¹⁵N₄-L-arginine HCl) as well as lysine- and arginine-free DMEM were purchased from Silantes GmbH. Dialyzed fetal calf serum was purchased from Invitrogen (26400-036). Anti-FLAG antibodies and anti-FLAG M2 affinity gel were purchased from Sigma (F3165 and A2220). Horseradish peroxidase-conjugated secondary antibodies were obtained from Jackson ImmunoResearch. Sequencing grade-modified porcine trypsin was from Promega.

Cloning of FLAG-SK43 and FLAG-SK51 constructs

Previously described constructs were used for transcriptional sphingosine kinase isoforms of 43 kDa (SK43) (25). For transcriptional sphingosine kinase isoforms of 51 kDa (SK51), RNA was isolated from human umbilical vein epithelial cells (HUVECs) using TRIzol reagent (Invitrogen; 15596–018) according to the manufacturer's protocol. RNA was then reverse transcribed to cDNA using a high-capacity cDNA reverse transcription kit (Applied Biosystems; 4368814). Primers were designed to amplify cDNA encoding the 51-kDa SK1 protein (NM_182965.1).

The forward primer (5′-tagaattcgccaccatgtccgctcaagttctggga-3′) incorporated an EcoRI restriction site sequence upstream of a Kozak consensus sequence to initiate translation. The reverse primers [5′-TACTCGAGACTTGTCATCGTCGTCCTTGTAGTCTAAGGGCTCTTCTG-3′ (FLAG tagged)] were designed with an XhoI restriction site sequence. PCR amplification was performed using a KAPA HiFi PCR kit (KAPA Biosystems; KK201) and PCR products visualized on a 1% agarose gel stained with ethidium bromide. PCR products were excised and purified from agarose using a MO-BIO UltraClean 15 DNA purification kit (MO BIO; 12100–300). Vector pcDNA3.1zeo (+) was prepared for cloning by digesting with XhoI and EcoRI. After agarose gel electrophoresis, PCR fragments were digested with EcoRI and XhoI before being extracted and purified. PCR fragments were ligated into vector pcDNA3.1zeo using T4 DNA ligase (New England Biolabs) overnight at 16°C. Sequence was confirmed via sequencing and restriction digest analysis.

Sphingosine kinase fluorescent activity assay

The sphingosine kinase assays were performed as described (37). Briefly, cells were lysed by a single freeze-thaw process in sphingosine kinase buffer containing 50 mM HEPES (pH 7.4), 15 mM MgCl₂, 10 mM KCl, 0.1% Triton X-100, 20% glycerol, 2 mM dithiothreitol, 10 mM NaF, 2 mM orthovanadate, 1 mM deoxypyridoxine, and EDTA-free complete protease inhibitor (Roche) followed by clearing at 21 800 × g for 15 minutes. Five micrograms of total lysate were used. Total sphingosine kinase activity was measured in a reaction buffer containing 50 mM HEPES (pH 7.4), 15 mM MgCl₂, 10 mM KCl, 10% glycerol, 5 mM NaF, 2 mM ATP, 1 mM deoxypyridoxine, and 10 μM 7-nitro-2–1,3-benzoxadiazol-4-yl (NBD)-sphingosine as substrate. SK1 activity was determined in 50 mM HEPES (pH 7.4), 15 mM MgCl₂, 0.5% Triton X-100, 10% glycerol, and 2 mM ATP. Kinase reactions were initiated with the addition of 5 μg of protein lysate in a 50-mL reaction volume incubated at 35°C for 30 minutes. S1P was extracted by adding 50 μL of 1 M potassium phosphate (pH 8.5), followed by 250 μL of chloroform/methanol (2:1) and cleared at 15 000 × g for 1 minute. One hundred microliters of the upper aqueous phase were removed and mixed with 100 μL of dimethylformamide for the detection of NBD fluorescence. The blank sample contains no enzyme. The fluorescence units were converted to nanomoles of S1P using NBD-S1P standards extracted as described above. NBD sphingosine and NBD-S1P were from Avanti Polar Lipids. HUVECs with and without TNFα were used as controls.

Immunofluorescent imaging

Cells were plated on coverslips at 2 × 10⁵ cells/well plates and left for 48 hours. Cells were fixed and stained as described previously (38). FLAG antibody was used for SK1 detection, and Alexafluor 488 and 594 and Hoechst 33342 were used for visualization. Images were taken on a Nikon inverted fluorescent confocal microscope at ×60 oil magnification.

Stable isotope labeling and cell culture

MCF-7 epithelial breast cancer cells (American Type Culture Collection 30–4500K) were maintained in complete medium: DMEM supplemented with 10% (vol/vol) fetal bovine serum and penicillin/streptomycin (50 U/mL). Cells were mycoplasma free tested using the MycoAlert kit (Lonza). MCF-7 cells were transfected with an empty vector (EV), FLAG-SK43, or FLAG-SK51 plasmids, and positive pooled clones were selected with 500 μg/mL zeocin. Cells were then cultured and split at a 1:5 dilution into one of three SILAC media formulations. For double-labeling experiments, EV-expressing cells were cultured in media containing normal isotopic abundance lysine (146 mg/L) and arginine (84 mg/L), whereas FLAG-SK43- and FLAG-SK51-expressing cells were cultured in ¹³C₆¹⁵N₂-lysine (152 mg/L) and ¹³C₆¹⁵N₄-arginine (88 mg/L).

For the triple-labeling experiments, EV-expressing cells were cultured in media containing normal isotopic abundance lysine (146 mg/L) and arginine (84 mg/L), whereas FLAG-SK43-expressing cells were cultured in ²H₄-lysine (150 mg/L) and ¹³C₆-arginine (86 mg/L), and FLAG-SK51-expressing cells were cultured in in ¹³C₆¹⁵N₂-lysine (152 mg/L) and ¹³C₆¹⁵N₄-arginine (88 mg/L). Cells were split three more times before harvesting, achieving close to 100% label incorporation (verified by MS analysis). On harvesting, cells were washed three times in cold PBS and then scraped in cold lysis buffer [1% (vol/vol) Nonidet P-40, 150 mM NaCl, 50 mM Tris-HCl (pH 7.4)] with protease inhibitors (Sigma; P3840) and phosphatase inhibitors (5 mM EDTA, 20 mM NaF, 1 mM Na₃VO₄, 5 mM Na₄P₂O₇). The lysate was cleared by centrifugation at 18 000 relative centrifugal force (rcf) for 15 minutes at 4°C, and protein concentration was measured by the Bradford method (Bio-Rad Laboratories).

