Abstract: Presumably any tumor cell population contains concomitantly, therapy-sensitive and therapy-resistant cells in contrast to the opinion, still held by many, that drug resistance develops only after repeated cycles of treatment. To test this hypothesis we selected a breast cancer cell line that under exposure to AGL2592 (Stat3 inhibitor) will result, along a wide concentration range, in steady levels of residual cell counts. MCF7 and T47D breast cancer cell lines showed a low ic50 (0.4 and 0.5 μM, respectively) from which we chose MCF7 because it showed only 10% residual cells compared to 40% in T47D. Proteomes of resistant and untreated MCF7 cells were run on 2D-gels for comparative analysis of protein expression. Calreticulin (CRT) and protein disulfide isomerase (PDI) showed lower abundance in treated cultures. Glycolytic enzymes, enolase1, aldolase A and glyceraldehydes 3 phosphate dehydrogenase (GAPDH) were more abundant and less modified in treated cultures. AGL2529-sensitivity was associated with part of the GAPDH molecules susceptible to cleavage at Ile170 and adenylation between Ile170 and Lys186. The differential profile exhibited by all the above proteins is (according to others) the hallmark of unfavorable prognosis, which suggests that therapy resistant and sensitive MCF7 populations may coexist in this cell line in its pre-selected form.
Introduction
Therapy resistance of relapsing cancer is one of the many unresolved issues in tumor cell biology. A frequent scenario in Breast Cancer (BC) hormonal and chemotherapy, is that an apparent BC eradication is followed by tumor-relapse that shows increased resistance to therapy (Tonetti and Jordan, 1995). In vivo BC cell populations initially respond to chemotherapy by increased apoptosis and cell proliferation, which decline within one month from the start of chemotherapy (Dowsett et al., 1999) and are later followed by BC relapse. The cells responsible for BC relapse are chemo-resistant residual cells. Similarly, chemical treatment of BC cells in culture frequently results in residual chemo-resistant tumor cell populations remaining in culture after the death of the chemo-sensitive cells (Palmari et al., 2000). Much of the effort in studying chemo-resistance mechanisms has been focused on multidrug resistance molecules (MDR). However, MDR glycoproteins are not always active in drug resistant tumors (Salvatore et al., 1999) and are not expected to play a role in irradiation-resistant tumor cells. This indicates that search for additional therapy-resistance mechanisms, is warranted. Cancer cells differ, from their normal counterparts, by reliance on the glycolytic metabolism similar to the strategy that normal cells adopt temporarily under anaerobic conditions. Increased consumption of glucose by tumors, while relinquishing the highly economic benefits of the Krebs cycle and oxygen-driven oxidative phosphorylation, has been noticed 80 years ago by Otto Warburg. The Warburg effect refers to the restricted use of glycolysis independently of oxygen presence and its fulfillment is associated with a change in the control of protein expression (Unwin et al., 2003). Normally hypoxia inducible factor-1 (HIF-1) regulates angiogenic factors and glycolitic enzymes expression, in response to low oxygen levels. Examples where HIF-1 expression may be regulated in disregard of oxygen, are its induction when c-Src is over-expressed (Karni et al., 2002) and by increased pyruvate synthesis (Lu et al., 2002) that inhibits HIF-1 degradation. Thus, the Warburg effect distinguishes tumor from normal cells, nevertheless, our question is which protein changes distinguish therapy resistant from therapy sensitive BC cells? In the later case therapy at least distinguishes between the 2 roughly defined phenotypes. Being influenced by the tumor stem-cell hypothesis, our experimental model assumes the concomitant existence of chemo-resistant and chemo-sensitive cell populations (even in cultured BC cell lines and) regardless of the nature of anti cancer therapy. This model is inline with the concept of carcinogenegisis in mammary stem-cells (Smalley and Ashworth, 2003). Hypothetically, stem-cells becoming tumors could proliferate continuously giving off progeny, half of which will lose "stem-ness" and become prone to apoptosis, i.e., becoming more sensitive to therapy, the rest would remain resistant stem-cells. Whether this turns out to be the case or not it seemed advisable to reveal differential proteomics in chemo-resistant cells. It is reasonable that chemo-resistant BC cells contain potential