Upregulation of ABC transporters contributes to chemoresistance of sphingosine 1-phosphate lyase-deficient fibroblasts.

Sphingosine 1-phosphate (S1P) is an extra- and intracellular mediator that regulates cell growth, survival, migration, and adhesion in many cell types. S1P lyase is the enzyme that irreversibly cleaves S1P and thereby constitutes the ultimate step in sphingolipid catabolism. It has been reported previously that embryonic fibroblasts from S1P lyase-deficient mice (Sgpl1−/−-MEFs) are resistant to chemotherapy-induced apoptosis through upregulation of B cell lymphoma 2 (Bcl-2) and Bcl-2-like 1 (Bcl-xL). Here, we demonstrate that the transporter proteins Abcc1/MRP1, Abcb1/MDR1, Abca1, and spinster-2 are upregulated in Sgpl1−/−-MEFs. Furthermore, the cells efficiently sequestered the substrates of Abcc1 and Abcb1, fluo-4 and doxorubicin, in subcellular compartments. In line with this, Abcb1 was localized mainly at intracellular vesicular structures. After 16 h of incubation, wild-type MEFs had small apoptotic nuclei containing doxorubicin, whereas the nuclei of Sgpl1−/−-MEFs appeared unchanged and free of doxorubicin. A combined treatment with the inhibitors of Abcb1 and Abcc1, zosuquidar and MK571, respectively, reversed the compartmentalization of doxorubicin and rendered the cells sensitive to doxorubicin-induced apoptosis. It is concluded that upregulation of multidrug resistance transporters contributes to the chemoresistance of S1P lyase-deficient MEFs.

and downregulation of class I HDACs, which contributed to the disturbed Ca 2+ homeostasis with enhanced Ca 2+ storage and elevated basal intracellular free Ca 2+ concentration ([Ca 2+ ] i ) in these cells ( 26 ).
The starting point of our present study was the hypothesis that this accumulation of S1P in S1P lyase-defi cient MEFs may induce counterregulatory mechanisms such as upregulation of S1P secretion, which would then keep cytosolic S1P at a normal level while nuclear pools of S1P probably have no access to the export mechanisms. Therefore, we studied the expression and functionality of the transporters that have been implicated in S1P export, i.e., spinster-2, Abcc1, Abca1, and Abcg2, and also of the multidrug resistance protein Abcb1, which is known to be regulated by SphK and S1P ( 27 ). We demonstrate that several of the transporters were upregulated in Sgpl1 Ϫ / Ϫ -MEFs and caused a sequestration of the ABC transporter substrates fl uo-4 and doxorubicin. Furthermore, we show that the compartmentalization of doxorubicin contributes signifi cantly to the chemoresistance of these cells. These results link S1P lyase to the regulation of multidrug transporters and suggest that this activity, in addition to S1P's ability to interfere with apoptotic signaling pathways, plays an important role in cancer cell chemoresistance.

Cell culture
Embryonic fi broblasts from S1P lyase-defi cient and corresponding wild-type mice had been prepared as described previously ( 25 ). The cells were cultured in DMEM/F12 supplemented with 100 U/ml penicillin G, 0.1 mg/ml streptomycin, and 10% FCS in a humidifi ed atmosphere of 5% CO 2 /95% air at 37°C. If not stated otherwise, the cells were kept in serum-free medium overnight before experiments.

Measurements of S1P and sphingosine in cellular supernatants
The cells were seeded onto 3.5 cm dishes and grown to near confl uence. They were kept in serum-free medium supplemented with 1 mg/ml fatty acid-free BSA in the absence or presence of 10 µM sphingosine for 16 h. Thereafter, the cellular supernatants (1.5 ml) were collected and centrifuged for 10 min at 1,800 × g and 4°C. The supernatants were transferred into fresh tubes and supplemented with 1.2 ml methanol containing 20 ng/ml D -erythro -C17-sphingosine and 20 ng/ml D -erythro -C17-S1P (Avanti Polar Lipids Inc., Alabaster, AL) as internal standards. Then 35 µl 1 M HCl, 70 µl 10% KCl, and 2 ml chloroform the transporter proteins that have been shown to export S1P, there are several members of the ABC transporter family such as the multidrug resistance-related protein Abcc1, the cholesterol transporter Abca1, and the breast cancer resistance protein Abcg2 (9)(10)(11). While these ABC transporters either do not regulate plasma concentrations of S1P or are redundant with respect to this activity ( 12 ), plasma S1P is regulated by the non-ABC transporter-related transport protein spinster-2, which is probably a specifi c S1P transporter ( 13,14 ).