Inhibition of dipeptidyl peptidase 2 (DPP2) and dipeptidyl peptidase 4 (DPP4) using Ala-BoroPro inhibitor

The dipeptidyl peptidase inhibitor Ala-BoroPro (39) was maintained at a concentration of 100 mM in buffer (pH 2) at −20°C and used at a concentration of 1 nM to 100 μM at pH 7 at 37°C. Cells were incubated from 1 to 4 hours, and SK51kDa and SK43kDa protein expressions were compared by a Western blot using an anti-FLAG mouse antibody (Sigma). MG132 and cyclohexamide (CHX) were used at concentrations of 10 μM. a trypan blue exclusion assay was used to determine the cell proliferation after the treatment with Ala-BoroPro (100 nM).

Immunoprecipitations and Western blots

One milligram of protein from nonlabeled FLAG-SK43 and FLAG-SK51 cell lysate was incubated with 50 μL of anti-FLAG M2 affinity gel for 2 hours at 4°C. The beads were collected by centrifugation at 1000 rcf for 2 minutes at 4°C and washed three times with 700 μL of wash buffer [1% (vol/vol) Nonidet P-40, 150 mM NaCl, 50 mM Tris (pH 7.4)]. The beads were then resuspended in 200 μL of competitive elution buffer containing 150 ng/μL FLAG competitor peptide in 150 mM NaCl, 50 mM Tris (pH 7.4) and incubated with shaking for 15 minutes. The supernatant was then extracted and the elution performed again, with the second eluate added to the first. Eluates were then concentrated using 3 kDa MWCO filters (Millipore), separated by 1-D SDS-PAGE and immunoblotted with anti-FLAG antibodies using established methods.

SILAC coimmunoprecipitations for MS analysis

For double-labeling experiments, 1.5 mg of protein from labeled FLAG-SK51- or FLAG-SK43-expressing cell lysates was combined 1:1 with 1.5 mg of protein from EV-expressing cells, and the combined lysates were immunoprecipitated as above with 80 μL anti-FLAG affinity gel. The same procedure was performed with different label configurations for triple-labeling experiments, and the lysates were combined 1:1:1. For post-IP mix triple-labeling experiments, 1.5 mg of protein from each labeled cell line was immunoprecipitated in isolation with 50 μL anti-FLAG affinity gel, and the products from each independent immunoprecipitation were then combined 1:1:1 after elution.

Gel liquid chromatography and tandem mass spectrometry (LC-MS/MS) and data analysis

Anti-FLAG coimmunoprecipitation products were concentrated using 3 kDa MWCO filters (Millipore) and then separated by SDS-PAGE on 4%–12% gradient gels and stained with blue silver (40). Lanes were cut into 10 pieces across the mass range, reduced, and alkylated prior to tryptic digestion (41), and extracted peptides were then separated by nano-LC using an Ultimate 3000 HPLC and autosampler system (Dionex). Samples (2.5 μL) were concentrated and desalted onto a micro-C18 precolumn (500 μm × 2 mm; Michrom Bioresources) with H₂O-CH₃CN [98:2 (vol/vol), 0.1% trifluoroacetic acid] at 15 μL/min. After a 4-minute wash, the precolumn was switched (Valco 10 port valve; Dionex) into line with a fritless nanocolumn (75 μ × ∼10 cm) containing C18 media (5 μ, 200 Å Magic; Michrom) manufactured according to Gatlin et al (42). Peptides were eluted using a linear gradient of H₂O-CH₃CN [98:2 (vol/vol), 0.1% formic acid] to H₂O-CH₃CN [64:36 (vol/vol), 0.1% formic acid] at 250 nL/min over 30 minutes. Two thousand volts were applied to the low volume tee (Upchurch Scientific) and the column tip positioned approximately 0.5 cm from the heated capillary (T = 275°C) of an Orbitrap Velos ETD (Thermo Electron) mass spectrometer. Positive ions were generated by electrospray and the Orbitrap operated in data-dependent acquisition mode. A survey scan mass to charge ratio (m/z) of 350–1750 was acquired in the Orbitrap (Resolution = 30 000 at m/z 400, with an accumulation target value of 1 million ions) with lock mass enabled. Up to the 10 most abundant ions (>5000 counts) with charge states of +2 or greater were sequentially isolated and fragmented within the linear ion trap. Collisionally induced dissociation was used with an activation q = 0.25 and activation time of 30 milliseconds at a target value of 30 000 ions. The m/z ratios selected for tandem mass spectrometry (MS/MS) were dynamically excluded for 30 seconds.

Data were processed using MaxQuant software (version 1.0.13.13) (43). MS/MS spectra were searched with the MASCOT search engine against the decoy IPI-human database (forward and reverse sequences). Search parameters were as follows: variable modifications of N-terminal acetylation and methionine oxidation, fixed modification of cysteine carbamidomethylation, peptides of minimum six or more amino acids, maximum of two missed cleavages, minimum two razor peptides for quantitation, and peptide and protein false discovery rate of 0.01.

Gene ontology analysis using GOMiner

Gene ontology (GO) terms for identified proteins were extracted, and overrepresented functional categories for differentially abundant proteins were determined by the GOMiner tool (44). All proteins that were identified in the coimmunoprecipitates served as the background list, and level 4 GO terms categories were used for statistical calculations.

Results

Analysis of SK1 isoform expression and activity in MCF-7 SK1 cell lines

Plasmids containing C-terminal FLAG-tagged 43-kDa or 51-kDa SK1 sequences (Figure 1A) were stably transfected into MCF-7 cells. The expression levels of the FLAG-fusion proteins were assessed by SDS-PAGE and Western blotting of cell lysates (Figure 1B, lanes 1 and 2). Distinct protein products were observed at the predicted masses of 51 kDa and 43 kDa. To confirm the utility of FLAG-based affinity purification as an enrichment strategy, anti-FLAG immunoprecipitations were performed on these lysates. SDS-PAGE analysis and immunoblotting indicated strong enrichment of the SK1 isoforms (Figure 1B, lanes 3, and 4). Expression of the two isoforms was visualized by immunofluorescent staining using a FLAG-TAG antibody (Figure 2). Interestingly, the 51-kDa isoform was located in the cytoplasm, and the 43-kDa isoform was located throughout the cell as shown by the confocal microscopic studies (Figure 2C). Sphingosine kinase fluorescent activity assays were performed using the 43- and 51-kDa transfected cell lines, showing significant increased basal activity compared with parent cells and cells transfected with the empty vector (Figure 3).