molecular targets for supplemental chemotherapeutic agent/s. Such putative targets would be difficult to discern by the use of comparative transcriptomics, because our current inability to predict post-translational modifications of proteins based on mRNA (cDNA) levels. This problem is also reflected by the difficulty to discern resistance from sensitivity by comparative transcriptomics in the different clinical stages in BC patients (Ma et al., 2003). The hypothesis that the BC cell population contains cells that apriori are resistant to chemotherapy, (which is well fitting into the BC stem-cell hypothesis (Smalley and Ashworth, 2003)) required pilot experiments to test its validity and to suggest a cell-culture model. We chose to use two tyrphostins AG 1478 and AGL2592 to inhibit the growth of 2 breast cancer cell lines, MCF7 and T47D, both express low levels of EGFR (ErbB2) (Daly et al., 1997) (target for AG1478). BC cells belong to a group of cancers that over-express Stat proteins (Real et al., 2002; Zhang et al., 2003), inhibition of Stat3 underscored the importance of its activity in BC (Li and Shaw, 2002). In normal breast epithelium 2 alternative activators of Stat3 were demonstrated, either by Jaks or by c-Src kinases, these were suggested to result in 2 different functional modules (Vultur et al., 2004). Stat3 inhibition via Jaks (but not via Src) antagonized the cell adherence-induced Stat3 phosphorylation. However in the BC cells Stat3 constitutive activity was blocked via both, Jak and Src. AGL2592 was shown to inhibit Stat3 phosphorylation in leukemic (Ben-Bassat et al., 2002) and in ovarian cancer cell lines (Arbel et al., 2003). In the present work we chose to inhibit Stat3 based on the above information, but also because it is an inducer of several anti-apoptotic (survival) proteins like Bcl-2 and ERK ½ which we could follow in the cells that survived treatment. We show that an MCF7 cell population resistant to AGL2592 exists (apriori) within untreated cultures. This resistant MCF7 population is differentially expressing several proteins in a manner that is consistent with the different pattern of their expression in malignant/metastatic relative to less malignant or to normal cells (Table 2).
Materials and Methods
Reagents
Antibodies against phosphorylated dual specificity kinase target (pT-E-pY)
and protease inhibitors were from Sigma, anti Bcl-2 antibodies were purchased
from Sant-Cruze. MCF7 and T47D breast cancer cells were from ATCC. Cells were
grown in Dulbeccos modified Eagles medium 10% fetal calf serum 50
mg mL-1 penicillin and streptomycin (Invitrogen). Tyrphostins AG1478
and AGL2592 were a gift from Alexander Levitzki.
Cell Count and Determination of IC50
Cells were plated at 3 or 5x103 cell/well in microtiter
plates and after 24 h medium containing tyrphostins was added to final concentrations
indicated in Fig. 1. The plates were fixed 48 h after exposure
to the tyrphostins by adding 0.5% glutaraldehyde. The cultures were stained
with 0.1% methylene blue in 0.1 M borate buffer pH 8.6 for 60 min. Surplus stain
was thoroughly washed with distilled water and the residual stain was eluted
with 0.1N HCL and counted by an ELISA counter at 620 nm.
Electrophoresis and Western Blot Development
Cells were seeded, 104 cm-2, in 25 cm2
flasks, after 24 h medium with different concentrations of AGL2592 was applied
to the cultures, 48 h later the cultures were harvested and subjected to cell
lysis. Flasks were placed on ice and washed twice with 6 mL of cold PBS removing
the washing solution to dryness. Ice-cold electrophoresis sample buffer (with
7% 2-mercaptoethanol and 3% sodium dodecylsulfate (SDS)) 100 μL/flask was
added to incubate for 30 min on ice. The lysates were quickly scraped off the
plastic with a policeman and transferred to cold tubes, for quick spinning,
2 min at 14,000 rpm, boiled for 5 min, cooled on ice and stored at -70°C
until further use. Samples of lysed cultures representing equal protein concentrations
and adjusted to equal volumes with plain sample buffer were subjected to polyacrylamide
gel electrophoresis. Gels were then electroblotted onto PVDF filters that were
blocked with 3% bovine serum albumin and 0.2% Tween 20. Filters were first incubated
with anti Bcl-2 antibody and after washing they were incubated with anti rabbit
IgG second antibody conjugated to peroxidase and the chemilluminescent was detected
by autoradiography. Immunoblot results were analyzed by scanning of chemilluminescent
autoradiographies and performing image analysis on the bands, using Scion-NIHimage
software.
