QTL analysis in the (B6 × CSS-13) F2 population
We first examined the average 24-h wheel running activity through 3 weeks of recording in 53 (B6 × CSS-13) F1 animals. The average wheel running activity in the F1 mice (24439 ± 960 revolutions/day) is similar to the CSS-13 parental strain (24077 ± 1459 revolutions/day) but is significantly less than the B6 parental strain (32288 ± 1163 revolutions/day, P < 0.01). To increase the resolution of genetic mapping, a total of 163 (B6 × CSS-13) F2 animals having at least one recombination between 23 and 83 Mb on chromosome 13 were selected for the QTL analysis. Single marker association analysis revealed a highly significant QTL with a peak LOD score of 8.18 at 39 Mb (rs3720620, P < 0.001), as shown in Fig. 1a. The LOD score threshold for significance (P = 0.05) was 3.50, calculated by 1000-time permutation. Descriptive analysis indicated that wheel running activity over 24 h in the F2 population was normally distributed with an average daily activity count of 26575 ± 389. Characterization of the F2 progeny indicated that animals carrying homozygous B6 alleles at the peak loci of the QTL had a significantly higher level of wheel running activity (29722 ± 695 revolutions/day) compared to the animals carrying homozygous A/J alleles (24532 ± 746 revolutions/day, P < 0.001), whereas animals with heterozygous alleles were intermediate (26148 ± 521 revolutions/day). This QTL accounts for 15% of the total variance in wheel running activity in the F2 population. In addition, an epistasis analysis in the F2 population using R/qtl did not detect any interactive pair of loci for the 24-h wheel running activity trait.
Figure 1. Genetic mapping of 24-h voluntary wheel running activity in a (B6 × CSS-13) F 2 population (a) and ISCS (b). (a) QTL analysis using a total of 14 SNP markers on chromosome 13 in 163 recombinant F2 animals. The peak of the QTL was mapped to 39 Mb (rs3720620) with a LOD score of 8.18. The LOD score threshold was designated by the dotted line. (b) The locations of genetic markers examined to generate the congenic interval boundaries are shown (in megabases) at the bottom. For each congenic strain, the donor B6 and A/J homozygous segment is shown in black and white, respectively, and the boundary between the B6 and A/J regions is shown in gray. The level of 24-h wheel running activity for each strain is illustrated in blue bars (high-activity strains) and red bars (low-activity strains) on the left side, with n representing the number of animals tested in each strain. Levels of activity are reported as mean ± SEM revolutions/day. The blue arrow designates the common B6 allele shared by the three high-activity congenic lines, HR-1, HR-2 and HR-3. The red arrow designates the common A/J allele shared by the four low running congenic lines, LR-1, LR-2, LR-3 and LR-4. The VRA QTL, highlighted in yellow, is narrowed down to a 3.76-Mb interval (38.84–42.60 Mb), which was identified in the overlapping region of the red and blue arrows.
Download figure to PowerPoint
QTL fine mapping using ISCS
To further narrow down the QTL, seven ISCS (HR-1, HR-2, HR-3, LR-1, LR-2, LR-3 and LR-4) were generated carrying recombination between the 76.01–83.00, 55.51–59.01, 38.84–40.95, 40.95–42.60, 47.00–55.51, 59.01–69.00 and 69.00–76.01 Mb intervals on chromosome 13, respectively (Fig. 1b). Daily wheel running activity of the congenic lines as well as the B6 and CSS-13 parental strains are shown on the left side of Fig. 1b. The average number of revolutions per day in the HR-1 (32549 ± 1319), HR-2 (30564 ± 1271) and HR-3 (29799 ± 739) strains was similar to the B6 strain (32288 ± 1163) but was remarkably higher than the CSS-13 strain (24077 ± 1459, P≤ 0.001 for HR-1, HR-2 and HR-3). The three high-activity ISCS shared a common B6 allele from 38.84 to 59.01 Mb, designated by the blue arrow in Fig. 1b. In contrast, the LR-1 (25293 ± 682 revolutions/day), LR-2 (24066 ± 1044 revolutions/day), LR-3 (25319 ± 1491 revolutions/day) and LR-4 (24021 ± 1097 revolutions/day) strains exhibited a significant reduction in daily activity compared to the B6 strain (P < 0.001 for the LR-1, LR-2, LR-4; P = 0.005 for the LR-3), but did not differ from the CSS-13 strain. The four low-activity ISCS shared a common A/J allele from 23.78 to 42.60 Mb, which is marked by the red arrow. Thus, the critical QTL of VRA (VRA QTL), highlighted in yellow, was identified in the overlapping region of the red and blue arrows. This result narrowed the genetic interval for the VRA QTL, to a maximal 3.76-Mb interval (38.84–42.60 Mb). The fact that neither daily activity levels in the HR-1, HR-2 and HR-3 strains were significantly different from each other nor were the activity levels in the LR-1, LR-2, LR-3 and LR-4 strains indicated that no additional QTL in the 23–83 Mb region on chromosome 13 contribute to the difference in the level of wheel running activity between the B6 and CSS-13 strains.