Figure 1. FLAG-tagged 43-kDa and 51-kDa isoforms of SK1 produce two distinct products upon transfection into MCF-7 cells.

A, The 43-kDa and 51-kDa transcriptional isoforms of the SK1 protein, as identified by cDNA sequencing. The two isoforms contain a common C-terminal 384-amino acid sequence, but the 51-kDa isoform possesses an extra 86 amino acids at its N terminus. A C-terminal FLAG tag (DYKDDDDK) was appended to both sequences used in this study. B, MCF-7 lysates from cells transfected with FLAG-SK43 and FLAG-SK51 (lanes 1 and 2) were resolved by SDS-PAGE and analyzed by blotting with an anti-FLAG antibody to confirm the expression levels of the fusion proteins and the relative sizes of the two SK1 isoforms. The FLAG-SK1 proteins were immunoprecipitated with anti-FLAG antibody (lanes 3 and 4) from these lysates, demonstrating a strong enrichment of the fusion proteins.

Figure 2. Representative images of SK1 expression in MCF-7 cells stably transfected with the 51-kDa and 43-kDa isoforms.

Cells were transfected either the 51-kDa or 43-kDa isoform. The two isoforms were visualized by indirect immunofluorescence staining using FLAG-TAG antibodies and Alexafluor 488 (Invitrogen) and the cytoskeleton visualized using phalloidin conjugated to Alexaflour 568 (Sigma). The nucleus was visualized using Hoechst 33342. A, Overlay. B, Blue, nucleus. C, Green, SK1-FLAG. D, Red, cytoskeleton. Fluorescence images were acquired using a Nikon inverted fluorescence microscope (×60 objective). Scale bar, 20 μm.

Figure 3. Overexpression of SK43 and SK51 increases S1P basal activation in MCF-7 cells.

SK1 activity was measured in MCF-7 cells stably transfected with SK43, SK51, or vector and parent cells using the SK-fluorescent enzyme assay. HUVECs with and without TNFα were used as positive and negative controls. All assays were performed in biological triplicate experiments (in technical duplicate). *, P > .005. Error bars represent SEM.

Double-labeling SILAC for quantitative analysis of SK1 coimmunoprecipitations

Transfected MCF-7 cells were metabolically labeled using SILAC. Cells expressing FLAG-SK51 or FLAG-SK43 were grown in media containing lysine-8 and arginine-10 (a heavy label set), whereas cells transfected with an EV were grown in media containing amino acids with natural isotope abundances (a light label set). MS/MS analyses indicated essentially complete incorporation of the metabolic labeling (data not shown).

As outlined in Figure 4, A and B, lysates from heavy-labeled FLAG-SK1-expressing cells (either SK51 or SK43) were mixed 1:1 with lysates from light-labeled EV controls. The combined lysates (ie, SK51:EV or SK43:EV) were then used for SK1 coimmunoprecipitations by the use of anti-FLAG antibody-conjugated beads. Coimmunoprecipitation products were separated by SDS-PAGE, stained with blue silver, cut into slices, and digested with trypsin prior to LC-MS/MS analysis (see Materials and Methods for details). This process resulted in the identification and quantitation of 620 and 611 proteins, respectively, for SK51:EV and SK43:EV analyses. From one set of metabolically labeled cells, the coimmunoprecipitation and LC-MS/MS steps were performed twice, producing two procedural technical replicates.

Figure 4. SILAC co-IP experimental designs.

A and B, Dual-labeling SILAC strategy for identifying the interaction partners of SK51 and SK43, respectively. C, Triple-labeling SILAC strategy for the simultaneous analysis of SK51 and SK43 interaction partners. D, Post-IP mix triple-labeling SILAC validation strategy.

The ratios of proteins identified in the coimmunoprecipitates followed a near-normal distribution (Figure 5, A and C) with a maxima of approximately 1.0 and a SD of approximately 1.2. As expected, nonspecific immunoprecipitation contaminants were responsible for the peak at 1.0. These were proteins immunoprecipitated in equal quantities from lysates from SK1-expressing as well as control EV cells and tended to be proteins of high abundance in the cell, such as metabolic enzymes. Environmental and experimental contaminants such as keratins and trypsin were within the region of ratios 0–0.2. The SK1 protein itself was observed at an average ratio of 29.1 in SK51:EV analyses, and a similar average ratio was observed in SK43:EV analyses (22.9). Pearson's correlation coefficients for the procedural technical replicates (Figure 5, B and D) were 0.993 and 0.964 for SK51 and SK43 analyses, respectively, indicating a high reproducibility of coimmunoprecipitation and MS analysis steps. However, these linear correlation coefficients are strongly influenced by the SK data point.

Figure 5. Dual-labeling SILAC analyses of SK51:EV and SK43:EV anti-FLAG coimmunoprecipitation products.

A and C, SK51 (heavy) vs EV (light) and SK43 (heavy) vs EV (light) protein ratio histograms. Normalized ratios from anti-FLAG co-IP products were binned into units of 0.05, and the total number of proteins in each bin was plotted against the ratio. Co-IPs and MS analysis were performed twice from one set of lysates. SK1 was observed at an average ratio of 29.1 in analyses using SK51 and at an average ratio of 22.9 with the SK43 isoform. B and D, SK51 vs EV and SK43 vs EV technical replicate scatterplots. Protein ratios from the two procedural technical replicates of the anti-FLAG co-IPs and MS analysis were plotted against each other. Pearson's correlation coefficients were 0.993 and 0.964 for SK51 and SK43 studies, respectively, indicating low technical variation in these analyses. SK51 and SK43 points are off scale for all graphs in this figure.

Proteins identified with ratios higher than the main peak of the distribution at 1.0 were of particular interest because they were enriched in these experiments and represented the potential interaction partners of SK1. A number of proteins were identified in this region, with approximately 10–15 proteins in each SK51 and SK43 isoform analysis possessing ratios greater than 2 SD away from the population mean. This correlated with observed ratios of higher than approximately 1.4. A list of the top 20 proteins with the highest ratios is provided in Supplemental Table 1, A and B.

