Chronic Reduction of the Cytosolic or Mitochondrial NAD(P)-malic Enzyme Does Not Affect Insulin Secretion in a Rat Insulinoma Cell Line*

The cytosolic malic enzyme (ME1) has been suggested to augment insulin secretion via the malate-pyruvate and/or citrate-pyruvate shuttles, through the production of NADPH or other metabolites. We used selectable vectors expressing short hairpin RNA (shRNA) to stably decrease Me1 mRNA levels by 80–86% and ME1 enzyme activity by 78–86% with either of two shRNAs in the INS-1 832/13 insulinoma cell line. Contrary to published short term ME1 knockdown experiments, our long term targeted cells showed normal insulin secretion in response to glucose or to glutamine plus 2-aminobicyclo[2,2,1]heptane-2-carboxylic acid. We found no increase in the mRNAs and enzyme activities of the cytosolic isocitrate dehydrogenase or glucose-6-phosphate dehydrogenase, which also produce cytosolic NADPH. There was no compensatory induction of the mRNAs for the mitochondrial malic enzymes Me2 or Me3. Interferon pathway genes induced in preliminary small interfering RNA experiments were not induced in the long term shRNA experiments. We repeated our study with an improved vector containing Tol2 transposition sequences to produce a higher rate of stable transferents and shortened time to testing, but this did not alter the results. We similarly used stably expressed shRNA to reduce mitochondrial NAD(P)-malic enzyme (Me2) mRNA by up to 95%, with severely decreased ME2 protein and a 90% decrease in enzyme activity. Insulin release to glucose or glutamine plus 2-aminobicyclo[2,2,1]heptane-2-carboxylic acid remained normal. The maintenance of robust insulin secretion after lowering expression of either one of these malic enzymes is consistent with the redundancy of pathways of pyruvate cycling and/or cytosolic NADPH production in insulinoma cells.

NADP ϩ -dependent isocitrate dehydrogenase has been reported to impair glucose-stimulated insulin secretion, pyruvate cycling, and NADPH rise in response to glucose (20). If confirmed, these results suggest either that isocitrate dehydrogenase is the primary source of cytosolic NADPH in these cells or that multiple routes for NADPH production exist but cannot fully compensate for each other. These results are consistent with the finding that normal mouse islets contain little, if any, cytosolic malic enzyme (21,22), whereas mouse, rat, and human islets all contain abundant cytosolic isocitrate dehydrogenase (13,21). Insulin levels and insulin release are normal in the islets of the mouse strain Mod-1, which lacks cytosolic malic enzyme in all tissues (11,13). It is interesting that a recent systems biology paper modeling beta cell metabolism predicted that deletion of the cytosolic malic enzyme would not affect ATP levels or oscillations of cytosolic metabolites and would only modestly affect pyruvate and malate levels (23).
We have stably integrated shRNA plasmids into the INS-1 832/13 cell line to achieve long term reduction of ME1 enzyme activity and also of ME2 protein and enzyme activity as detected by immunoblotting and a newly developed enzyme assay. We find no evidence for altered insulin release in our stable lines and no evidence for compensatory increases in other enzymes that produce cytosolic NADPH.

Cell Culture and Transfection
The INS1 832/13 rat insulinoma cell line (24) was a gift of Dr. Chris Newgard (Duke University Medical Center) at population doubling 42, and was grown an additional 10 -14 passages prior to transfection. Cells were cultured in RPMI 1640 containing 10% fetal bovine serum, supplemented with 1 mM pyruvate, 10 mM HEPES, 2 g/liter sodium bicarbonate, 50 M 2-mercaptoethanol, 100 IU/ml penicillin, and 100 g/ml streptomycin (25). Cells were transfected at ϳ60 -80% confluence in individual wells of a 6-well plate, using 10 l of Lipofectamine TM 2000 (Invitrogen) with 4 g of plasmid DNA, according to the manufacturer's directions, in complete medium lacking antibiotics, for 5 h or overnight. Cells were trypsinized and transferred to one or two 75-cm 2 culture flasks after 24 h and allowed to grow an additional day prior to selection with hygromycin B (100 g/ml). Transferent lines were maintained in selective medium. Preliminary studies were performed with siRNA prepared by in vitro transcription using the Silencer TM siRNA construction kit (Ambion). In these experiments, siRNA was added to the cells at 40 nM with Oligofectamine TM (Invitrogen) according to the manufacturer's directions.