While the role of ABC transporters in chemoresistance of cancers is widely recognized ( 1,8,15 ), the role of the enzymes that catalyze S1P formation and degradation in cancer growth and survival appears to be less clear. SphK1 is upregulated in many cancers, and cancer cells might have a "non-oncogene addiction" for this enzyme ( 16 ); however, new highly potent SphK1 and SphK1/2 inhibitors failed to inhibit cancer cell proliferation and growth of tumor xenografts in mice ( 17,18 ). Recently, also S1P lyase, which is an endoplasmic reticulum (ER) resident enzyme that cleaves S1P irreversibly and catalyzes the ultimate step in sphingolipid catabolism, has been connected with cancer growth and chemoresistance (19)(20)(21). Early studies have shown that overexpression of S1P lyase rendered the cells sensitive to apoptosis induced by serum deprivation or chemotherapeutic agents, while siRNA-induced knockdown diminished apoptosis at baseline and in response to chemotherapy (22)(23)(24). Recent studies show that S1P lyase is downregulated in human colon cancers ( 24 ), human melanoma cell lines ( 19 ), and human prostate cancers ( 21 ). Fibroblasts lacking S1P lyase were able to form colonies in soft agar and to induce tumors in immunocompromised mice ( 19 ). In human prostate cancers, S1P lyase expression and activity was inversely correlated with clinical malignancy scores, and cell death of prostate cancer cell lines, induced by irradiation or chemotherapy, was reduced by S1P lyase knockdown and potentiated by S1P lyase overexpression ( 21 ).
Although a low expression of S1P lyase thus appears to be favorable for proliferation and survival of cancer cells, the presence of the enzyme is essential for development and survival of the mammalian organism as a whole. Thus, a full knockout of S1P lyase in the mouse induced growth retardation and early death after a few weeks, accompanied by severe organ damage, immunosuppression, a major disturbance of lipid homeostasis, and a proinfl ammatory phenotype [for review, see Aguilar and Saba ( 20 )]. On the other hand, mouse embryonic fi broblasts (MEFs) from S1P lyase-defi cient mice ( Sgpl1 Ϫ / Ϫ -MEFs) proliferated without retardation and, in the absence of serum, even better than MEFs from wild-type mice ( 19 ). The resistance of Sgpl1 Ϫ / Ϫ -MEFs to apoptosis induced by doxorubicin and etoposide was ascribed to upregulation of the antiapoptotic proteins Bcl-2 and Bcl-xL because a combined knockdown of both proteins with siRNA improved the sensitivity of the cells ( 19 ). In our own studies on these S1P lyase-defi cient fi broblasts, we observed that S1P accumulated by ‫ف‬ 5-fold compared with wild-type cells when whole cells were extracted and ‫ف‬ 50-fold in nuclear preparations ( 25,26 ). Furthermore, we observed a reduced HDAC activity RevertAid fi rst strand cDNA synthesis kit (Fermentas, St. Leon-Rot, Germany). Real-time PCR was performed with the Applied Biosystems 7500 Fast Real-Time PCR System. Probes, primers, and the reporter dyes 6-FAM and VIC were from Applied Biosystems (Darmstadt, Germany). The cycling conditions were 95°C for 15 min (1 cycle), followed by 95°C for 15 s and 60°C for 1 min (40 cycles). mRNA expression levels were analyzed by the ⌬ ⌬ C t method with GAPDH as reference.

Data analysis and presentation
Averaged data are means ± SEM from the indicated number (n) of independent experiments or means ± SD from a representative experiment performed with n replicates. Graphical presentation and statistical analysis were performed with Prism-5 (GraphPad Software, San Diego, CA). Microscopic images are representative for at least three similar experiments and were edited using the LSM Image Browser or the ZEN software (http:// www.zeiss.com/micro).