Triple-labeling SILAC for comparative analysis of SK1 isoforms

The double-labeling strategy described above established a set of putative interaction partners for SK51 and SK43. However, this approach did not provide a means of directly comparing these interaction partners and thus identifying proteins uniquely associated with either isoform. To address this, a triple-labeling strategy was thus used, using the introduction of a third label set: lysine-4 and arginine-6 (a medium label). Outlined in Figure 4C, the three cell lines (EV, FLAG-SK51, and FLAG-SK43) were each metabolically labeled (SK51 heavy, SK43 medium, and EV light), and anti-FLAG coimmunoprecipitation was performed on the 1:1:1 combined lysates prior to the MS analysis. Resultant protein ratios showed the expected normal distributions (Figure 6, A, C, and E), with SK1 observed at an average ratio of 40.5 in SK51 vs EV analyses and 99.9 in SK43 vs EV analyses. A number of proteins were observed at ratios greater than the main peak of the distribution at 1.0, representing putative SK1 interaction partners. A list of the top 20 proteins with the highest ratios is supplied in Supplemental Table 2, A and B.

Figure 6. Triple-labeling SILAC analyses of SK51:SK43:EV anti-FLAG co-IP products.

A, C, and E, SK51 (heavy) vs EV (light), SK43 (medium) vs EV (light), and SK51 (heavy) vs SK43 (medium) protein ratio histograms. Normalized ratios from anti-FLAG co-IP products were binned into units of 0.05, and the total number of proteins in each bin was plotted against the ratio. Experiments were performed twice with biological replicates. SK1 was observed at average ratios of 40.5 and 99.9 in SK51 vs EV and SK43 vs EV analyses, respectively. B, D, and E, SK51 vs EV, SK43 vs EV, and SK51 vs SK43 biological replicate scatterplots. Protein ratios from two biological replicates of the anti-FLAG co-IPs and MS analysis were plotted against each other. Pearson's correlation coefficients were 0.925 and 0.998 for SK51 vs EV and SK43 vs EV comparisons, respectively, with a score of 0.50 for SK51 vs SK43 comparisons. SK51 and SK43 points are off scale for all graphs in this figure.

Biological replicates were performed with the triple-labeling strategy. The amount of biological variation in these analyses (Figure 6, B and D; Pearson's correlation coefficients of 0.925 and 0.998 for SK51 vs EV and SK43 vs EV comparisons, respectively) was small, indicating a high reproducibility of labeling, coimmunoprecipitation, and MS analysis. The direct comparison of the coimmunoprecipitation products from SK51 and SK43 isoforms (Figure 6, E and F) was of particular interest. A number of proteins identified as putative SK1 interaction partners were observed at ratios outside the normal distribution of nonspecific binding proteins. This suggested they were specifically or preferentially associating with a single SK1 isoform and could be important in functionally differentiating the two isoforms of SK1.

Generation of a high-confidence list of SK1 interaction candidates by meta-analysis

To produce a high-confidence list of candidate SK1 interaction partners, data from double-labeling and triple-labeling experiments were combined. Ratios from technical replicates of double labeling were averaged to give measures for one biological replicate; these were used in combination with the two further biological replicates performed in triple-labeling studies. Proteins with the highest average SK51 vs EV or SK43 vs EV ratios were then sorted into candidate lists as SK51 and SK43 interaction partners and then filtered. To be accepted as an interaction partner, a protein had to be observed at least twice, with a relative SD (RSD) less than 50%.

As expected, both the SK51 and SK43 interaction candidate lists contained SK1 as the protein with the highest average observed ratio (36.7 for SK51 (Table 1), 74.2 for SK43 (Table 2)). Among the SK43 interaction candidates with high SILAC ratios was a known SK1 interaction partner: the protein phosphatase 2a (PP2A) protein. The PP2A protein was observed at a ratio of 1.41 in SK43 vs EV meta-analyses (Table 2), and this ratio was then used as a cutoff to produce a final list of the top 30 enriched SK43-associated proteins for further investigation. This same cutoff of the top 30 proteins with the highest ratios was then applied to the SK51 interaction candidates list. The final top 30 proteins with the highest ratios (both SK51 vs EV and SK43 vs EV) are shown in Tables 1 and 2.

Table 1.

SK51-Associated Proteins With the Highest SILAC Ratios

IDa	Ratiob	RSDc	Percentage to 43d	Post-IP Mix Ratioe	Name
Q9NYA1	36.7	64%		26.3	Sphingosine kinase 1
P11142	2.60	43%	+169%	5.73	Heat shock cognate 71 kDa protein
O95425	2.52	19%	−14%	1.93	Supervillin
O43795	2.27	12%	+81%	2.84	Myosin-Ib
Q16643	1.96	6%	−5%	4.22	Drebrin
Q9BQI0	1.90	4%	+135%	4.88	Allograft inflammatory factor 1-like protein
Q6IAU5	1.80	6%	+24%	13.1	PPM1G protein
O95816	1.79	1%	+19%	4.88	BAG family molecular chaperone regulator 2
Q7Z4W1	1.75	15%	+8%	25.8	L-xylulose reductase
B3KPC7	1.71	30%	+67%	2.95	Actin-related protein 2/3 complex subunit 5
Q9UHL4	1.67	18%	+309%	NA	Dipeptidyl peptidase 2
Q6IAT9	1.61	26%	−27%	4.78	Proteasome subunit-β type
Q14766	1.60	24%	+259%	1.40	Latent-transforming growth factor-β-binding protein 1
Q7Z6Z7	1.60	17%	−14%	4.16	E3 ubiquitin-protein ligase HUWE1
B0YJ74	1.59	8%	+53%	4.30	Proteasome subunit-α type
O15145	1.59	26%	+30%	2.64	Actin-related protein 2/3 complex subunit 3
O60762	1.58	8%	+64%	2.93	Dolichol-phosphate mannosyltransferase
P49006	1.58	14%	−11%	NA	MARCKS-related protein
O75116	1.56	11%	+27%	NA	ρ-Associated protein kinase 2
O00268	1.56	23%	+12%	2.17	Transcription initiation factor TFIID subunit 4
O14578	1.53	26%	+98%	2.40	Citron ρ-interacting kinase OS
O95793	1.53	11%	+6%	1.44	Double-stranded RNA-binding protein Staufen homolog 1
P60709	1.51	14%	+7%	2.29	Actin, cytoplasmic 1
P25788	1.51	16%	+52%	4.30	Proteasome subunit-α type-3
P49721	1.49	24%	+60%	4.63	Proteasome subunit-β type-2
Q13257	1.48	17%	+39%	2.08	Mitotic spindle assembly checkpoint protein MAD2A
O94832	1.48	17%	−27%	1.35	Myosin-Id
P62993	1.47	7%	+8%	2.56	Growth factor receptor-bound protein 2
P63261	1.46	15%	+2%	2.17	Actin, cytoplasmic 2
P04181	1.45	0.10	+28%	0.66	Ornithine aminotransferase, mitochondrial
Q13501	1.45	0.14	+23%	1.49	Sequestosome-1