Comparative Proteomics
Comparative 2 dimensional electrophoresis of AGL2592 treated and control
MCF7 breast cancer cells. MCF7 cells were seeded in 150 cm2 culture
flasks, treated with 10 μM AGL2592 or carrier and harvested on ice. The
cell lysates where precipitated in cold 95% ethanol 5% de-ionized water overnight
at -20°C The precipitate was washed in 90% ethanol 10% water and air-dried
at -20°C. The protein pellets were dissolved in 300 μL of isoelectric
focusing (IEF) sample buffer and 1.3 mg of each sample were adsorbed into 24
cm IPG strips (IEF at pH 3-10 range) by in-gel re-hydration at low voltage for
3 h. IEF was performed during 24 h by a total of 55000 volt-hours and equilibrated
15 min with the DTT and SDS-containing 2nd dimension sample buffer followed
by 15 min equilibration with iodoacetamide. The samples of AGL2592 and control
were ran concomitantly on 2 identical 10% polyacrylamide gels and stained with
Coomassie blue. Protein spots that showed clear difference between the two 2D
gels were removed and placed into acetonitrile until trypsin in-gel digestion
was performed. After peptide extraction and desalting mass spectrometry for
the specific proteins was done with collision induced dissociation of amino
acid residues, detecting their time of flight on a Q-TOF-2 device. Proteins
were identified with Mascot and Global-Server software.
Results
Selection of Chemo-therapy Compound
MCF7 and T47D breast cancer cells were exposed to decreasing concentrations
of AGL2592 and AG1478, 24 h after they have been seeded in microtiter plates.
The cultures were fixed and stained for counting 48 h after exposure to these
compounds. Figure 1a shows that MCF7 cells are more sensitive
to the EGFR kinase inhibitor (AG1478) with an IC50 of 17 μM
then the T47D cells (IC50 = 27 μM, Fig. 1b).
However, AGL2592 was a much more effective cytotoxic agent than AG1478 in both
cell lines (4 and 5 μM, Fig. 1a).
Fig. 1: | Dose response of MCF7 and T47D, breast cancer cell lines, to tyrphostins. Cells were seeded at 5x103/well and 24 h later the medium was brought to indicated concentrations of the tyrphostins, a = AG1478, b = AGL2592. Forty eight hours after exposure to these compounds, the cultures were fixed and stained with methylene blue for cell counting. Cell counts are expressed as indices relative to untreated wells (n = 10 per point) |
Fig. 2: | Time course of MCF7 cells response of Bcl-2 and dual-specificity kinase (ERK1/2) motif under AGL2592. MCF7 cells were seeded, 106cells/25 cm2 flask and 24 h later the medium was adjusted to 3 μM AGL2592 or the equivalent DMSO carrier concentration as controls. The cultures grew in the presence of the compound for indicated time intervals and then harvested with 100 μL RIPA on ice, 40 μg of protein equivalents were loaded on 10% SDS acrylamide gels. The Western blots were developed with indicated antibodies (a.) and the radiographic bands were scanned and their image (b = Bcl-2, c = Anti dual specificity kinase phospho-motif) was analyzed for quantitative density |
Although the IC50 obtained for the Stat3 activation inhibitor AGL2592 was similar in both cell lines (0.4 μM for MCF7 and 0.5 μM for T47D cells), the level of residual cell counts at 10 μM AGL2592 was fourfold higher in T47D cells (40%) than that of residual MCF7 cells (only10%). This indicates that T47D cells contain a fourfold larger chemo-resistant population than MCF7 cells. Therefore, if we were to compare the balk of untreated cells (that hypothetically) coexist with the resistant cells, then MCF7 residual cells are expected to differ from the untreated cells in a more pronounced manner and with a lower background than in the case of T47D cells. Hence, MCF7 cells seemed more suitable for a pilot study to explore proteomic markers of chemo-resistance.
