Epigenetic Changes in Mitochondrial Superoxide Dismutase in the Retina and the Development of Diabetic Retinopathy

Wistar rats (male, 200 g) were randomly assigned to normal, diabetic, or experimentally galactosemic groups. Diabetes was induced with streptozotocin (55 mg/kg body wt), and 3 to 4 days after induction of diabetes, the rats were divided into three groups: group 1 rats were in poor glycemic control (glycated hemoglobin [GHb] >12%) for 4 months (PC group), group 2 rats were in poor control for 2 months, followed by good control for 2 additional months (PC-Rev group), and group 3 rats were in good control (GC group, GHb ∼7.0%). The PC rats received 1 to 2 units of insulin (Humulin; Eli Lilly, Indianapolis, IN) four to five times a week, and the GC rats received insulin twice daily for a total of 7 to 8 units. Experimental galactosemia was induced in normal rats (Galac group) by feeding 30% galactose supplemented diet. The Galac group rats received the galactose diet for the entire 4 months of the study, and the Galac-Rev group received galactose diet for the first 2 months and a normal diet for 2 additional months. The control group comprised normal age-matched rats.

Blood was obtained by weekly tail vein puncture to measure glucose by Freestyle Glucometer (Bayer, Tarrytown, NY), and GHb (every 2 months) using a kit from Helena Laboratories (Beaumont, TX). These procedures conformed to the Association for Research in Vision and Ophthalmology Resolution on Treatment of Animals in Research and are routinely used in our laboratory (7,12,19–22). At the end of the experiment, rats were killed by carbon dioxide asphyxiation, and the retina was removed immediately. About 50% of the retina was crosslinked with 1% paraformaldehyde, and the remaining was stored in liquid nitrogen.

Endothelial cells prepared from bovine retina were cultured in medium consisting of 15% FBS (heat inactivated), 5% replacement serum (Nu-serum; BD Bioscience, San Jose, CA), heparin (50 μg/mL), endothelial growth supplement (25 μg/mL), and antibiotic/antimycotic in Dulbecco’s modified Eagle’s medium (9,23). Confluent cells from passages four to five were incubated in high glucose (20 mmol/L) for 4 days, followed by normal glucose (5 mmol/L) for 4 additional days (20-5 group). The controls included cells incubated in normal or high glucose for 8 days of the experiment, and 20 mmol/L mannitol served as an osmotic control. At the end of the incubation, the cells were processed for chromatin immunoprecipitation, as described below.

Endothelial cells from the passages three to five were transfected with small interfering RNA (siRNA) of SUV420h2 or NF-κB p65 using transfection reagents from Santa Cruz Biotechnology (Santa Cruz, CA) as routinely performed in our laboratory (26). Briefly, after incubation of siRNA transfection complex for 45 min at room temperature, the cells were incubated with the transfection complex for 8 h at 37°C. Parallel incubations were done using nontargeting scrambled RNA. For sod2 overexpression, the cells were incubated with 3 μg of sod2 plasmid and effectene transfection reagent (Qiagen, Valencia, CA), as described by us (23). At the end of the transfection, the cells were washed with PBS and incubated in 5 mmol/L or 20 mmol/L glucose for 4 days. The transfection efficiency was verified by semiquantitative PCR using control primers from Santa Cruz Biotechnology.

Rat retina or isolated endothelial cells were crosslinked with 1% paraformaldehyde for 15 min. The fixed sample was resuspended in lysis buffer containing 1% SDS, 10 mmol/L EDTA, 50 mmol/L Tris-HCl, pH 8.1 (ChIP Assay Kit, Millipore, Temecula, CA), and protease inhibitor cocktail (Sigma-Aldrich, St. Louis, MO), and sonicated four to six times for 10 s each. Supernatant was diluted in ChIP dilution buffer containing 0.01% SDS, 1.1% Triton X-100, 1.2 mmol/L EDTA, 16.7 mmol/L Tris-HCl, pH 8.1, and 167 mmol/L NaCl, and precleared with protein A agarose/salmon sperm DNA for 30 min.