Abbreviation: ID, identification. Listed are the top 30 proteins with the highest SK51 vs EV ratios from all experiments.

^aUniprot accession ID.

^bAverage ratio.

^cRelative standard deviation.

^dPercentage difference in association with SK43 to SK51, and the

^ePost-IP mix SILAC co-IP ratio for each protein are included. Identification data for all proteins are provided in Supplemental Table 3A.

Table 2.

SK43-Associated Proteins With the Highest SILAC Ratios

IDa	Ratiob	RSD	Percentage to 51c	Post-IP Mix Ratiod	Name
Q9NYA1	74.2	63%		101	Sphingosine kinase 1
O95425	3.14	38%	+17%	4.45	Supervillin
Q9UH62	2.55	16%	+61%	2.73	Armadillo repeat-containing X-linked protein 3
Q16643	2.17	36%	+5%	4.20	Drebrin
Q7Z4W1	1.96	29%	−7%	7.14	L-xylulose reductase
O94832	1.95	18%	+37%	2.71	Myosin-Id
Q13085	1.90	42%	+38%	NA	Acetyl-CoA carboxylase 1
Q7Z6Z7	1.88	22%	+17%	4.41	E3 ubiquitin-protein ligase HUWE1
Q6P1M3	1.65	6%	+11%	1.67	Lethal (2) giant larvae protein homolog 2
P21291	1.65	10%	+47%	1.93	Cysteine and glycine-rich protein 1
Q5T2W1	1.64	19%	+194%	1.53	Na (+)/H (+) exchange regulatory cofactor NHE-RF3
P49006	1.63	33%	+13%	NA	MARCKS-related protein
Q9NYL9	1.57	31%	+15%	2.27	Tropomodulin-3
P60981	1.53	16%	+18%	1.53	Destrin
P63261	1.53	17%	−2%	1.08	Actin, cytoplasmic 2
Q6IPJ9	1.52	21%	+7%	1.98	Ladinin 1
P35579	1.50	2%	+9%	1.90	Myosin-9
O95793	1.49	1%	−6%	1.50	Double-stranded RNA-binding protein Staufen homolog 1
P98179	1.48	2%	+18%	NA	Putative RNA-binding protein 3
P60709	1.48	19%	−6%	1.24	Actin, cytoplasmic 1
O95816	1.47	9%	−16%	2.55	BAG family molecular chaperone regulator 2
O00159	1.44	10%	+30%	3.60	Myosin-Ic
Q6IAU5	1.44	10%	−19%	4.88	PPM1G protein
B4E1G6	1.43	2%	+45%	1.71	Galactokinase 1
Q5SSJ5	1.43	20%	+24%	1.36	Heterochromatin protein 1-binding protein 3
Q6IB91	1.42	9%	+122%	NA	PCK2 protein
P22626	1.42	28%	+30%	1.57	Heterogeneous nuclear ribonucleoproteins A2/B1
P49411	1.42	27%	+23%	1.38	Elongation factor Tu, mitochondrial
Q15286	1.41	24%	+9%	NA	Ras-related protein Rab-35
Q99873	1.41	17%	+50%	1.17	Protein arginine N-methyltransferase 1
P63151	1.41	1%	+88%	1.22	Serine/threonine-protein phosphatase 2A 55-kDa regulatory subunit Bα isoform

Abbreviations: ID, identification; NA, not available; NHE-RF3, Na(+)/H(+) exchange regulatory cofactor. Listed are the top 30 proteins with the highest SK43 vs EV ratios from all experiments.

^aUniprot accession ID.

^bAverage ratio.

^cPercentage difference in association with SK51 to SK43.

^dPost-IP mix SILAC co-IP ratio for each protein are included. Identification data for all proteins are provided in Supplemental Table 3B.

The PP2A enzyme is one of the most well-established SK1 binding partners and can dephosphorylate SK1 through its phosphatase activity via a physical interaction (45, 46). Other reported SK1 interaction partners apart from the PP2A protein were not identified in our experimental system. This could be due to these proteins not being present in sufficient quantities for detection by MS or simply because these reported interactions are cell line specific.

Validation of SILAC coimmunoprecipitations

To validate the analyses described above, a subtle but significant variation to the SILAC coimmunoprecipitation experiment was introduced. Anti-FLAG coimmunoprecipitations were performed on each labeled lysate in isolation before combining the three final coimmunoprecipitation products 1:1:1 for MS analysis, constituting a post-immunoprecipitation (IP) mix experiment (Figure 4D). This differs from the previous pre-IP mix experiments in this study in which the labeled lysates were combined for a single coimmunoprecipitation reaction. Such an approach has been previously used in SILAC co-IP-based studies to identify protein interaction partners (47). This yielded a broader distribution of protein ratios (Figure 7, A–C), with a considerably more pronounced right shoulder. This was particularly significant in the SK51 isoform analysis (Figure 7A). However, because the post-IP mix strategy involves performing separate coimmunoprecipitation reactions with each labeled lysate (rather than a single combined coimmunoprecipitation step), it does not control for variations and irreproducibility inherent in coimmunoprecipitations. The resultant ratios from the post-IP mix experiment were added to the above-mentioned high-confidence SK1 interaction candidates list yielded by a meta-analysis (Tables 1 and 2) to show that an orthogonal coimmunoprecipitation technique could be used to validate the SK1 interaction candidates deduced from previous experiments.