Insulin Release
Insulin release experiments were performed as previously described (26). Total insulin was assayed using a modification of published methods (24). Cells were lysed in 0.5 ml of water. An equal volume of 2 M acetic acid plus 0.2% bovine serum albumin was added, and the mixture was briefly sonicated. Aliquots were diluted and assayed for insulin. Replicate wells were washed in Krebs-Ringer buffer to remove protein and lysed by pipetting in 0.5 ml of water, and aliquots were taken for the determination of total protein. All conditions were performed in quadruplicate in each experiment.

Tol2 Vector Construction
pCMV-Tol2 and pminiTol2-MCS (27), containing the transposase and the transposable element of Oryzias latipes 2 (Tol2) (28), respectively, were gifts of Dr. Stephen C. Ekker. The pSilencer 2.1 U6 hygromycin vectors (Ambion) containing the malic enzyme (Me1-753) or control (HygC) targeting inserts were digested with PscI and PdmI. The DNA fragment containing the shRNA cassette and the selectable marker was ligated between the NcoI site and the blunted HindIII site of the pminiTol2-MCS vector to produce the Me1-753ϩTol2 and hygromycin control ϩ Tol2 (HygCϩTol2) vectors. The Me2-654 and Me2-2124 constructs also contain the Tol2 transposable element. The Tol2 shRNA vectors were cotransfected with an equimolar amount of the transposase expression plasmid, pCMV-Tol2, using Lipofectamine 2000 (Invitrogen), as above. Cells were selected with hygromycin B, as above.

Enzyme Assays
Cytosolic Enzymes-Cells from one 75-cm 2 flask were trypsinized, resuspended in tissue culture medium, and washed twice in phosphate-buffered saline. The cell pellet was lysed by a single freeze/thaw cycle and pipetting in 150 l of buffer containing 220 mM mannitol, 70 mM sucrose, 5 mM HEPES, and 1 mM dithiothreitol, adjusted to pH 7.5 with KOH (KMSH buffer). The homogenate was centrifuged at 20,000 ϫ g for 10 min at 4°C, and the supernatant was collected for assay of cytosolic enzymes to give a protein concentration of 3-6 mg/ml cytosol. Enzymes were assayed in a buffer consisting of 50 mM Tris chloride, pH 7.6, 6 mM MgCl 2 , 0.5 mM NADP ϩ , and 1 mM dithiothreitol. Protein concentration was measured using the Bio-Rad protein assay. For ME1 activity, 10 l of cytosol was added to 1 ml of prewarmed buffer, and the absorbance was monitored at 340 nM on a SpectraMax M2 spectrophotometer (Molecular Devices). The slope was determined in the presence and absence of 1 mM malate and analyzed with the kinetic protocol of the SoftMax Pro software (Molecular Devices). Isocitrate dehydrogenase activity was measured similarly, except that 20 l of cytosol was measured with and without 1 mM isocitrate. Glucose-6-phosphate dehydrogenase activity was assayed similarly, using 10 l of cytosol with or without glucose 6-phosphate at a final concentration of 2.9 mM. Activities for all enzymes were calculated in nmol of NADPH produced/min/mg of protein.
Mitochondrial Malic Enzyme (ME2)-Cells were harvested by trypsinization from 3-4 75-cm 2 flasks, resuspended in culture medium, and washed twice in phosphate-buffered saline. The cell pellet was resuspended in 1 ml of KMSH buffer (see above) containing a protease inhibitor mixture (Halt Protease Inhibitor Cocktail, EDTA-free (Thermo-Scientific Pierce)) and homogenized with 40 strokes of a Potter-Elvehjem homogenizer. The homogenate was centrifuged at 600 ϫ g for 10 min at 4°C, and the supernatant fraction was placed in a new tube and centrifuged a second time at 600 ϫ g. The second supernatant fraction was carefully removed and centrifuged at 10,000 ϫ g for 10 min. The 10,000 ϫ g supernatant fraction ("cytosol") was removed, and the pellet ("mitochondria") was washed in 0.5 ml of homogenization buffer, and the 10,000 ϫ g centrifugation was repeated. The supernatant was discarded, and the mitochondrial pellet was resuspended in 150 l of homogenization buffer and vortexed vigorously. The mitochondrial fraction was frozen at Ϫ20°C prior to assay. Mitochondrial malic enzyme (ME2) activity was measured by an assay recently developed in our laboratory (29). The enzyme reaction mixtures contained 50 mM Tris-HCl, pH 7.4, 10 mM MgCl 2 , 0.3 mM NADP ϩ , and 1 mM dithiothreitol (background control mixture) or the same mixture with either the addition of 1 mM malate or with the addition of both 10 mM malate and 5 mM fumarate (an allosteric activator of ME2). Rates were measured simultaneously in the three mixtures at 37°C in a 96-well plate, with 200 l of the reaction mixture and 4 l of mitochondrial preparation per well. Measurements were performed at ϳ22-s intervals for at least 10 min with mixing prior to each measurement. The initial 5 min was considered as equilibration time, and the rate remained constant for at least 10 min after this period. The absorbance was read at 340 nM at 37°C using the SpectraMax M2 spectrophotometer (Molecular Devices). The background rate was subtracted from the rates in both the 1 mM malate and 10 mM malate plus 5 mM fumarate. An enzyme rate in the presence of 1 mMmalatehigherthanthebackgroundratecouldbeduetoNADPdependent mitochondrial malic enzyme (ME3), to ME1 from residual cytosolic contamination of the mitochondria, and/or to partial activity of mitochondrial ME2 malic enzyme. (see "Results" and Fig. 3). ME2 activity should be maximal in the presence of 5 mM fumarate plus 10 mM malate. As a mitochondrial control, the same preparations were assayed for the mitochondrial glycerol phosphate dehydrogenase as described previously (30).