After a small aliquot of protein-DNA complex (15–20 μg) was collected for starting chromatin input, the protein-DNA complex was immunoprecipitated with the antibody against NF-κB p65 or p50, acetyl H3K9, H4K20me3 or rabbit normal IgG (Abcam, Cambridge, MA), or anti-general transcription factor IIB (TFIIB, Millipore Corporation, Temecula, CA). The immunoprecipitate collected with protein A agarose/salmon sperm DNA was washed sequentially with low-salt buffer, high-salt buffer, and LiCl buffer, according to the vendor’s protocol. After the immunoprecipitate was washed twice with Tris-EDTA buffer, the sample was used for Western blotting or was extracted twice with 1% SDS containing 0.1 mol/L NaHCO₃.

Elutes and input were heated at 65°C for 4 h to reverse the formaldehyde crosslinking and digested with protease K at 45°C for 1 h. DNA fragments were recovered by phenol-chloroform-isoamyl alcohol extraction and ethanol precipitation, and resuspended in 20 μL water for PCR. Gene expression of sod2 promoter and enhancer were quantified by SYBR green-based real-time quantitative PCR (q-PCR). Promoter of β-actin occupied by TFIIB and the off-target region of sod2 were used as ChIP positive and negative controls, respectively, and normal rabbit IgG was used as the negative antibody control and DNA from input as the internal control. Each ChIP measurement was made in five to six rats per group or four to five cell preparations.

Gene expressions were confirmed by q-PCR using 7500 Real Time PCR System (Applied Biosystem, Foster City, CA) with SYBR green Master Mix (Applied Biosystem, Branchburg, NJ) or TaqMan primers.

Amplification of the sod2 promoter and enhancer was performed in 1 μL of ChIP-purified DNA using the primers provided in Table 1. PCR was performed by denaturing at 95°C for 10 min, followed by 40 cycles of denaturation at 95°C for 15 s, annealing and extension at 60°C for 60 s, and this was followed by 95°C for 15 s, 60°C for 60 s, 95°C for 15 s, and 60°C for 15 s. The q-PCR values in each immunoprecipitate were normalized to the Ct value from the input sample using the ΔΔCt method. Relative fold-changes were calculated by setting the mean fraction of normal rats as 1. For quantifying the gene expressions of SUV420h2 and sod2, TaqMan primers were used. Each sample was measured in triplicate. Fold-change in mRNA abundance was calculated with the ΔΔCt method as routinely used by us (22,23).

Chromatin pellets or retinal homogenate were boiled for 5 to 10 min in SDS-PAGE sample buffer, and separated on a 4–20% gradient polyacrylamide gel. The proteins blotted onto membranes were blocked and incubated with the primary antibody against acetyl H3K9, or mono-, di-, or trimethyl H4K20 (H4K20me1, me2, and me3, respectively), or SUV420h2 (Abcam). After enhancing the target proteins with enhanced chemiluminescence reagent (Thermo Scientific, Rockford, IL), they were determined by autoradiography. The membranes were reprobed with normal rabbit IgG or β-actin (Sigma-Aldrich, St. Louis, MO) or histone H4 (Cell Signaling, Danvers, MA). The band intensity was quantified using Un-Scan-It Gel digitizing software (Silk Scientific, Orem, UT), and protein expression was calculated relative to normal rabbit IgG in ChIP or β-actin or histone H4 in the same samples.

For immunoprecipitation, 100–200 μg of retinal protein was incubated overnight with 1 μg antibody against NF-κB p65 or H4K20me3. The antibody-protein complex was collected with Protein A/G Plus-agarose (Santa Cruz Biotechnology) and separated on a 4–20% gradient gel. The membranes were probed for H4K20me3 or acetyl H3K9, and NF-κB p65 or H4K20me3 served as loading controls for their respective membranes. The precipitate obtained from normal rabbit IgG (Abcam) served as the negative control.

Results are presented as mean ± SD, and analyzed using the nonparametric Kruskal-Wallis test, followed by the Mann-Whitney test for multiple group comparison. Similar conclusions were achieved by using ANOVA with Fisher or Tukey tests. P < 0.05 is considered as significant.