RESULTS
Because S1P accumulated in S1P lyase-defi cient MEFs, we hypothesized that the cells would try to counteract this by upregulating S1P secretion. Therefore, we analyzed the expression of the multidrug transporters Abcc1 (= MRP1), Abca1, and Abcg2, which all have been associated with S1P extrusion, of Abcb1 (= MDR1 or p-glycoprotein), which is known to be regulated by S1P metabolism ( 27 ), and of spinster-2, which appears to be a specifi c S1P transporter ( 13,14 ). Indeed, the mRNA levels of Abcc1 and Abca1 were upregulated by ‫ف‬ 2-to 3-fold in Sgpl1 Ϫ / Ϫ -MEFs, while Abcg2 remained unaltered ( Fig. 1A ). Of the two transcripts of Abcb1, Abcb1a was strongly induced by ‫ف‬ 10-fold, while Abcb1b was upregulated by ‫ف‬ 2-fold ( Fig. 1A ). Spinster-2 mRNA was induced by ‫ف‬ 3-fold in Sgpl1 Ϫ / Ϫ -MEFs ( Fig. 1A ). On the protein level, both the unglycosylated form at ‫ف‬ 130 kDa and the glycosylated mature form at ‫ف‬ 170 kDa of Abcc1 were added, and after thorough mixing and centrifugation, the organic phase was collected. The aqueous phase was reextracted two times with chloroform, and the organic phases were combined and dried down. The lipids were redissolved in 200 µl DMSO containing 2% HCl, and LC-MS/MS was performed as previously described ( 28 ). The cell pellets were washed, scraped into lysis buffer, and subjected to protein measurements.

H]sphingosine in cells and supernatants
The cells were seeded onto 3.5 cm dishes and grown to near confl uence. Before experiments, they were kept for 16 h either in serum-free medium supplemented with 10 mg/ml fatty acid-free BSA or in medium containing 10% FCS. Labeling was performed in the respective media for 2 h with 0.5 µCi/ml [ 3 H]sphingosine. Then, the cells were washed twice and incubated for a further 4 h either in serum-free medium supplemented with 10 mg/ml fatty acid-free BSA or in medium containing 10% FCS. Thereafter, the cellular supernatants (1 ml) were collected, and 1 ml methanol, 70 µl 10% KCl, 35 µl 1 M HCl, and 2 ml chloroform were added. Cell monolayers were washed with ice-cold PBS and scraped into 1 ml methanol. The dishes were washed with 1 ml methanol, and 1.6 ml of high salt solution (0.74% KCl, 0.04% CaCl 2 , 0.034% MgCl 2 ), 35 µl 1 M HCl, and 2 ml chloroform were added. Lipid extraction was performed as described for the LC-MS/MS measurements. The dried samples were redissolved in 50 µl methanol and separated by TLC with 1-butanol:acetic acid:water 3:1:1. Areas containing S1P and sphingosine, respectively, were identifi ed with nonradioactive standard samples and scraped off the TLC plates, and radioactivity was quantifi ed by liquid scintillation counting. Separate dishes were used for protein measurements.

Fluorescence microscopy
For microscopic analysis, the cells were cultured on 8-well chambered coverslides (µ-slide; ibidi GmbH, Martinsried, Germany) coated with poly-L -lysine. If not stated otherwise, the cells were washed with HBSS and kept in HBSS at room temperature during the measurements. Confocal laser scanning microscopy was performed with a Zeiss LSM510 Meta system equipped with an inverted Observer Z1 microscope and a Plan-Apochromat 63×/1.4 oil immersion objective (Carl Zeiss MicroImaging GmbH, Göttingen, Germany). The following excitation (ex) laser lines and emission (em) fi lter sets were used: fl uo-4, ex 488 nm, em 505 nm long pass fi lter or 505-530 nm band pass fi lter (when measured in combination with tetramethylrhodamine); tetramethylrhodamine and LysoTracker Red DND-99, ex 543 nm, em 560 nm long pass fi lter; ER-Tracker Blue-White DPX, ex 364 nm, em 475 nm long pass fi lter; Hoechst 33342, ex 405 nm, em 420-480 nm band pass fi lter; and doxorubicin, ex 488 nm, em 575-630 nm band pass fi lter. Simultaneous staining with more than one dye was analyzed in multitracking mode.