RNA Quantification
Total RNA was prepared using the RNeasy minikit (Qiagen), with on-column deoxyribonuclease digestion (RNase-free DNase set, Qiagen). cDNA was prepared from 1.5 g of total RNA using the RETROscript kit (Ambion/Applied Biosystems) with oligo(dT) primers. Real-time PCR was performed on a Bio-Rad MyiQ TM single color real-time PCR detection system. A standard curve prepared from INS-1 832/13 cDNA was included on each run for relative quantitation. Glutamate dehydrogenase was used in some experiments as an internal control (see "Results") after testing showed it to have the least variation of the potential control genes tested (data not shown). Primers used for real time quantification are shown in supplemental Table 2. Interferon pathway pilot studies were performed semiquantitatively, by amplification of serially diluted cDNAs for 26 cycles and examination of the products on 2% agarose gels. Oas1a and Irf7 mRNA levels were then examined by real-time PCR as above. Interferon pathway primers are shown in supplemental Table 3.

Western Blotting
15 g of mitochondrial protein were loaded per lane and separated by SDS-PAGE. After transfer to nitrocellulose, the membrane was blocked with buffer containing 10 mM Tris buffer, pH 8.0, 150 mM NaCl, and 5% nonfat powdered dry milk. The polyclonal rabbit anti-ME2 antibody (HPA008880, Sigma) was used at 1:500, and was detected with a horseradish peroxidase-conjugated goat anti-rabbit IgG (Thermo Scientific), used at 1:13,000. The signal was detected using the Immobilon Western chemiluminescent horseradish peroxidase developer (Millipore). As a mitochondrial control, the blot was stripped with Restore Western blot stripping buffer (Thermo Scientific) prior to reprobing with a polyclonal antiserum to the mitochondrial GPD2 (glycerol-3-phosphate dehydrogenase) (31) at 1:10,000, with detection as above.

Statistics
Results are shown as means Ϯ S.E. Statistical analysis was performed using Student's t test with two-sided comparisons.