Voluntary physical activity in the HR-3 and LR-1 strains
The HR-3 and LR-1 congenic strains were of particular interest because across the genome, they were only different in the 3.76-Mb interval of interest, yet showed a significant difference in their daily wheel running activity levels. Representative actograms of mice for the B6, CSS-13, HR-3 and LR-1 strains are shown in Fig. 2a–d, respectively. Figure 3 illustrates the hourly distributions of activity in the two congenic strains and the B6, CSS-13 parental strain over a 24-h day. We have previously reported that the CSS-13 strain showed a significantly lower level of wheel running activity and total cage activity, as well as an attenuated diurnal distribution of activity compared to the B6 strain (Yang et al. 2009). However, the total cage activity levels in the HR-3 and LR-1 congenic strains were not significantly different from the B6 strain (B6 = 11299 ± 547, HR-3 = 11109 ± 922, LR-1 = 12369 ± 618 counts/day). In addition, both HR-3 and LR-1 congenic lines had around 95% of the wheel running activity in the dark phase, similar to the B6 strain. Body weights of HR-3 and LR-1 mice were not significantly different before or after wheel running (before running: HR-3: 22.34 ± 1.31 g, LR-1: 22.70 ± 1.89 g; after running: HR-3: 24.56 ± 1.59 g, LR-1: 25.59 ± 1.95 g). Therefore, these results from our high-resolution mapping showed a separation of different aspects of physical activity at a genetic level. The VRA QTL has a large and specific effect on the amount of daily voluntary wheel running activity but not on total cage activity in the absence of a running wheel.
Figure 2. Representative wheel running activity records of a B6 (a), CSS-13 (b), HR-3 (c) and LR-1 (d) mouse. Activity is shown for 4 weeks under a 14:10 LD cycle. Wheel revolutions are indicated by the black tic marks, with the height of marks reflecting the level of activity. The horizontal bar above each activity recording signifies the light (white) and dark (black) periods.
Download figure to PowerPoint
Figure 3. Hourly distribution of wheel running activity of the B6 (open squares) and CSS-13 (light gray diamonds), HR-3 (dark gray circles), LR-1 (black triangles) mice under a 14:10 LD cycle. The B6 and HR-3 strains exhibited significantly higher levels of activity than the LR-1 and CSS-13 strain. However, the HR-3 and LR-1 strains were similar in the diurnal distribution of wheel running activity, whereas different from the CSS-13 strain, which shows onset of activity 1–3 h before light off.
Download figure to PowerPoint
Cocaine-induced locomotor responses
To test the hypothesis that functional alterations in the dopaminergic pathway contribute to a different propensity for voluntary wheel running activity in the ISCS, locomotor responses to acute cocaine administration were examined in the HR-3, LR-1 and B6 mice using an open field apparatus. Cocaine, a dopamine transporter blocker, increases the extracellular concentration of dopamine and induces an increase in locomotor activity. The responses of the HR-3, LR-1 and B6 strains to saline injection (Day 1) in the open field apparatus were indistinguishable from one another (data not shown). However, the LR-1 strain exhibited a significantly higher response to cocaine injection compared with the HR-3 and B6 strains, which is illustrated in Fig. 4a. A one-way anova revealed marked strain differences in normalized activity counts during 30-min postcocaine injection period (F2,31 = 6.176, P = 0.006). Furthermore, a Tukey post-hoc analysis showed a stronger response in the LR-1 mice than HR-3 (P < 0.01) or B6 (P < 0.001) mice, whereas HR-3 and B6 mice were not significantly different (Fig. 4b).