Immunocytochemistry
The subcellular distribution of Abcb1 was analyzed by immunocytochemistry. Cells grown on 8-well chambered coverslides were fi xed with formaldehyde and stained with anti-p-glycoprotein antibody (Sigma-Aldrich Chemie GmbH) followed by Alexa-Fluor 555-conjugated anti-mouse antibody (Invitrogen GmbH). The fl uorescence was monitored by confocal microscopy as described above using the 543 nm excitation laser line and a 575-630 nm emission band pass fi lter. To further address the functional signifi cance of ABC transporter upregulation, we analyzed next whether Sgpl1 Ϫ / Ϫ -MEFs were able to extrude the substrate of Abcc1, fl uo-4. However, we observed only a slightly diminished fl uo-4 loading in Sgpl1 Ϫ / Ϫ -MEFs (data not shown). Most interestingly, when analyzing the subcellular distribution of fl uo-4, we observed that fl uo-4 was strongly compartmentalized in Sgpl1 Ϫ / Ϫ -MEFs, while it was evenly distributed throughout the cytosol and nuclei of wild-type cells ( Fig. 3 ). Two different cellular staining patterns of fl uo-4 were observed in Sgpl1 Ϫ / Ϫ -MEFs, a rather punctuate and a rather reticular staining pattern, respectively ( Fig. 4 ). The fl uo-4-containing organelles did not colocalize with tetramethylrhodamine, the mitochondrial marker ( Fig. 4A ). However, the reticular staining pattern overlapped with an ER tracker, while the punctuate staining pattern colocalized with a lysosomal tracker ( Fig. 4B, C ). It has to be taken into account that these MEFs are polyclonal cell populations. Therefore, the two different staining patterns could represent two main different cell types.
It has been shown previously that Sgpl1 Ϫ / Ϫ -MEFs are resistant to chemotherapy-induced apoptosis, in particular to apoptosis induced by etoposide and doxorubicin ( 19 ). Doxorubicin is a substrate of Abcb1 and Abcc1, but the were upregulated in Sgpl1 Ϫ / Ϫ -MEFs ( Fig. 1B ). The expression of the mature form was quantifi ed by densitometry and found to be enhanced by ‫ف‬ 2-fold ( Fig. 1B ). Because it is known that HDAC inhibitors can upregulate multidrug resistance transporters in tumor cells, and because HDAC activity and expression of class I HDACs were decreased in Sgpl1 Ϫ / Ϫ -MEFs, we analyzed whether a treatment with the pan-HDAC inhibitor TSA induced an upregulation of the transporters in wild-type cells. However, TSA did not upregulate Abcc1, Abcb1, or Abca1 in either Sgpl1 +/+ -or Sgpl1 Ϫ / Ϫ -MEFs (data not shown). In contrast, TSA strongly induced the expression of spinster-2 by ‫ف‬ 10-fold in both Sgpl1 +/+and Sgpl1 Ϫ / Ϫ -MEFs ( Fig. 1A ). Spinster-2 expression thus appears to be regulated by histone acetylation.
Because the transporters that could potentially transport S1P were upregulated in Sgpl1 Ϫ / Ϫ -MEFs, we analyzed whether S1P was elevated in the supernatants of these cells. However, although total S1P levels were ‫ف‬ 5-fold higher in S1P lyase-defi cient MEFs ( 25 ), the concentrations of S1P in the supernatants were similar in wild-type and lyase-defi cient MEFs ( Fig. 2A ). This is in agreement with a previous report ( 29 ). In addition, the concentrations of sphingosine in the supernatants were similar in both cells types ( Fig. 2A ), although total cellular sphingosine levels were ‫ف‬ 2-fold higher in the knockout MEFs ( 25 ). Moreover, treatment with 10 µM external sphingosine overnight to further challenge S1P metabolism in the lyase-defi cient MEFs did not signifi cantly elevate S1P in the supernatants, and although sphingosine was generally higher in supernatants of the treated cells, there was again no difference between Sgpl1 +/+ -and Sgpl1 Ϫ / Ϫ -MEFs ( Fig.   Fig. 1

. Upregulation of ABC transporters and spinster-2 in Sgpl1
Ϫ / Ϫ -MEFs. A: mRNA expression was determined by quantitative real-time PCR in Sgpl1 Ϫ / Ϫ -and Sgpl1 +/+ -MEFs. In some experiments, the cells were treated with vehicle or 1 M TSA as indicated for 16 h before isolation of mRNA. The data were normalized to untreated wild-type cells according to the ⌬ ⌬ C t method with GAPDH as reference (means ± SEM of the indicated number of experiments; * P < 0.05, ** P < 0.01 in one-sample ttest). B: Protein expression of Abcc1 was analyzed by Western blotting. Shown is a representative blot, which was reprobed with anti-␤ -actin, and a densitometric analysis of the expression of the glycosylated mature form of Abcc1 at 170 kDa (n = 3 independent experiments each performed in duplicate; means ± SEM; * P < 0.05 in one-sample t -test).