RESULTS
Interferon Pathway Induction-In our preliminary experiments using cells transfected with Oligofectamine TM and 40 nM siRNA prepared by in vitro transcription, we found induction of four genes in the interferon pathway. The mRNAs of Oas1a (2Ј,5Ј-oligoadenylate synthetase 1A) and Irf7 (interferon regulatory factor 7) were induced ϳ16-fold, mRNA of Prkr (protein kinase, interferon-inducible double-stranded RNA-dependent) was induced 4 -16-fold, and mRNA of Mx1 (myxovirus resistance 1) was strongly induced by siRNA but was undetectable in control cells. Three other genes tested (Irf1, Irf2, and Wars) were not induced. Induction of the interferon pathway by siRNAs transcribed with T7 RNA polymerase has been reported to be due to the presence of a 5Ј-triphosphate on the siRNA (32). Using these findings as a guide, we tested our cell lines expressing stable shRNA constructs for similar induction. No induction of the interferon pathway was seen in cells stably expressing shRNA constructs when tested for Prkr (seven transferent lines tested), Irf7 (15 lines), and Oas1a (17 transferent lines tested, including HygC, Me1-753, Me1-1575, Me2-654, Me2-2124, and the Tol2 versions of these constructs). In addition, total RNA quantity from transferent cell lines was similar to that of non-targeted 832/13 cells. Growth rates were similar in the parental 832/13 cell line, the control shRNA line HygC, and the Me1 and Me2 shRNA lines (data not shown). Fig. 1 shows our results using the pSilencer 2.1 U6 hygromycin plasmids. Malic enzyme mRNA levels were reduced by 80 Ϯ 5% in the Me1-753 line and by 84 Ϯ 2% in the Me1-1575 line when compared with the non-targeting shRNA (HygC) control (p Ͻ 0.02 for each). No difference in Me1 mRNA levels was seen between the non-targeting shRNA control and the INS-1 832/13 parental cells. No significant changes in levels of mRNAs or enzyme activities were found between the lines in levels of glucose-6-phosphate dehydrogenase or the cytosolic isocitrate dehydrogenase, the two other sources of cytosolic NADPH (Fig. 1A). Malic enzyme (ME1) activity was reduced 86 Ϯ 8% in Me1-753, and 81 Ϯ 9% in Me1-1575 (p Ͻ 0.02) and was not changed by the HygC control shRNA. No compensatory changes were seen in the activity of glucose-6phosphate dehydrogenase or the cytosolic isocitrate dehydrogenase (Fig. 1B). Insulin release stimulated by glucose was lower in the line expressing the non-targeting insert (HygC) than in either the parental 832/13 line or the Me1-753 line (p Ͻ 0.05 for both comparisons), but glutamine plus BCH-stimulated insulin release did not differ significantly between either control and the two Me1-targeted lines (Fig.  1C). Total insulin content of the HygC line was found to be 74% of that of the 832/13 control (p Ͻ 0.01) and 78% of that of Me1-753 (p Ͻ 0.01). No significant differences were found among any of the cell lines when insulin release was expressed as -fold glucose-or glutamine plus BCH-stimulated insulin secretion over base-line (unstimulated) insulin secretion (Fig. 1D).