Survival Protein in Resistant Cells
Anti-apoptotic proteins like Bcl-2 and ERK1/2 are Stat3-inducible and were
followed as a marker to the AGL2592 effect. If cell death under AGL2592 in 90%
of the MCF7 cells is accompanied, during the first exposure day, by a decrease
in a survival protein e.g. Bcl-2, then 48 h later the 10% surviving cells should
exhibit a Bcl-2 level that would be higher than the level in the dying cell
population, (that can be observed only during the first 24 h). Indeed, Fig.
2 shows that at 48 h Bcl- 2 levels become much closer to the non-treated
controls than seen after 10 min, 30 min and 24 h of exposure (Fig.
2a and b). Antibodies that specifically recognize the
dual phosphorylation target characteristic of ERK1/2 show a steady level during
the first 30 min (Fig. 2c), a low level after 24 h and a "recovery"
after 48 h. This is consistent with a state in which the AGL2592-resistant population
coexisted with the sensitive cell population before the treatment has been started,
here the resistant cells reflect a straight forward selection of an existing,
rather than a gradually developing, phenotype.
Differentially Exhibited Proteins in Agl2592-resistant Cells
The putative constant level of Bcl2 in residual MCF7 cells (relative to
untreated) raised the following question: Is the sustained unchanged level of
Bcl-2 (untreated Vs resistant cells at 48 h) in selected cells accompanied by
proteomic dissimilarities between the chemo-resistant (residual) cells and the
untreated cell population? To test this possibility, we treated MCF7 cells with
10 μM AGL2592 for 48 h and parallel neighboring cultures were treated with
the DMSO carrier to serve as untreated controls. The extracted proteomes of
AGL2592 and control cultures were run on a preparative two-dimensional (2D-gel)
electrophoresis and the twin gels were stained with Coomassie blue. Fig.
3 shows a differential protein abundance in highly expressed (or presented)
proteins that is prominent relatively to the background of many quantitativelly
similar spots. We have picked some of the differentially expressed protein spots
for identification by mass-spectrometry. Table 1 identifies
the indicated protein spots. Calreticulin and PDI (disulfide isomerase) showed
a lower abundance in the residual cells, 4.2 and 18 fold, respectively. Glyceraldehyde
phosphate dehydrogenase (GAPDH), aldolase-A and enolase-1 showed in the residual
cells a higher abundance than in the untreated controls, 1.97, 2.5 and 13.3
fold, respectively. In addition to the quantitative differences in protein abundance
two remarkable phenomena can be observed, one relates to the appearance of bovine
albumin in the gels and the other concerns the shift in the first dimension
(IEF) mobility of aldolase-A and GAPDH. Several versions of BSA molecules (bovine
serum albumin supplied through the fetal calf serum) appeared in a differential
manner between both cell populations. In the residual cells the balk of BSA
appeared mainly at its proper and expected relative-mobility (mr)
around 69 kDa, whereas in the gel of the untreated cells there was less BSA
at the mr of 69 kDa, however there was a substantial amount of BSA
at lower mr, possibly dimmers (140 kDa). In addition there were bands
of BSA running in between the 69 and 140 kDa with a tendency to have a progressively
slower mobility towards the acidic side of the 1st dimension. This kind of BSA
bands is not observed in the residual cell gel. The gel of the residual cells
showed in total 40% more albumin than that of the untreated cells.
Fig. 3: | Total proteomic 2D-gels, comparative between untreated and AGL2592-resistant MCF7 cells. MCF7 cells seeded 104 cells/cm2 in 175 cm2 flasks were exposed to 10 μM AGL2592 for 48 h and harvested on ice with RIPA in the presence of protease inhibitors. The cell lysates were precipitated in 95% ethanol at -20°C overnight and washed with 90% ethanol. The precipitates were suspended in IEF sample buffer and 1.3 mg protein from the treated and the untreated control cells and were run on 24 cm IEF strips with a pH range of 3-10. After equilibration with SDS/DTT and SDS/ iodoacetamide the 2nd dimension was run on 10% gels, fixed and stained with Coomassie blue. Left panel= AGL2592-treated residual cells, Right panel= untreated control cells (DMSO carrier). Arrows indicate enumerated spots identified in Table 1 that were picked for in-gel digestion, mass-spectrometry analysis and identification |
Table 1: | Differentially expressed protein spots in the comparative 2D-gels |
# Arbitrary density units, ## BSA with relative mobility of dimmers, * BSA with intermediate mobility, **GAPDH adjacent to aldolase |
The second remarkable phenomenon is the shift in aldolase-A and GAPDH to the acidic side in the untreated cell population. It should be noted that in both gels these two enzymes moved together on the IEF and were separated only during the 2nd dimension run, after equilibration with SDS+DTT followed by iodoacetamide treatment. The co-migration in the first dimension was seen also in the T47D gels (not shown). The difference in GAPDH abundance between the residual and control MCF7 populations was only 1.97 fold in favor of the residual cells.