the transporter was indeed only partially localized at the plasma membrane and instead was situated mainly at intracellular punctuate compartments ( Fig. 6B ). Finally, we analyzed the infl uence of a combined pretreatment with the specifi c Abcb1 inhibitor zosuquidar and the Abcc1 inhibitor MK571. Again, this treatment effectively prevented the nuclear exclusion of doxorubicin by Sgpl1 Ϫ / Ϫ -MEFs, and some but not all cells showed small condensed nuclear staining previous study did not fi nd a difference in cellular uptake of doxorubicin between Sgpl1 +/+ -and Sgpl1 Ϫ / Ϫ -MEFs ( 19 ). Therefore, we wondered whether doxorubicin was not extruded but rather compartmentalized in the knockout MEFs, and we analyzed the subcellular distribution of doxorubicin by monitoring its autofl uorescence by confocal microscopy. As shown in Fig. 5 , doxorubicin, which intercalates into the DNA, initially localized to the nuclei in both Sgpl1 +/+ -and Sgpl1 Ϫ / Ϫ -MEFs; only minor staining outside the nuclei was observed after 1 h. However, after 16 h of incubation, there was a major difference between the two cell types. Doxorubicin was still bound to the DNA in wild-type MEFs, which had small condensed nuclei suggestive of apoptosis. In contrast, doxorubicin was completely excluded from the nuclei of Sgpl1 Ϫ / Ϫ -MEFs and strongly compartmentalized in predominantly punctuate structures ( Fig. 5 ). Double staining revealed that there was a partial overlap between the compartments that accumulated fl uo-4 and doxorubicin, respectively (data not shown). To verify that doxorubicin compartmentalization was caused by ABC transporters localized on cellular organelles, we applied an inhibitor of organic anion transporters, probenecid; the Abcc1 inhibitor MK571; and the nonspecifi c Abcb1 inhibitor verapamil ( Fig. 6A ). Of these inhibitors, only verapamil was able to partially prevent doxorubicin compartmentalization. As shown in Fig. 6A , some of the knockout MEFs that had been treated with doxorubicin in the presence of verapamil had small condensed nuclei containing doxorubicin, suggestive of apoptosis, while other cells kept doxorubicin compartmentalized but at the same time could not fully exclude doxorubicin from their nuclei ( Fig. 6A ). These observations suggest that Abcb1 at least partially accounted for doxorubicin compartmentalization in Sgpl1 Ϫ / Ϫ -MEFs. Staining of the cells with an antibody against Abcb1 revealed that 3 H]sphingosine in either serum-free medium containing 10 mg/ml BSA or in medium containing 10% FCS, as indicated. After washing and further incubation for 4 h in the respective media, the lipids were extracted from cells and supernatants, separated by TLC and quantifi ed by liquid scintillation counting (means ± SD from a representative experiment; n = 3; * P < 0.05, ** P < 0.01, *** P < 0.001 in t -test comparing Sgpl1  proapoptotic agent staurosporine reduced the viability of wild-type cells by more than 75% at 10 nM and more than 90% at у 30 nM, while the viability of Sgpl1 Ϫ / Ϫ -MEFs was reduced by ‫ف‬ 50% at 10-500 nM, and the chemoresistance of the S1P lyase-defi cient cells was overcome at 1 µM of staurosporine ( Fig. 8A ). A combined inhibition of Abcc1 and Abcb1 by pretreatment with zosuquidar plus MK571 had no further infl uence on reduction of cell viability by doxorubicin or staurosporine in wild-type MEFs ( Fig. 8B ). Furthermore, doxorubicin-induced caspase-3 cleavage was minimally infl uenced, if at all, and staurosporine-induced caspase-3 cleavage was not infl uenced by the treatment with the transporter inhibitors in these cells ( Fig. 8C ). In contrast, in Sgpl1 Ϫ / Ϫ -MEFs, the treatment with zosuquidar plus MK571 rendered the cells sensitive to doxorubicin ( Fig. 8B ). Similarly, the minor caspase-3 cleavage induced by the doxorubicin treatment in Sgpl1 Ϫ / Ϫ -MEFs was strongly augmented ( Fig. 7 ). The treatment with zosuquidar plus MK571 in the absence of doxorubicin did not alter the nuclear morphology in either cell type, as shown in the insert of Fig. 7 . These observations confi rm the important role of Abcb1 and Abcc1 in nuclear exclusion of doxorubicin in Sgpl1 Ϫ / Ϫ -MEFs and suggest a role for this mechanism in chemoresistance of the cells.