No Inhibition of Insulin Release with Cytosolic Malic Enzyme (Me1) Knockdown-
Because genetic drift can occur on prolonged culture and because selection and amplification of the cells required approximately 1 month between transfection and the start of testing, we attempted to shorten the time to testing and increase the number of stable transferents through the use of the Tol2 transposase system (27). By putting Tol2 transposition sites into our selectable vector, the plasmid DNA sequence that is integrated should be uniform, although the sites of integration in the genome should be essentially random (27). The rate of stable integration was improved, as evidenced by an estimated 20 -100fold increase in hygromycin-resistant clones and approximately onethird shortening of time to testing (from 30 -34 days to ϳ22 days). Fig.  2A shows that Me1 mRNA was decreased 86 Ϯ 2% in the Me1-753ϩTol2 line when compared with the HygCϩTol2 control (p Ͻ 0.002) and that ME1 enzyme activity was decreased 78 Ϯ 2% (p Ͻ 0.001) (Fig. 2B). No compensatory changes were seen in either the mRNA level or enzyme activity of isocitrate dehydrogenase or glucose-6-phosphate dehydrogenase. The rat NAD(P) ϩ -Me2 and the NADP ϩdependent Me3 mRNA levels were also tested in the Tol2 transferent lines. Although a small difference in Me2 mRNA was seen between the 832/13 parental line and the Me1-753ϩTol2 line (1.36-fold, p Ͻ 0.05), the Me2 mRNA level of the Me1-753ϩTol2 line did not differ significantly from that of the shRNA control vector HygCϩTol2. We also directly examined the internal control, glutamate dehydrogenase, for evidence of an effect of the Me1 knockdown. RNA from each of the transferent or control lines was prepared on three different dates, and 1.5 g of total RNA from each of the nine samples were reverse transcribed on the same day. The glutamate dehydrogenase mRNA values, relative to an INS-1 832/13 standard, were as follows: 832/13, 1.05 Ϯ .02; HygC ϩ Tol2, 1.02 Ϯ 0.07; Me1-753 ϩ Tol2, 1.03 Ϯ 0.06. This showed that the internal control was not affected by the transfection or selection. No significant difference was found in insulin release between any of the lines, whether expressed as units released per mg of protein or as -fold stimulation (Fig. 2, C and D). Fig. 3A shows that Me2 mRNA is decreased by 95% with Me2-654 shRNA and by 82% with Me2-2124 (p Ͻ 0.01 for both). The Me2-224 and Me2-269 shRNA constructs did not significantly alter Me2 mRNA levels in transfected cells (data not shown). Levels of the mRNAs encoding glutamate dehydrogenase, cytosolic malic enzyme, and the mitochondrial NADP-malic enzyme (Me3) mRNA did not differ between Me2-654 and Me2-2124 and the control shRNA-treated cells (HygC). NAD(P)-malic enzyme (ME2) protein was not detected in Western blots of mitochondria from the Me2-654 line, and a faint band was detected in the Me2-2124 cell line only intermittently, whereas normal ME2 protein bands were seen in the two unsuccessfully targeted cell lines, Me2-224 and Me2-269 (Fig. 3B) (data not shown). An assay for ME2 enzyme activity corroborated the large decrease in ME2 protein shown by the Western blot. In this assay, we attempted to discern whether malic enzyme activity detected in mitochondria was due to ME2 or to either contamination of the mitochondrial preparation by cytosolic malic enzyme (ME1) or activity of the mitochondrial NADP-malic enzyme, ME3. Malic enzyme activity was measured both in the presence of a low concentration of malate (1 mM) that would detect maximal activity of ME1 and nearly maximal activity of ME3 and in the presence of 10 mM malate plus 5 mM fumarate, an allosteric activator of ME2, to maximize ME2 activity (Fig. 3D). In mitochondria from the Me2-654 and Me2-2124 cell lines, in which Me2 mRNA was successfully lowered, the enzyme rates in the presence of 10 mM malate plus fumarate were severely lowered (p Ͻ 0.01 for each), compared with the shRNA control cell line (HygC) and the parental INS-1 832/13 cell line. In the Me2-654 line, the malic enzyme rate obtained in the presence of 1 mM malate was nearly the same as the rate in the presence of 10 mM malate, suggesting that most of the residual malic enzyme activity was due to either ME3 or ME1, whereas the ratio of activity in 1 mM malate to 10 mM malate was 21.7% in 832/13 and 17.4% in the HygC line. The rate with 1 mM malate in the INS-1 832/13 cells differed from that of both targeted lines (p Ͻ 0.04), although the rate in the HygC control did not differ significantly from the other lines. This suggests that, although a small amount of the reduction in malic enzyme activity in the presence of 1 mM malate was due to the decreased levels of ME2, much of the malic enzyme activity detected with 1 mM malate was due to ME3 or ME1. If the rate in the presence of 1 mM malate in the targeted lines is assumed to be due primarily to ME1 and ME3, and if this rate is also subtracted from the activity of the control lines, calculated reductions of ME2 activity in the targeted lines, compared with the HygC control, are 91% in Me2-654 and 67% in Me2-2124, which agree fairly closely with the reductions in Me2 mRNA in these cell lines. ME2 enzyme activity in the unsuccessfully targeted cell lines Me2-224 and Me2-269 was similar to that in the control cells (data not shown). No significant differences were found among the cell lines in respect to the enzyme activity of the internal mitochondrial control, mitochondrial glycerol phosphate dehydrogenase, and no compensatory changes were seen in the activity of the cytosolic malic enzyme (Fig. 3C). Insulin content varied among the cell lines, resulting in a higher absolute insulin release in the HygC line compared with the parent cell line INS-1 832/13 (Fig. 4A); however, insulin release was unchanged by ME2 loss when measured as -fold stimulation of insulin by  DECEMBER 18, 2009 • VOLUME 284 • NUMBER 51

JOURNAL OF BIOLOGICAL CHEMISTRY 35363
either glucose or glutamine plus BCH (Fig. 4B). The Me2-654 line had a lower insulin content than the other lines, but when insulin secretion was examined as -fold stimulation or as a fraction of the total insulin content, secretion was normal (Fig. 4, B  and C). In summary, we could not confirm a significant impairment of insulin release with the long term loss of either ME1 or ME2 enzyme activity induced by stable expression of shRNA.