Fig. 4: | Mass spectra comparison between glycerophoaphate dehydrogenase (GAPDH) protein spots. Panel a. shows the spectrum of GAPDH of AGL2592-resistant cells gel. Panel b. shows mass spectrum of untreated control cells. The Arrow indicates the broken and adenylated fragment, I170-K186 |
Table 2: | Differential expression in tumor cells of proteins mentioned in Table 1 |
However, a fraction of the GAPDH molecules in the control cells was cleaved in an illegitimate residue (isoleucine170) that is not due to the trypsin in-gel treatment (Fig. 4). This residue is in midst of the catalytic domain of the GAPDH enzyme for which the MS analysis revealed an additional mass of 134 Daltons attached to the Ile170-K186 residues fragment. The additional 134 Daltons are identified as adenine that must be covalently bound to Ile170-K186 fragments (Fig. 4b), which were absent from the canonical GAPDH digest in the AGL2592-resistant cell population. This may constitute an added effect to the 1.97-fold abundance of GAPDH in favor of the residual cells because it might impede the normal enzyme activity in the control (sensitive) cells, for example in a negative dominant manner.
Table 2 shows that all the 5 proteins that were differentially expressed in the residual cells relative to the untreated cells are also known to be differentially expressed in malignant cells relative to their normal or less malignant counterparts. This indicates that after a minimal selection time of only one cell division (48 h) the pattern of differentially expressed proteins most probably belongs to cells that coexisted with the general MCF7 population and were simply selected by AGL2592.
Discussion
AGL2592 is a bistyrphostin (Ben-Bassat et al., 2002) that has been shown to inhibit stat3 phosphorylation, perhaps by virtue of interference with dimerization of each of two putative Stat3 molecules on their way to become activated. The efficiency of growth arrest and cell death induction in breast cancer cells by this compound, relative to AG1478 (EGFR inhibitor) is consistent with the claims (Garcia et al., 1997) that Stat3 is an important mediator of proliferation in breast and other cancer types (Catlett-Falcone et al., 1999). The low proportion (10%) of residual cells under AGL2592, that were seen in MCF7 cultures made this cell line a suitable candidate for proteomic comparisons exploiting this case in which resistant cells coexist with the sensitive population. This is simply because the lower the resistant cell percentage the less prominent will be the contamination effect of its differentially expressed proteins on the negative control of the untreated population. Stat3 can activate transcription of several anti-apoptotic proteins like Bcl-2, Bcl-X, survival proteins and ERKs (Alas and Bonavida, 2003). Therefore, inhibition of Stat3 is consistent with a lower Bcl-2 abundance (relatively to controls) during the first 24 h interval that is included in the doubling time of the first cell-generation. Beyond this first cell doubling time (48 h) the Bcl-2 in the residual cells equalized with that of the untreated total population that reflects stability in this protein level and supports our hypothesis of a selection effect. The dual specificity-kinase target, the Thr-Glu-Tyr (TEY for short) is a relatively uncommon motif among proteins, in addition to ERK1/2 it is found in the metabotropic receptor of CNS cells and in Bcl-2 (T178-E-Y180). This motif is thus associated with these anti-apoptotic proteins and its decreased phosphorylated version at the 24 h interval in MCF7 cells and the "recovery" after 48 h may be well suited for the survival of the residual cells. There must be a specific mechanism (probably other than the multidrug-resistance glycoprotein) that enables some cells to ignore the AGL2592 potential to inhibit Bcl-2 expression and TEY motif phosphorylation. In spite of the interest one may find in the effect of AGL2592 on apoptotic signaling pathways, its importance for the present work is only marginal. The main purpose of the present work is to show that the expression pattern of various proteins associated with tumor resistance to therapy may be present in a restricted population of cells as early as one cell division after the therapy starts. This hard core of resistant cells may very well coexist with the sensitive cells even before the start of therapy.