Finally, we analyzed whether the observed upregulation of the multidrug transporters contributed to the resistance of Sgpl1 Ϫ / Ϫ -MEFs against chemotherapy-induced apoptosis by measuring the infl uence of ABC transporter inhibitors on cell viability and caspase-3 cleavage. As shown in Fig. 8A , the viability of wild-type MEFs was strongly reduced by incubation for 16 h with 1 µM doxorubicin, while Sgpl1 Ϫ / Ϫ -MEFs were not signifi cantly affected by this treatment, in agreement with the results of the microscopic analysis described above and with the data of Colié et al. ( 19 ). Incubation for 16 h with the  sporine-induced caspase-3 cleavage ( Fig. 8C ). Taken together, these data suggest that a large part of doxorubicin resistance of S1P lyase-defi cient MEFs was caused by upregulation of the multidrug transporters Abcc1 and Abcb1, while the sensitivity to staurosporine-induced apoptosis was not dependent on Abcc1 or Abcb1 in these cells.

DISCUSSION
For more than 20 years, ceramide and S1P have been implicated in regulation of cell survival, apoptosis, and chemoresistance, and yet surprisingly little is known about the mechanisms by which the biostat works on the molecular level. While G protein-coupled S1P receptors account for activation of survival signaling pathways by extracellular S1P, such as protein kinase B (Akt)/mechanistic target of rapamycin (mTOR) or ERK, the direct targets involved in antiapoptotic signaling by intracellular S1P are less clear ( 2,3,30 ). Recently, a direct activation of the proapoptotic effector molecules BAK and BAX by S1P and hexadecenal, respectively, has been demonstrated in a reconstitution system with purifi ed mitochondria ( 31 ); however, this mechanism can explain the occasionally proapoptotic but not the widespread antiapoptotic effect of S1P, although it might be in line with an antiapoptotic role of S1P lyase knockout with reduced formation of hexadecenal. Indeed, the product of the S1P lyase reaction, trans -2-hexadecenal, induced apoptosis in several cell lines, and this indeed also by zosuquidar plus MK571 ( Fig. 8C ). Compared with doxorubicin, the effect of staurosporine was not signifi cantly affected by the transporter inhibitors ( Fig. 8B ). Furthermore, the treatment with the inhibitors was not able to augment stauro-  expression. Furthermore, extracellular S1P stimulated Abcb1 transport activity via S1P 1 and S1P 3 receptors, and as data not shown, it was mentioned that exogenous S1P did not modulate Abcb1 expression, suggesting differential activities of intracellular SphK1 and extracellular S1P ( 27 ). On the other hand, Cannon et al. ( 33 ) and Miller ( 34 ) showed recently that exogenous S1P decreased Abcb1 activity via S1P 1 in isolated mouse brain and spinal cord capillaries, as well as in isolated killifi sh renal proximal tubules, respectively. Because this effect was rapid and reversible, it did not rely on transcriptional regulation, but either on alteration of the transporter turnover number or on traffi cking of the transport protein away from the exterior surface of the luminal plasma membrane ( 33 ). In line with this activity, exogenous cell permeable C 2 -ceramide was able to increase cholesterol effl ux to apolipoprotein A-I by increasing the cell surface presence of Abca1 ( 35 ). However, these data provide no possible explanation for why several diverse ABC transporters were upregulated in S1P lyase-defi cient MEFs.