Figure 7. Triple-labeling SILAC analysis of post-IP mix SK51:SK43:EV anti-FLAG co-IP products.

Coimmunoprecipitations were performed on each labeled lysate in isolation, and the separate products were combined 1:1:1 for MS analysis. A, B, and C, SK51 (heavy) vs EV (light), SK43 (medium) vs EV (light), and SK51 (heavy) vs SK43 (medium) protein ratio histograms. Normalized ratios from anti-FLAG co-IP products were binned into units of 0.05, and the total number of proteins in each bin was plotted against the ratio. Sphingosine kinase was observed at ratios of 27 and 101 in SK51 vs EV and SK43 vs EV analyses, respectively; these points are not shown in these graphs.

The ratios observed in the post-IP mix experiment generally supported the rankings of SK1 interaction candidates. Only one of the proteins established as a putative interaction partner of SK51 and SK43 in previous experiments was observed at a ratio below 1.0 in this validation (ornithine aminotransferase, with a ratio of 0.66), whereas a number of proteins showed strongly enhanced ratios. The l-xylulose reductase protein presents a striking example: it averaged moderate to highly enriched ratios of approximately 1.8 in all prior SK43 and SK51 analyses but was observed at the extremely highly enriched ratios of approximately 25 with the SK51 isoform and approximately 7 with the SK43 isoform in a post-IP mix analysis. Many other proteins identified as putative interaction partners of SK1 also exhibited an increased presence, including protein phosphatase, Mg²⁺/Mn²⁺ dependent, 1G (PPM1G), allograft inflammatory factor 1-like protein, E3 ubiquitin protein ligase HUWE1, and various identified proteasome subunits. Some proteins opposed this trend, however, by showing ratios closer to 1.0 in the post-IP mix SILAC coimmunoprecipitation. This was more apparent in the SK43 interaction candidates, in which proteins such as actin (cytoplasmic 1 and 2) dropped from ratios of 1.48 and 1.53 to 1.24 and 1.08, respectively. Associations between SK1 and actin, as well as other cytoskeletal features, have been previously reported (48).

DPP2/4 inhibitor Ala-BoroPro increases SK51kDa expression and has no effect on SK43kDa expression

The DPP2 belongs to a family of serine proteases that exhibits posttranslational proline-dependent cleavage specificity at the N terminal of protein sequences (49). Ala-BoroPro is a potent inhibitor of DPP2 and DPP4 (50). From the SILAC data analysis (Table 1), the DPP2 protein was observed to be greatly (309%) more abundant in association with the SK51 isoform than SK43. Disruption of the DPP-SK51kDa interaction using Ala-BoroPro increased SK51 protein expression within 1–4 hours of treatment similar in intensity to that observed for MG132, a generic protease inhibitor, whereas CHX treatment (4 h) inhibited protein expression of the 51-kDa isoform (Figure 8A). The 43-kDa isoform was shown to be a more stable protein, whereas 4 hours of treatment with Ala-BoroPro, CHX, and MG132 showed little or no change in the expression of the shorter isoform.

Figure 8. DPP inhibition increases SK51 expression and has no effect on SK43 expression.

A, Representative Western blots of SK1 isoforms, p21, and β-actin from MCF-7 cells stably expressing SK43 and SK51 treated with control (C; PBS), AL-BoroPro [DP₁ (10 nM) and DP₂ (100 nM)], CHX (10 μM), MG132 (M; 10 μM). Experiments were repeated in three independent biological experiments with similar results. B, The Ala-BoroPro inhibitor had no adverse effect on cell proliferation in MCF-7 parent and SK51- and SK43-transfected cell lines. Graph shows trypan blue exclusion counts of biological replicate experiments (±SD).

DPP2/4 inhibitors had no overt effect on morphology and cell proliferation/apoptosis in MCF-7 cells

Because Ala-BoroPro inhibitor increased the expression of SK51kDa expression and SK1 is a potential oncogene, the consequences of Ala-BoroPro treatment on cell cycle was evaluated. Downstream consequences of inhibiting DPP with Ala-BoroPro was evaluated by observing the effect on p21, a cell cycle regulator in MCF-7 cells (38), and the trypan blue exclusion cell proliferation assay (51). The Ala-BoroPro inhibitor had little or no effect on p21 protein expression in cell lines stably expressing the SK43 and SK51, whereas Ala-BoroPro increased SK51 expression under the same experimental conditions as shown by Western blot (Figure 8A). Treatment with CHX decreased p21 expression and MG132 increased p21 expression (Figure 8A). Using trypan blue exclusion, no significant difference was observed in cell proliferation (Figure 8B) and no change in cell morphology (data not shown). Furthermore, no change in the localization of SK51 and SK43 was observed upon treatment with Ala-BoroPro: SK51 remained largely cytoplasmic, whereas the SK43 isoform was nuclear and cytoplasmic (data not shown).

GO classification of high-confidence SK1 interaction candidates

The GO system was used to determine whether the final top 30 lists of the newly identified SK51 and SK43 interaction candidates showed enrichment for specific functions. Using the GOMiner tool (52), the GO terms that were enriched among the top 30 SK51- and SK43-binding proteins relative to all identified proteins were determined (Figure 9, A and B). Enriched functional groups common between the two isoforms included proteins involved in assembling or organizing cell-cell junctions, a process that SK1 has been reported to influence (53). Additionally, there was a particularly strong enrichment of proteins involved in cell motility/locomotion in both SK51 and SK43 candidates, supporting previous reports characterizing SK1's promigratory effects in breast cancer (16) and suggesting a possible avenue of investigation for this phenomenon. Previous reports have characterized the localization of SK1 (recruitment from the cytoplasm to association with the plasma membrane) as of significant importance in regulating its signaling activity (54), and the enrichment of proteins involved in the process of intracellular component movement seemed to reflect this. The precise molecular mechanisms for how the localization of SK1 is dynamically regulated remain unknown (55–58), and these data suggest components that could play a role in this process. Candidates include the Rho-associated protein kinase 2 because it coordinates cytoskeletal dynamics that have been observed to regulate SK1 localization and activity (59, 60).

Figure 9. GO enrichment analysis of SK1-binding proteins.

All proteins identified in coimmunoprecipitates, and the final top 30 lists of SK1-interacting proteins with the highest ratios, were analyzed for functional relationships using GOMiner (see Materials and Methods). Graphs for the level 4 terms are found to be most enriched in SK51 (A) and SK43 (B) top 30 lists and are shown relative to all identified proteins (∼800). Categories common to both SK51 and SK43 interaction candidates are listed first.