With the hypothesis of coexistence of therapy sensitive and resistant cell populations in mind we looked at the Coomassie blue detectable proteins in differential 2D-gels. The most striking differences between the gels turned out to be the various versions of bovine serum albumin (BSA) that one would normally dismiss as a culture artifact derived from the fetal calf serum. However, the striking differences between the mobility in several BSA molecular species in the untreated MCF7 cells in contrast with their absence in the resistant cell gels, deserves a different logic. Based on the longstanding clinical information on serum albumin in cancer one could refer to BSA as a surrogate albumin replacing in culture the human serum albumin (HSA).
Serum albumin deficiency (inversion of albumin/globulin ratio) is a reliable marker for unfavorable prognosis in individuals suffering from one of several malignant tumors (Lis et al., 2003) e.g., breast cancer. Albumin, synthesized mainly in the liver, is the most abundant serum protein; it is functioning as an oncotic-pressure protein by keeping a proper intra-vascular volume. Albumin binds lipids, fatty acids, ingested chemicals and a variety of small compounds, perhaps serving as a buffer or carrier of drugs to peripheral tissues. Albumin can cross (in vivo) the vascular wall into the Extra Cellular Fluid (ECF) compartment across endothelial cells by a process known as transcytosis (Schnitzer, 1992) which is thought to be mediated by cell surface albumin-binding proteins. These binding proteins are considered to be albumin receptors, i.e., gp60 (albondin) (Schnitzer et al., 1988), osteonectin (SPARK), gp18 and gp30 (Schnitzer and Oh, 1992). Albumin is a target for modifications (in addition to phosphorylation), it may undergo oxidation of its SH residues (Havea et al., 2000) to form adducts and aggregates, albumin dimers may result also from reaction with hydroxyl radicals via generation of intermolecular bityrosins (Davies and Delsignore, 1987), nitration is also a possible modification (Greenacre et al., 1999; Koeck et al., 2004). It has been suggested that non-modified albumin is more avidly bound by gp60, in contrast, albumin modification by radicals increases its affinity to gp30 and gp18 (Schnitzer et al., 1992) and makes it more susceptible to breakdown by proteases (Davies et al., 1987). Aggregated and adduct albumin molecules can be distinguished by SDS-PAGE according to their mobility that differs from the unmodified molecules, these can be better exhibited in 2D gels (Havea et al., 2000, 2001) and picked for mass spectrometry analysis. Recently it has been shown that several albumin receptors belong to the nuclear heterogeneous ribonuclear proteins (hnRNP) family (Fritzsche et al., 2004) such as hnRNP A2/B1 which are known to increase in tumor cells (Satoh et al., 2000) and were identified as the above mentioned gp30 and gp18, respectively. In addition calreticulin (CRT) was also found to bind albumin, this raises the possibility that the untreated MCF7 cells with higher CRT expression were more efficient in removing the unmodified albumin, perhaps explaining their lower level in untreated cells. Significantly, the resistant cells were more efficient in removing the modified slow migrating albumin species and less so in removing the normally migrating molecules. CRT, in addition to its act as an albumin binding protein (Fritzsche et al., 2004), may interact with glucocorticoid and androgen hormone receptors. By this activity it could influence the function of these receptors in BC cells especially in MCF7 cells that are known as a sex hormone responsive model. CRT has been recently shown to inhibit the expression of GLUT-1 (glucose transporter-1) by negative regulation at the un-translated (UTR) GLUT-1 mRNA level(Totary-Jain et al., 2005). This may explain part of the metabolic aspect of the resistance in the residual cells that due to lower CRT might express GLUT-1 more abundantly (for a high glucose uptake) which would support the glycolytic energy metabolism, a known advantage in malignant cells.
Protein Disulfide Isomerase (PDI) that has a chaperon effect by regulation of protein structure around disulfide bonds, among others it is important as a member of a tetramer with α-Peptidyl-prolyl hydroxylase (α-PPH) for protection and efficient secretion of collagen. Therefore a lower expression of PDI may be interpreted as a lesser dependence of the therapy resistant cells on extra cellular matrix collagen with an increased propensity in detachment to metastasize. PDI may perhaps interact as a chaperon with β-PPH, which senses levels of oxygen controlling HIF-1 turnover. Lower expression of PDI in the therapy resistant cells is consistent with upgrading of the Warburg effect in which the tumor cells can prevail in much lower oxygen levels and may afford decreased sensitivity to changes in oxygenation.