The body weights of the rats in the PC group were significantly lower, and their GHb values were more than twofold higher compared with the age-matched normal control rats (317 ± 32 vs. 439 ± 68 g and 12.3 ± 2.0 vs. 5.9 ± 0.9%, respectively). The rats in the GC group had body weight and GHb values (445 ± 14 g and 7.0 ± 0.4%, respectively) that were similar to the normal rats (P > 0.05). In PC-Rev group, the body weights and GHb levels before initiation of good glycemic control were not different from the rats in the PC group (272 ± 30 g and 11.8 ± 2.0%, respectively). However, after initiation of good control, the values became similar to those in the normal group (395 ± 35 g and 6.9 ± 0.3%). The GHb value in the Galac group was slightly higher than in the normal group (∼8%), but was significantly lower than in the PC group (P < 0.05). The average body weights of the Galac group were ∼400 g, and became similar to the normal rats after termination of the galactose diet (447 ± 21 g, GHb 6.1 ± 0.3%).

Diabetes inhibited the activity of retinal MnSOD (1) and also decreased its gene transcripts by 60% (Fig. 1). H4K20me3 and acetyl H3K9 were elevated by 6–11-fold at both the promoter and the enhancer of sod2 (Fig. 2A and B). Because H4K20me3 was increased at sod2, the protein levels of H4K20me1, -me2, and -me3 were also quantified. As shown in Fig. 2C, methylation of H4K20 was increased in diabetes, and the increase in H4K20me2 and -me3 was significantly higher than that in H4K20me1. Gene and protein expressions of SUV420h2, the enzyme with dynamic functions for di- and trimethylation (16,27), were significantly elevated (Fig. 2D and E). In the same retina samples, although NF-κB p65 was elevated by more than 10-fold at both the promoter and enhancer of sod2 (Fig. 3A), p50 was elevated only at its enhancer (Fig. 3B).

To verify the specificity of the proteins binding to the promoter and enhancer of sod2, the same ChIP DNA from normal rats was amplified using an off-target primer (amplified from +175 to +488 of sod2) by semiquantitative PCR (Fig. 3C) and confirmed by q-PCR (Fig. 3D). As shown in Fig. 3C and D, despite substantially enriched H4K20me3, acetyl H3K9, and NF-κB p65 at the promoter of sod2, off-target region existing in input and genomic DNA was not occupied by them and nonspecific precipitate by normal IgG was almost undetectable. H4K20me3 and acetyl H3K9 proteins, co-immunoprecipitated with p65 in chromatin, were significantly increased (Fig. 4A).

To verify interaction of H4K20me3 or acetyl H3K9 with NF-κB p65, retinal homogenate was immunoprecipitated with NF-κB p65 or H4K20me3 antibodies. Figure 4B shows that diabetes significantly increased the interaction of H4K20me3 or acetyl H3K9 with NF-κB p65. Owing to the increased expression of retinal NF-κB p65 in the PC and PC-Rev groups compared with normal (consistent with our previous results showing the activation of retinal NF-κB p65 in diabetes [12]), despite much higher interactions depicted in the Western blots, the accompanying histogram shows about a twofold increase in the ratio. However, retinal protein precipitated with rabbit IgG (negative control) yielded negligible expressions of H4K20me3 or acetyl H3K9 (Fig. 4C). In addition, as shown in Fig. 4D, in the same animals, diabetes also increased the interaction of H4K20me3 and acetyl H3K9.

As with diabetes, gene expression of retinal sod2 were significantly reduced in the retina obtained from experimentally galactosemic rats, and H4K20me3 and acetylated H3K9 at both the promoter and enhancer of sod2 were elevated (Fig. 2). This was accompanied by concomitant increases of p65 and p50 at the enhancer of sod2, and only p65 at the promoter of sod2 (P < 0.05 compared with normal, Fig. 3).

Reinstitution of good glycemic control for 2 months after 2 months of poor glycemic control (PC-Rev) failed to provide any benefit to retinal sod2 (Fig. 1). It had no significant effect on retinal H4K20me3 and acetyl H3K9 at the promoter and enhancer of sod2, and SUV420h2 expression remained elevated (Fig. 2). The diabetes-induced increase in p65 at the promoter and enhancer of sod2, and H4K20me3 and acetyl H3K9 proteins co-immunoprecipitated with p65 in chromatin, and also in retinal homogenate, remained significantly elevated (P > 0.05 compared with PC, Figs. 3 and 4). The levels of H4K20me1, -me2, and -me3, and the expression of SUV420h2 remained elevated (Fig. 2C–E). In concert, termination of experimental galactosemia (Galac-Rev) also failed to prevent elevations in H4K20me3, acetyl H3K9, and p65 at the promoter and enhancer of sod2 (Figs. 2 and 3).