ME1 and Insulin
Release-Cytosolic NADPH is produced by the cytosolic NADP ϩ -dependent malic enzyme (ME1), cytosolic NADP ϩ -isocitrate dehydrogenase, and glucose-6-phosphate dehydrogenase. Short term siRNA-mediated inhibition of ME1 activity by 54 -58% in the rat insulinoma cell line INS-1 832/13 was reported to decrease insulin secretion by 40% with either glucose or glutamine plus leucine stimulation (10). This result was recently confirmed in the same cell line in a report that short term shRNA expression reduced ME1 enzyme activity by 60% and glucosestimulated insulin secretion by 30%, although there was no change in glucose oxidation (7). In contrast, we have found that a chronic reduction of ME1 enzyme activity by greater than 80% does not interfere with insulin release in the rat insulinoma line INS-1 832/13 (Fig. 1). Although chronic reduction of enzyme activity may differ from that of acute reduction due to compensatory changes, we found no evidence of alterations in the activity of the two other enzymes that could also provide cytosolic NADPH in these cells. This finding suggests that if cytosolic NADPH production is important for insulin release, either residual malic enzyme activity is sufficient, or other sources, such as the cytosolic isocitrate dehydrogenase, provide sufficient NADPH without compensatory changes. Similar redundancy could exist for pyruvate cycling between the pyruvate-malate, pyruvate-citrate, and pyruvate-isocitrate pathways. The evidence suggesting a role for the pyruvate-isocitrate pathway in glucose-stimulated insulin release and against the role of pyruvate cycling via the pyruvatemalate or pyruvate-citrate pathways has recently been reviewed (33). The pyruvate-isocitrate pathway also produces cytosolic ␣-ketoglutarate, which could enter the malateaspartate NADH shuttle, a pathway previously shown to be important for glucose-induced insulin secretion (34,35).
It should also be noted that negative results suggestive of compensatory changes or redundancy of pathways are not universally found with long term inhibition by shRNA. In a preliminary test of this system, we used an shRNA to reduce the mRNA of the mitochondrial malate dehydrogenase, a component of the tricarboxylic acid cycle, by 75%, with inhibition of insulin release to glucose, pyruvate, and amino acids by 40 -45% when compared with parental INS-1 832/13 cells. 3 We have previously published reports of inhibition of insulin release by long term shRNA-mediated reduction of pyruvate carboxylase (36) and acetoacetyl-CoA synthetase (9). The blot was stripped and reprobed for the GPD2 protein as an internal control for loading. C, GPD2 activity was assayed in the same mitochondrial preparations used in D to ascertain the equivalence of the mitochondrial preparations. ME1 activity was measured in cytosol to look for evidence of compensatory changes. Enzyme activities are expressed relative to the HygC control. These activities did not differ significantly between the lines. D, ME2 activity was assayed in mitochondria. Activity of each preparation was measured in three enzyme reaction mixtures: without substrate to determine the background rate; in the presence of 1 mM malate (which should measure ME3 activity and any contaminating ME1 activity and may include partial activity of ME2); and in the presence of 10 mM malate plus 5 mM fumarate (an allosteric activator of ME2) to determine ME2 maximal activity plus any ME1 and/or ME3 maximal activities if present (see "Experimental Procedures" and "Results"). Background activity with no malate was subtracted to yield malic enzyme rates. Total ME activity in 10 mM malate plus 5 mM fumarate in the 832/13 line was 20. Mitochondrial Malic Enzymes-In addition to the cytosolic malic enzyme, two malic enzymes have been described in mammalian mitochondria. The mitochondrial NAD(P)-malic enzyme, ME2, can use either NAD ϩ or NADP ϩ to catalyze the oxidative decarboxylation of malate to pyruvate. This enzyme is often present in tumor cells (37,38), and the level of this enzyme has been linked to tumor progression (39). It is also present in normal cells that actively divide, such as small intestinal epithelium, spleen, thymus, and testis (40). This enzyme is inhibited by ATP in the physiologic range and is activated by fumarate, which lowers the K m for NAD ϩ , NADP ϩ , and malate (38). Because of its tissue distribution and activity, it is thought that ME2 allows glutamine and glutamate to be used as a source of respiratory energy when acetyl-CoA levels are low (41). Me2 mRNA is found at similar levels in INS-1 832/13 cells and normal rat islets (10) (data not shown). The presence of ME2 could provide an explanation for the ability of glucose or glutamine plus BCH to strongly stimulate insulin release in cells lacking ME1. Conversely, the presence of ME1 may explain the normal glucose-induced and glutamine plus BCH-induced insulin release in our cell lines with severely knocked down