Enrichment of protein folding proteins (chaperones) and ubiquitination are likely artifactual features present due to the overexpression of the SK1 protein, although there are reports of chaperone proteins associating with SK1 and the degradation of SK1 occurring through the proteasomal pathway (32, 61).

Discussion

The aim of this work was to use quantitative proteomics to address two questions: what are the protein interaction partners of SK1 in breast cancer and the interaction partners of the two isoforms of SK1. Previous approaches for the identification of SK1-interaction partners, such as the yeast two-hybrid system, have been hindered by issues such as the spurious activation of reporter genes leading to a high rate of false-positive results (62). Equally, coimmunoprecipitation and MS-based approaches can be problematic due to the difficulties in distinguishing between genuine interactions with the bait protein and environmental or nonspecific contaminants (63). Relative to more conventional, qualitative techniques for identifying protein-protein interactions, SILAC has advantages in providing an unbiased assessment of binding specificity (64). Additionally, inherent limitations in more conventional techniques preclude direct and simultaneous analyses of two protein isoforms, as can be achieved with triple-labeling SILAC. Using a combination of SILAC-based double- and triple-labeling approaches, we have identified novel common and isoform-specific interaction partners of the two major isoforms of SK1. These interactions have implications in the cellular signaling pathways in breast cancer.

An issue addressed here with SILAC coimmunoprecipitations is the effect of the step at which work flows are combined. Two models are predominant: one in which the labeled lysates are combined for a single immunoprecipitation reaction (a pre-IP mix) and one in which separate immunoprecipitations are performed on each labeled lysate, with the separate immunoprecipitates later being combined for MS analysis (a post-IP mix). Studies have used either the former or the latter seemingly interchangeably (64–67). We sought to address this issue by using post-IP mix immunoprecipitations to validate our data produced from pre-IP mix experiments. The two approaches yielded comparable results: proteins that we had previously identified with high ratios retained this property under validation. There were numerous cases of proteins exhibiting higher ratios in the post-IP mix experiment, the most dramatic of which were the l-xylulose reductase and PPM1G proteins. It is difficult to attribute this to a single specific cause, but it is likely that combining lysates for a single immunoprecipitation reaction (ie, in the pre-IP mix approach) decreases the concentration of the ligand or bait protein and therefore has a diluent effect on protein-protein interaction equilibria. When each lysate is immunoprecipitated separately in a post-IP mix SILAC co-IP, this does not occur, resulting in higher ratios being observed. It would be prudent for further SILAC-based IP studies to bear this issue in mind.

The PP2A serine/threonine phosphatase that we found associated with SK43 is perhaps the most thoroughly characterized SK1 interaction partner (45), responsible for dephosphorylating (and therefore deactivating) SK1. The observed PP2A SILAC ratios served as a cutoff for the generation of our high-confidence SK1 interaction candidates, although this was probably quite conservative (35).

Because this is the first attempt to identify SK1 interaction partners in breast cancer cells and the first to use the 51-kDa isoform of SK1, we were unsurprised to identify few previously reported interactions. However, a close examination of the candidate SK1 interaction partners from this study (Tables 1 and 2) revealed that they shared striking characteristics with other proteins previously reported as SK1-interacting proteins from other cell lines or organism. For example, the δ-catenin/neural plakophilin-related armadillo repeat protein was identified in a yeast two-hybrid screen using a rat brain cDNA library as a SK1-interacting protein (27). Further immunoprecipitation experiments revealed that an armadillo-repeat sequence within this protein was responsible for its binding to SK1. We identified the armadillo repeat-containing X-linked protein 3 as being the second most highly enriched protein in SK43 co-IPs. This protein is an integral membrane protein, possessing many of the repeat armadillo sequences that were previously identified to interact specifically with SK1, and has been implicated in epithelial tumorigenesis (68). The allograft inflammatory factor 1-like protein, a protein we observed as exclusively associating with the SK51 isoform, was another candidate that shared characteristics with previously reported SK1 interaction partners. In this case, it was the combination of plasma membrane localization and calcium-binding activity that it shared with the calcium- and integrin-binding protein 1 (58).

One of the key observations of the functional role of SK1 in breast cancer has been its promotion of cell migration. Yet how this function is achieved is unknown (16). We identified the supervillin protein, a key regulator of cell motility (69, 70), as one of the most highly enriched proteins present in the lists of SK1 interaction candidates. The colocalization of supervillin with signaling proteins in cholesterol-rich lipid raft membranes (71) is particularly interesting because these regions are rich in the sphingosine substrate for SK1 activity. Localization of SK1 to these regions has been proposed as a component of the mechanism of SK1 signaling (54). Supporting this, we also identified the myristoylated alanine-rich C-kinase substrate (MARCKS)-related protein among the SK1 interaction candidate lists. The MARCKS-related protein has been reported to localize to lipid rafts upon stimulation by the phospholipid product of SK1 and is implicated in coordinating cell adhesion and migration (72).

Supervillin has also been reported to interact with the androgen receptor protein (73), which suggests a possible role in the influence of SK1 in hormone signaling in both breast and prostate cancer (17, 74, 75). Most notable of all is the identification of the growth factor receptor-bound protein 2 (GRB2) in our SK1 interaction candidates, a signaling protein regulating the response to epidermal growth factor (76). Supervillin is necessary for sustained ERK phosphorylation in response to epidermal growth factor activation (71), and the action of GRB2 could potentially have some role in maintaining the activation and translocation of SK1 in response to ERK phosphorylation (77). It is therefore interesting to note that the stimulation of SK1 activity through its phosphorylation leads to an increased migration in breast cancer cells (78).

The sustained phosphorylation of SK1 has been found to be a strong component of its oncogenic signaling (57, 77). The identification of the serine/threonine phosphatase PPM1G in our SK1 interaction candidates list was therefore of interest because this suggests an alternative dephosphorylation mechanism of SK1, independent of the PP2A phosphatase. A member of the large PP2C family of protein phosphatases, the oncogenic PPM1G protein is a negative regulator of the p53 tumor suppressor response pathway (79). Whether it has any phosphatase activity on SK1 will be interesting to determine, particularly in the context of recent studies linking ERK-mediated phosphorylation of SK1 with breast cancer progression through increased cell growth and migration (80).