Aldolase-A glycolytic enzyme has been shown to appear in metastasis of BC cells. In the gel of the sensitive cells it has moved almost one pH unit to the acidic side, indicating a possible modification.
Glyceraldehyde 3-phosphate dehydrogenase (GAPDH), another glycolytic enzyme that ran toward the acidic side, in the gel of the pretreated sensitive cells, perhaps due to modification.
Fig. 5: | Consensus sequences of adenine ring-binding enzymes aligned with MCF7 GAPDH I170-K186 fragment. Consensus sequences 1-5 and 7-8 represent Adenylate-photoaffinity-protected sequences of enzymatic digests as shown in several publications (Chavan et al., 1993; Salvucci et al., 1992; Sankaran et al., 1996). The marked residues V178, A180, I181, conform to the adenine-ring-binding consensus of aligned enzymes according to polar/non-polar amino acid classification. The low-case residues do not conform to the aligned motifs |
The adenine-conjugate in the I170-K186 fragment that was seen in the therapy sensitive cell population may reflect part of the post translational modifications responsible for the shift in first dimension migration of the putative complex GAPDH/aldolase. It is difficult to argue that the adenine molecule is a residue broken off the NAD+ coenzyme normally utilized by the GAPDH, because according to electron density analysis (Cowan-Jacob et al., 2003) the adenine residue of NAD+ has affinity to amino acids close to but still excluded from this fragment. The nicotinic moiety of NAD is bound to Cys149 (Cys152 in human GAPDH) on the N-terminal side of I170 and the adenine moiety of NAD may be in contact with His179 and other residues much closer to the C-terminus of GAPDH. In a monomeric form Cys152 and His179 are potentially bound by a hydrogen bond (Nagradova, 2001). It is not clear whether NAD or hydrogen bonding could staple together temporarily 2 cleavage products (cleaved at I170) to enable their co-migration with intact GAPDH monomer molecules on the 2nd dimension under electrophoretic strain. GAPDH can fulfill functions unrelated to glycolysis and appear in forms of oligomeric structures, it may be active in membrane fusion, DNA repair and replication, RNA transport, apoptosis and microtubule bundling (Sirover, 1999). Its DNA repair capacity could be reflected in adenine binding of the above described fragment similar to canonical DNA ligases that utilize a lysine residue as an adenylation site for their activity, which makes such an adenylated lysine resistant to trypsin cleavage (Tomkinson et al., 1991). However, the I170-K186 fragment doesn't contain a trypsin un-cleaved lysine-end, therefore the adenine is probably bound to a residue other than lysine. In Fig. 5 this GAPDH-derived I170-K186 fragment is aligned with consensus sequences that were found to be adenine-ring binding motifs. Motifs 1-5 (Chavan et al., 1993) and motifs at No.8 (Sankaran et al., 1996) are derived from hydrogenases and motifs at No. 7 (Salvucci et al., 1992) are from adenylate kinase. According to this alignment the three marked residues in this GAPDH fragment may bind a dephosphorylated adenine-ring. Poly (ADP-ribose) polymerase (PARP) can bind and inhibit the activity of oxidation-modified GAPDH (Du et al., 2003), it might underlay the adenylation of the I170-K186 fragment.
Note that the microtubule bundling ability of GAPDH may reflect its co-migration with aldolase A during IEF. These two enzymes may be non-covalently bound to each other and to cytoskeleton components (Volker et al., 1995) to form microtubules networks as an enzyme complex, supporting their catalytic activities on their consecutive substrate/product targets.
Enolase 1 (phosphopyruvate hydratase) appeared at its expected isoelectric point in the therapy-resistant cells and it does not appear in the respective spot of the untreated cell population. Taken together, the relative abundance of these glycolytic enzymes in the AGL2592-treatment residual cells indicate that part of the proteomic basis for therapy resistance in MCF7 cells is associated with increased intensity of the Warburg effect, or the ability of this effect to persist under the therapeutic stress.
Acknowledgement
We thank Dr. Ofra Moshel, from the Hebrew University Medical School Proteomic Service, for her help in mass spectrometry.