In contrast, when good glycemic control was instituted immediately after induction of diabetes in rats (GC group), values for retinal H4K20me3, acetyl H3K9, and NF-κB p50 were significantly different from those obtained from the rats in PC or Galac groups. Although p65 at the enhancer and promoter of sod2 was slightly higher than observed in normal rats, the values remained significantly different from those obtained from rats in the PC or PC-Rev groups (P < 0.05, Figs. 2 and 3). The levels of methylated H4K20me2 and -me3, and SUV420h2 were not different from those obtained from normal rats (Fig. 2C–E). Similarly, the values for H4K20me3 and acetyl H3K9 proteins, co-immunoprecipitated with NF-κB p65 in chromatin, were also lower than those observed in the PC group (Fig. 4A).

High glucose (20 mmol/L) exposure of retinal endothelial cells significantly decreased the gene expression of sod2 (P < 0.05, Fig. 5A), increased the gene expression (Fig. 5B) and protein levels (Fig. 5C) of SUV420h2, and elevated H4K20me3 and acetyl H3K9 at the promoter and enhancer of sod2 compared with the values obtained from the cells incubated in normal glucose (Fig. 6A and B). These changes were not due to increased osmolarity experienced by the cells in a high glucose medium because the addition of 20 mmol/L mannitol instead of 20 mmol/L glucose failed to produce any effect on SUV420h2 gene expression and the enrichment of H4K20me3 and acetyl H3K9 at sod2 (Figs. 5B, 6A and B).

Figure 6B and C shows that glucose exposure increased acetyl H3K9 at the sod2 promoter, but the negative antibody control, IgG, produced almost negligible PCR products for the sod2 promoter, and the yield of ChIP IgG was less than 2% of the product from ChIP with the acetyl H3K9 antibody. In addition, the amplification of TFIIB (positive control) at the β-actin promoter was similar in low and high glucose conditions (Fig. 6C). The increased SUV420h2 gene expression and the enrichment of H4K20me3 and acetyl H3K9 at sod2 persisted even when 4 days of high glucose was followed by 4 days of normal glucose exposure (20-5 group).

Because hyperglycemia increases oxidative stress, and oxidative stress can result in epigenetic modifications, we investigated the effect of sod2 overexpression. Figure 6A shows that overexpression of sod2 successfully blocked the increase of H4K20me3 at the sod2 promoter and enhancer. To further evaluate the effects of H4K20me3 and NF-κB p65 on the expression of sod2, SUV420h2 or NF-κB p65 were inhibited by their respective siRNAs. Figure 5B and C shows that high glucose increased the gene and protein expressions of SUV420h2, and this was successfully prevented by siRNA-SUV420h2. Transfection with siRNA-SUV420h2 also prevented an increase in H4K20me3 at the sod2 enhancer and ameliorated a decrease in sod2 mRNA (Figs. 5 and 7). In addition, transfection of cells with siRNA–NF-κB, which inhibited glucose-induced increase in p65 (Fig. 5D), further downregulated sod2 gene expression (Fig. 5A).

The mitochondrial superoxide scavenging enzyme, MnSOD, has an important role in protecting mitochondria from oxidative damage. Overexpression of sod2 in diabetes protects retinal mitochondrial DNA damage (23) and prevents the development of retinopathy (3). The enzyme activity continues to be compromised after reinstitution of normal glycemic control in diabetic rats, suggesting it is important in the metabolic memory phenomenon (7). This study provides novel data showing diabetes-induced epigenetic regulation of retinal sod2 and increased H4K20me3, acetyl H3K9, and NF-κB p65 at its promoter/enhancer. Increased acetyl H3K9 and NF-κB p65 at the promoter/enhancer of sod2 interacts with H4K20me3, suggesting a loss of their activation function. Furthermore, reinstitution of good control for 2 months after 2 months of hyperglycemia has no beneficial effect on increased methylation of H4K20 and acetylation of H3K9 and p65 at the promoter/enhancer of sod2, suggesting the role of epigenetic modifications of sod2 in the metabolic memory phenomenon associated with the progression of diabetic retinopathy.