It is known that inhibition of HDACs can lead to the induction of a broad range of multidrug transporters, which constitutes a major problem in the development of HDAC inhibitors for cancer treatment [see, e.g., Hauswald et al. ( 36 )]. We had observed a reduced expression of class I HDACs and a decreased HDAC activity in Sgpl1 Ϫ / Ϫ -MEFs ( 26 ), in line with the role of nuclear S1P in regulating HDAC activity reported previously ( 37 ). Therefore, it was reasonable to assume that upregulation of the transporters could be a result of HDAC inhibition in these cells. However, the involved BAX activation, but it was dependent on upstream activation of c-Jun N-terminal kinase (JNK) and was prevented by the antioxidant N -acetylcysteine ( 32 ). The chemoresistance of MEFs isolated from S1P lyase knockout mice, however, has been traced back to upregulation of Bcl-2 and Bcl-xL ( 19 ). Here, we add another facet to the picture by demonstrating a link between S1P lyase and transcriptional regulation of multidrug transporters, which we demonstrate to be functionally relevant for the wellknown chemoresistance of S1P lyase-defi cient MEFs.
Following the hypothesis that accumulation of S1P in S1P lyase-defi cient MEFs might induce a counterregulatory upregulation of S1P secretion, in our investigation we focused on ABC transporters that have been associated with transport of S1P (Abcc1, Abca1, and Abcg2) or regulation by S1P (Abcb1). We observed that Abcc1 mRNA and protein, and Abca1 as well as Abcb1 mRNA, were upregulated in Sgpl1 Ϫ / Ϫ -MEFs, while Abcg2 mRNA expression was not signifi cantly altered. Although, with the exception of the Abcb1a transcript, which was strongly induced, the upregulation on the mRNA level was only ‫ف‬ 2-to 3-fold, the microscopic images displaying the compartmentalization of the fl uo-4 and doxorubicin in Sgpl1 Ϫ / Ϫ -but not wild-type MEFs clearly demonstrate the functional effectiveness of the transporters in the knockout cells.
Not much is known so far about the link between multidrug transporters and S1P metabolism. In a pioneering study, Pilorget et al. ( 27 ) have shown that overexpression of SphK1 in a cerebral endothelial cell line induced upregulation of Abcb1 protein and Abcb1b, but not Abcb1a, mRNA In a previous study, we had used the fl uorescent Ca 2+ sensor, fura-2, for [Ca 2+ ] i measurements in the MEFs ( 25 ). In this context, it is important to note that fura-2, unlike fl uo-4, is a substrate of organic anion transporters [see, e.g., Di Virgilio et al. ( 39 )] and was not compartmentalized in Sgpl1 Ϫ / Ϫ -MEFs (data not shown). Furthermore, we had complemented our fura-2 experiments with a series of experiments using the cameleon Ca 2+ sensor, which is a cytosolic protein, and we obtained similar results with both methods ( 25 ). This indicates that fura-2 is suitable for [Ca 2+ ] i measurements in Sgpl1 Ϫ / Ϫ -MEFs, while fl uo-4 is not suited. It is well-known that there are large differences between the low-molecular-weight Ca 2+ sensors with regard to their subcellular sequestration ( 40 ).
We provide several lines of evidence that the multidrug transporters contributed to the resistance of S1P lyasedefi cient MEFs against doxorubicin. Treatment with the unselective Abcb1 inhibitor verapamil and combined treatment with the selective Abcb1 inhibitor zosuquidar, plus the Abcc1 inhibitor MK571, decreased doxorubicin compartmentalization, increased nuclear localization of doxorubicin, and induced the occurrence of nuclear condensation in doxorubicin-treated Sgpl1 Ϫ / Ϫ -MEFs. When used without zosuquidar, the inhibitor of Abcc1, MK571, did not have a signifi cant infl uence on the doxorubicin sequestration and nuclear morphology of the cells, probably because the anthracyclin doxorubicin is a substrate of several ABC transporters, among which are Abcb1, Abcc1, and Abcg2 ( 41 ). Furthermore, the treatment with zosuquidar plus MK571 was highly active in many but not all cells, underlining the previously described heterogeneity of the MEFs ( 25 ) and suggesting that other multidrug transporters, which were not analyzed in the present study, contributed to the effect. Finally, the combined treatment with zosuquidar plus MK571 rendered the Sgpl1 Ϫ / Ϫ -MEFs sensitive to doxorubicin-induced cytotoxicity and apoptosis. Interestingly, the infl uence of the inhibitors on staurosporine-induced cytotoxicity was not signifi cant, and staurosporine-induced caspase-3 cleavage was not enhanced by zosuquidar plus MK571. These data also suggest that other antiapoptotic mechanisms are active in these cells, which could be other not yet considered ABC transporters or Bcl-2 and Bcl-xL upregulation as described ( 19 ).