ME2 protein and enzyme activity. Thus, our data suggest that ME1 and ME2 may each be present at levels sufficient for insulin release in the absence of the other or that neither enzyme is required for insulin release in the INS-1 832/13 cell line. In regard to the redundancy of malic enzymes, because malate and many of its metabolites are rapidly transported across the inner mitochondrial membrane, in cells deficient in ME1, such as normal mouse beta cells (21,22), or all tissues of a Mod-1 mouse (11,13), ME2 could supply metabolites that can be exported from mitochondria to supply all cytosolic pathways except for the generation of NADPH equivalents and pyruvate cycling by ME1. Conversely, malate formed in the cytosol by ME1 and its metabolites can be rapidly transported into mitochondria in cells deficient in ME2 to supply mitochondrial pathways. The mRNA encoding the mitochondrial NADP ϩ -malic enzyme (ME3) appears to be present in INS-1 832/13 cells and islets but at lower levels than the mRNAs that encode ME2 or ME1 (10, 42), 4 and ME3 protein has not yet been definitively demonstrated in these cells. The malic enzyme activity that we measured in the presence of 1 mM malate is ϳ20% of the total mitochondrial malic enzyme activity, consistent with the published report that Me2 RNA is present at 5 times the level of Me3 RNA (10). The RNA and enzyme activity data together suggest that if ME3 is present in INS-1 832/13 cells, its level is substantially lower than that of ME1 or ME2.
Specificity of RNA Interference Effects-In addition to this report, other papers have recently been published that reported conflicting results when using siRNA or shRNA to study insulin release. One of the papers that reported decreased insulin secretion following Me1 shRNA treatment also reported a decrease in glucose-stimulated insulin secretion by an shRNA for ATP-citrate lyase (7). However, two other papers reported shRNA-or siRNA-mediated reduction of ATP-citrate lyase to a greater extent (8,9) but found no inhibition of glucose-stimulated insulin secretion. Conflicting results could arise from methodological differences, including the siRNA concentration used, or differences between nonspecific or sequence-specific off-target effects of the reagents used in these experiments. Predicting and detecting all of these effects remains difficult. Nonspecific off-target effects include the induction of the interferon response, with resultant inhibition of protein synthesis and RNA degradation (43)(44)(45). The interferon response does not require long double-stranded RNAs and has been reported with lentiviral vectors producing 21-mer shRNAs (46). Although we directly tested for and found induction of a subset of genes in the interferon pathway in our cells that were treated with siRNA, we ruled out induction of these genes in our lines transfected with shRNA plasmids. Such testing was not reported in the papers that demonstrated an effect of Me1 or Me2 knockdown. Nonspecific off-target effects, with altered expression of many genes, have been reported with the lipid transfection reagent alone and may be enhanced by siRNA (47). Both shRNAs and siRNAs can alter gene expression nonspecifically through competition with endogenous microRNAs (miRNAs) for limiting components in the miRNA processing or gene silencing pathways (48 -50). Because miRNAs have been reported to affect islet development (51)(52)(53), insulin production (53,54), and insulin secretion (55,56), interference with miRNA function could increase or decrease insulin secretion. Specific (sequence-dependent) off-target effects are also apparently frequent with either siRNA or shRNA, with dozens to hundreds of genes partially silenced or induced for each sequence introduced (44,(57)(58)(59), and only a 6 -7-nucleotide "seed region" match (positions 2-7 or 2-8) is required for these effects (60,61).
The use of stable transferents has some advantages. Because the testing of the transferent lines is removed from the time of transfection (often by a month or more), the cells could recover from any transient, transfection-related effects. Sequences that cause an interferon response as an siRNA in lipid transfection may not produce this response when expressed from a polymerase III promoter as an shRNA (62). Stable transferent lines can be tested repeatedly, and large numbers can be grown for additional experiments (proteins, metabolites, etc.). Potential problems include the possibility of selection of a population of cells that has diverged from the original, especially if there are few surviving transferent cells or if the knockdown creates a growth disadvantage. Off-target effects of the expressed shRNA (due to seed region homology or interference with endogenous miRNA processing) can still occur. Although both siRNA and shRNA are valuable tools, caution must be used in the design and interpretation of knockdown experiments, and testing of a variety of techniques is warranted.