Histone modifications regulate gene expression, acetyl H3K9 is considered to open up the chromatin (15), and H4K20me3, a modification with more general function in chromatin management, compacts it and inhibits genes by blocking the access of other transcription regulators to the genes (28,29). In cancer cells, decreased acetyl H3K9 at the promoter and enhancer of sod2 is associated with decreased sod2 (30). In diabetes, despite decreased activity of retinal MnSOD (7) and reduced transcripts, we show that acetyl H3K9 and H4K20me3 are both increased at the promoter and enhancer of sod2. In support, multiple histone modifications can act concurrently to specify distinct chromatin structure, and modifications for activation and repression may bind to the same gene simultaneously, suggesting a bivalent domain. The coexistence of H3K4me3 (activator) and H3K27me3 (repressor), in addition to downregulating genes can also activate them, and the same lysine residue can be methylated to different degrees with different functional consequences (31). Furthermore, we cannot rule out the possibility that increased acetyl H3K9, despite decreased sod2 transcript, could be due to an interaction of H4K20me3 with acetyl H3K9 that abolishes the activation function of acetyl H3K9.

The transcription-initiation site of sod2 is preceded by a promoter with the enhancer elements located in the second intron containing multiple potential regulatory elements, including several SP1 sites, two NF-κB sites, and an antioxidant-response element (32). Homodimer and heterodimer units of NF-κB bind to DNA binding sites with different affinities within the target genes (14). NF-κB p65/p50 binds to the enhancer and is important for cytokine-induced sod2, and also, the binding of NF-κB p65/p50 to the sod2 promoter is associated with its activation (30). In contrast, p50/p50 represses sod2 by competing with p65/p50 to the enhancer of sod2. Sp1 at the promoter and NF-κB at the enhancer are integrated by a connection protein, nucleophosmin, which binds to an 11-G single-strand loop structure in the sod2 promoter (13,14,33), suggesting strong interactions among various regulators of sod2. NF-κB acts as a proapoptotic factor, and the expressions of the p65 and p50 subunits are elevated in the retina in diabetes (12,34,35). We show that hyperglycemia increases p65 at both promoter and enhancer, and p50 only at the enhancer of sod2, and that genetic manipulation of p65 (siRNA) in retinal endothelial cells further decreases sod2 transcripts, supporting the role of activation of p65 at sod2.

In addition, our results show that the p65 subunit of NF-κB also interacts with H4K20me3, raising the possibility that this interaction could contribute to the decreased MnSOD activity in diabetes and that the binding of H4K20me3 in sod2 may facilitate the recruitment of p50 to its enhancer, decreasing the transcription. In support, epigenetic modifications are postulated in altering gene expression patterns associated with various diseases, including diabetes (17,36–38).

Reinstitution of normal glycemic control for 2 months after 2 months of hyperglycemia failed to normalize the increased H4K20me3, p50, p65, and acetyl H3K9 at sod2, and sod2 gene expression continued to be subnormal. The persistent epigenetic regulation of retinal sod2 in the normal glycemic state after a period of hyperglycemia is consistent with the results showing persistent epigenetic regulation of NF-κB in aortic cells (17,24). Here we show that hyperglycemia increases SUV420h2 and methylation of H4K20, and reversal of hyperglycemia fails to normalize it. This could be due to regulation of histone modification by increased oxidative stress experienced by the retina in diabetes, and once these modifications are formed, they are inherited by the next cell cycle independent of the actual glycemia (17,24,36–38). In support, our recent study demonstrated a significant role of histone acetylation in the development of diabetic retinopathy and in the metabolic memory phenomenon (25), and others have shown that short-term hyperglycemia, followed by normal glucose, have sustained an increase in H3K4me1 and a decrease in H3K9me3 at the promoter of p65 (24,36–38). The possibility that the increased NF-κB p65 at sod2 promoter might be due to increased levels of p65 in hyperglycemic milieu cannot be ruled out.

Experimentally, galactosemic rats do not experience insulin deficiency or lipid/protein dysmetabolism but present retinal histopathology similar to that observed in diabetic rats (39). Similar epigenetic modifications in galactosemic animals strongly imply that hyperglycemic insult is sufficient to initiate epigenetic modifications of retinal sod2 and that the process continues even after the insult is halted.