Taken together, our study presents the multidrug transporters as important targets of S1P metabolism and components of the chemoresistance of S1P lyase-defi cient cells. Because S1P lyase downregulation has been associated with cancer cell survival and malignancy (see the introduction), S1P lyase stimulation has been suggested as a strategy to sensitize cancer cells to anticancer treatment ( 21 ). Therefore, the infl uence of S1P lyase overexpression, and of yet to be identifi ed activators or inducers of this enzyme, on multidrug transporter expression deserves comprehensive investigation. Further studies are required to analyze in greater detail the infl uence of SphKs, S1P phosphatases, and their respective inhibitors on multidrug transporter expression, subcellular localization, and activity.
The authors thank Nicole Kämpfer-Kolb and Luise Reinsberg for expert technical assistance, and Nerea Ferreirós and Sandra unspecifi c HDAC inhibitor TSA did not upregulate the investigated transporters in wild-type cells (data not shown). Because expression of ABC transporters might be differentially affected by different HDAC isoforms, and because cell typespecifi c factors also appear to play a role ( 36 ), it remains unclear at present whether downregulation of HDAC activity and expression in Sgpl1 Ϫ / Ϫ -MEFs was responsible for the observed effect on the ABC transporters. On the contrary, spinster-2 expression was strongly induced by TSA in both wild-type and S1P lyase-defi cient MEFs, indicating that this transporter is subject to regulation by histone acetylation and suggesting that its induction in Sgpl1 Ϫ / Ϫ -MEFs relies on the reduced HDAC activity in these cells.
Interestingly, although nearly all known S1P transporters were upregulated in Sgpl1 Ϫ / Ϫ -MEFs, we did not observe altered concentrations of S1P or sphingosine in the cellular supernatants, in line with data from Karaca et al. ( 29 ). Even when high concentrations of potential acceptors for the lipids were added to the supernatants (i.e., 10 mg/ml BSA or 10% FCS), there was no enhanced secretion of [ 3 H]S1P or [ 3 H]sphingosine from S1P lyase-defi cient MEFs. We assume that this is due to the fact that the transporters were not localized at the cell surface but instead at intracellular compartments. By immunocytochemistry, we show that at least Abcb1 is localized to a large extent at intracellular vesicular structures in Sgpl1 Ϫ / Ϫ -MEFs. Furthermore, the distribution of fl uo-4, which is a substrate of Abcc1, strongly suggests that in these cells Abcc1 also resides mainly at intracellular structures, which we could identify as ER and lysosomes. Although we did not further analyze the localization of Abca1 and spinster-2, it is obvious that a dysfunctional insertion of the transporters into the plasma membrane would explain why S1P secretion was not enhanced in Sgpl1 Ϫ / Ϫ -MEFs. Furthermore, it explains why Colié et al. ( 19 ) did not observe a difference in doxorubicin uptake between wild-type and Sgpl1 Ϫ / Ϫ -MEFs. An interesting question is why Colié et al. did not observe doxorubicin compartmentalization even though they analyzed their cells by fl uorescence microscopy [ Fig. 3 in Colié et al. ( 19 )]. However, in our study, doxorubicin was initially localized in the nuclei of either cell type, and only with time was it redistributed into intracellular compartments in Sgpl1 Ϫ / Ϫ -MEFs. Therefore, the selected time window for analysis of the cells in the study of Colié et al. was probably unfavorable for detection of the drug's compartmentalization.
The reason why Abcb1 and Abcc1 were localized at intracellular compartments and not at the plasma membrane in Sgpl1 Ϫ / Ϫ -MEFs remains unclear at present. Abcb1 is known to be localized at the ER, where it is synthesized and folded; at the Golgi, where it is glycosylated; and at various endosomes as well as lysosomes ( 38 ). Its transport to the plasma membrane is dependent on the cytoskeleton, and its endocytosis and recycling is dependent on guanosine-5 ′ -triphosphate hydrolases (GTPases) of the Rab family. ( 38 ). It is tempting to speculate that there is a general membrane transport defect in S1P lyase-defi cient MEFs, as amyloid precursor protein processing is also impaired in these cells ( 29 ), but the observed effects could also be due to altered Rab signaling and/or defective recycling of the multidrug transporters.