As far as we know, this is the first study using multiplexing-based SILAC coimmunoprecipitations to directly compare the interactional capacities of two isoform species of the same protein. As predicted, GO analysis indicated the functional roles of the interaction candidates for both SK1 isoforms were very similar (Figure 8, A and B), illustrating that the interactions of these two isoforms had much in common. Because the SK43 and SK51 protein isoforms differ by only an 8-kDa segment in the N terminus, this was to be expected. However, we did observe a number of isoform-specific interactions, which were detailed by deriving a percentage figure of the relative abundance of each SK1 interaction candidate in association with each SK1 isoform, based on the observed SK51 vs SK43 ratios. For example, our data indicated the PP2A phosphatase associated considerably more strongly (88%) with the conventional SK43 isoform than with the SK51 isoform. This interaction has been established with only the 43-kDa SK1 isoform (45, 46), and these data produced the first example of an isoform-specific interaction. The diminished association of PP2A with the SK51 isoform raises questions as to whether the two SK1 isoforms are independently regulated. Considering that constitutive SK1 activity has been suggested to be oncogenic (81), this could be an observation of importance.

There were a few other cases in which a protein associated more strongly with the 43-kDa isoform of SK1. The armadillo repeat-containing X-linked protein 3 protein (detailed above) was one such protein; it was 61% more abundant in association with SK43 than SK51. This property was also observed with the Na(+)/H(+) exchange regulatory cofactor protein, a primary decidual zone-domain containing protein that localizes at cell junctions in epithelial cells (82), which associated significantly (194%) more strongly with SK43 than SK51.

In general however, isoform-specific interactions were more commonly observed with the longer 51-kDa isoform of SK1. Examples include the citron rho-interacting kinase (80%), latent-transforming growth factor β-binding protein 1 (259%), and various proteasomal subunits (50%–60%). The allograft inflammatory factor 1-like protein, a calcium-binding protein, was one of the strongest SK51-exclusive interaction candidates. It exhibited both a high ratio relative to the EV control (1.9) and much higher abundance in association with SK51 (135%) relative to 43 kDa SK isoform. It is difficult to draw functional segregations between the two SK1 isoforms on this basis, however, because the conventional 43-kDa SK1 isoform exerts a regulatory role in calcium mobilization and signaling (83–86).

The unique association of the SK51 isoform with DPP2 detection by SILAC analysis was of particular interest, considering previous reports describing the cleavage of SK1 by cathepsin B (87, 88), suggesting interactions with other proteases were not unexpected. In this study the increase and stabilization of SK51 through inhibiting DPP2 would increase the propensity for increased S1P activity, and elevated SK1 has been shown to be associated with an increased risk of breast cancer (7, -Ohotski J, Edwards J, Elsberger B, Watson C, Orange C, Mallon E, Pyne S, Pyne NJ. Identification of novel functional and spatial associations between sphingosine kinase 1, sphingosine 1-phosphate receptors and other signaling proteins that affect prognostic outcome in estrogen receptor-positive breast cancer. Int. J. Cancer. 2013: 132, 605–616.) and resistance to chemo- and hormone therapies (22, -Sukocheva O, Wadham C, Xia, P. Role of sphingolipids in the cytoplasmic signaling of estrogens. Steroids. 2009; 74: 562–7). Increased SIP is also strongly associated with processes such as inflammation leading to neovascularization and cancer progression reviewed elsewhere (91). There is some debate whether DPP inhibitors, currently used in diabetic treatment, increase cancer risk (92). In this study administration of the dipeptide Ala-BoroPro, a potent inhibitor of DPP2 and DPP4, had no adverse effect on cell proliferation consistent with its potential to increase SK1 activity. The findings in this current study whereby DPP inhibitors stabilize and increase SK51kDa in breast cancer cells may have implications for cancer progression in the increasing number of patients with breast cancer and diabetes adding to the conundrum of comorbidity treatments. More functional studies are needed, especially to determine combinational therapy treatments with the emerging interest of SK1 inhibitors in cancer treatments. There are limited studies on the functional differences of the enzyme kinetic properties of the two SK1 isoforms. One such study reports SK1 isoform specific inhibitor sensitivity in androgen-sensitive prostate cancer cells whereby SK1 51 kDa (SK1b) confers resistance to the Ski inhibitor, whereas SK1 43 kDa (SK1a) undergoes apoptosis (33).

In part, the differences in subcellular distribution of the SK1 51-kDa and 43-kDa proteins observed may facilitate differential roles in endocrine resistance and recurrence. This suggestion is supported by the recent report showing nuclear localization and protein interactions of SK1 may affect prognostic outcome, significantly reducing disease-free survival and recurrence in estrogen receptor-positive breast cancer (22). Although nuclear export signals are located in the N terminal common to both SK1 isoforms (94), speculatively, the 86-amino acid of the 51-kDa may prevent the longer isoform from translocation back into the nucleus, whereas the 43-kDa isoform lacks the 86-amino acid N-terminal region making it more receptive to translocation between the two compartments. Increased SK51 expression in cells treated with Ala-BoroPro had no effect on SK51 localization.

Many of our identified SK1 interaction candidates play roles in cell signaling, adhesion, and migration, supporting and highlighting the oncogenic role of aberrant SK1 expression in breast and potentially many other cancers. This report is the first to identify SK51 isoform-specific interacting partners and identify regulation of expression of the SK51 isoform independent of 43 kDa. An increase in 51-kDa SK protein expression with the inhibition of DPP2, not observed with the 43-kDa protein is a unique finding and supports the SILAC approach to the identification of differential as well as common interacting partners for the two SK1 proteins. These findings add to the evidence that the different isoforms perform different as well as similar functions in the cell. This would strengthen the concept that the different SK1 signaling pathways represent potential therapeutic targets for treatment of resistant and migratory breast and other tumors.

Funding Statement

This work was supported by a Cancer Institute New South Wales Career Fellowship (to E.M.M. and P.X.) and the Sydney Cancer Centre, Royal Prince Alfred Hospital, Sydney (to E.M.M. and P.X.). M.R.W. acknowledges support from the EIF Super Science Scheme, the New South Wales State Government Science Leveraging Fund, the Australian Research Council, and the University of New South Wales.