Good glycemic control, if initiated soon after the induction of diabetes, protected epigenetic changes in retinal sod2. This is consistent with our previous results that show that if the rats are maintained in good glycemic control immediately after the induction of diabetes, the retina escapes from increased oxidative stress and nitrotyrosine accumulation (7,19) and imply that the epigenetic modifications observed in PC-Rev group during good glycemic control are not influenced by high insulin administered to maintain such glycemic control. This further strengthens the importance of early and sustained good glycemic control for diabetic patients.

The retina is a complex tissue with multiple layers and cell types, and capillary cells are the major targets of the histopathology characteristic of diabetic retinopathy (40). Results from our laboratory and others have shown that retinal capillary cells incubated in high glucose show similar abnormalities as those observed in the retina in diabetes, and these cells present similar metabolic memory phenomenon as observed in the retina from diabetic rodents (9,12,21,23,41–46). The active p65 subunit of NF-κB is linked with persisting epigenetic changes in aortic endothelial cells, and these changes are maintained even hyperglycemia is removed (17). We show that genetic manipulation of SUV420h2 prevents a glucose-induced increase in H4K20me3 at the sod2 enhancer and a decrease in sod2 transcript, and overexpression of sod2 prevents an increase in H4K20me3 at sod2. This suggests that epigenetic regulations of sod2 and SUV420h2 are under the control of superoxide radicals. Failure of these modifications to reverse after termination of hyperglycemia further confirms the role of sod2 epigenetic modifications in the development of diabetic retinopathy and the metabolic memory phenomenon.

The retina and its capillary cells experience mitochondrial dysfunction in diabetes, and the pathogenic mechanisms that are postulated in the development of complications have a single unifying process—superoxide production from the mitochondria (47). In rodent models, compromised MnSOD activity and mitochondrial dysfunction precede the development of diabetic retinopathy (1,3,8,48), and the mitochondria remain dysfunctional when the histopathology is seen in the retinal vasculature (3,4,7). Overexpression of sod2 protects mitochondrial DNA from being damaged and the retinal vasculature from the pathology (6,8,9,23), suggesting a major role of MnSOD in the development of diabetic retinopathy. Thus, the significance of epigenetic regulation of sod2, observed in the current study in the retina and in its capillary cells, further supports its contributions in the development and progression of diabetic retinopathy.

Histone modifications influence the overall chromatin structure and result in functional consequences in cellular processes, and methylation of H4K20 has been implicated in DNA damage, chromatin maintenance, and transcriptional repression. Although H4K20me1 is associated with cell cycle regulation, H4K20me3 is associated with cancers and aging (49,50). Our data demonstrate that diabetes increases total retinal H4K20 methylation, and increases in H4K20me2 and H4K20me3 are significantly greater compared with H4K20me1. Reversal of hyperglycemia fails to provide any benefit. Furthermore, we show that, SUV420h2, an enzyme important for di- and trimethylation, is increased, and that inhibition of SUV420h2 prevents H4K20me3 at the sod2 enhancer and ameliorates a decrease in sod2 gene expression. However, others have shown glucose-induced monomethylation of H3K4 by Set7 in aortic endothelial cells is associated with the transcription activation and regulation of NF-κB and is important in the sustained increase in the NF-κB p65 gene (17,24,37). Our data showing a sustained increase of SUV420h2 in the retina and its capillary cells in hyperglycemia, and prevention of H4K20me3 at sod2 by its siRNA, strongly suggest SUV420h2 is important in the metabolic memory phenomenon associated with diabetic retinopathy. We cannot, however, rule out the regulation of retinal sod2 by other histone and nonhistone modifications, including methylation of H3K9 and activation of Set7/9.

In conclusion, this is the first report showing that retinal sod2 is epigenetically regulated in diabetes and that these modifications continue after termination of hyperglycemia, suggesting their role in the metabolic memory phenomenon associated with the continued progression of diabetic retinopathy. Modulation of epigenetic changes by pharmaceutical or molecular means may provide a potential strategy to retard the development/progression of diabetic retinopathy.

This study was partly supported by grants from the National Institutes of Health, Juvenile Diabetes Research Foundation, the Thomas Foundation, and Research to Prevent Blindness.

No potential conflicts of interest relevant to this article were reported.

Q.Z. researched data and wrote the manuscript. R.A.K. wrote the manuscript and reviewed and edited the manuscript.

The authors appreciate the technical assistance of Mamta Kanwar, Yakov Shamailov, and Dr. Julia Santos, Wayne State University, for help with PCR.