Figure 4. The effect of acute cocaine administration (20 mg/kg) on locomotor activity in the B6, HR-3 and LR-1 strains. Activity was monitored for 60 min and normalized by the average of baseline activity of preinjection. (a) Cocaine-induced hyperactivity was observed immediately after injection and peaked in 10–20 min in all strains. The LR-1 mice (black triangles) exhibited significantly higher responses to cocaine injection compared with the HR-3 mice (gray circles) and B6 controls (open squares), whereas the responses in the latter two were indistinguishable. (b) During 30 min of postinjection, normalized locomotor activity in the LR-1 mice was substantially higher than the HR-3 and B6 mice, whereas HR-3 and B6 mice are not significantly different. Results are reported as mean ± SEM. **P < 0.01, ***P < 0.001 compared with the LR-1 strain.
Download figure to PowerPoint
Expression of dopamine genes in the dorsal striatum and nucleus accumbens
To further investigate possible alterations in the dopaminergic pathway, mRNA levels of the key dopamine-related genes (Drd1a, Drd2, Drd3, Drd4, Drd5 and Slc6a3) were examined in the dorsal striatum and nucleus accumbens in the naÏve HR-3 and LR-1 congenic strains. Expression of Drd1a, normalized by Gapdh, was significantly higher in the LR-1 strain than the HR-3 strain in both the dorsal striatum (P < 0.001, Fig. 5a) and nucleus accumbens (P = 0.05 Fig. 5b). There were no significant differences in the expression of Drd2, Drd3, Drd4, Drd5 and Slc6a3 in any of the two tissues. The differential expression of Drd1a was confirmed by another internal housekeeping gene, Actb, in both the dorsal striatum (P < 0.001, Fig. 5c) and nucleus accumbens (P < 0.001, Fig. 5d).
Figure 5. Expression analysis of dopamine-related genes in the dorsal striatum (a, c) and nucleus accumbens (b, d) of the naÏve HR-3 (gray) and LR-1 (black) congenic strains using Qrt-PCR. Expression levels of dopamine-related genes were normalized by two different housekeeping genes, Gapdh (a, b) and Actb (c, d). Drd1a was expressed significantly higher in the LR-1 mice than the HR-3 mice in both the striatum and nucleus accumbens. *P < 0.05, ***P < 0.001 compared with the HR-3 strain.
Download figure to PowerPoint
Candidate genes in the QTL
On the basis of the Ensembl Gene 60 (NCBIM37) database, we identified 18 known genes and 7 predicted genes within the maximal VRA QTL interval (Fig. 6, Table S1, supporting information). Of the 25 genes, 17 are protein coding and the other 8 are known or predicted noncoding RNA genes. These candidate genes are involved in diverse cellular functions including transmembrane transport, DNA-binding transcription factor activity, regulation of signal transduction, cell adhesion etc. We used the evidence of brain expression as an initial filter in prioritizing potential candidate genes involved in the central regulation of voluntary physical activity. Using the GNF database and the Allen Brain Atlas, as well as our Qrt-PCR data in four brain tissues (frontal cortex, hypothalamus, striatum and cerebellum), we confirmed that 15 of these candidate genes, including Slc35b3, AC125223.1, U1, Tcfap2a, Gcnt2, AC133496.1, Pak1ip1, Tmem14c, Mak, Elovl2, Nedd9, AC133159.1, U6, Hivep1 and Edn1, were expressed at detectable levels in brain. The expression of the remaining 10 candidate genes was too low to be reliably detected in any of the four brain tissues by real-time PCR [i.e., lower than 1/30 000 (≅2−15) of the Gapdh expression level]. In concordance with our data, they were found to have very low levels of expression in brain in the GNF database and the Allen Brain Atlas. Among those expressed in the brain, mRNA levels of U1 and U6 was difficult to assess, as they are represented multiple times in the genome, and a unique query on chromosome 13 is not possible. Genotype-dependent expression of the 13 primary candidate genes was assessed in four brain regions of the HR-3 and LR-1 congenic strains. Two genes, Slc35b3 and Mak, showed evidence of brain regionally specific differential expression: expression of Slc35b3 was significantly higher in the frontal cortex of the LR-1 than HR-3 strain (P < 0.004), and expression of Mak was significantly higher in the hypothalamus of the HR-3 than LR-1 strain (P < 0.004). However, the magnitude of difference in expression was only 11 and 6% in Slc35b3 and Mak, respectively.
Figure 6. SNP histogram and locations of candidate genes in the VRA QTL. Pair-wise sequence comparison was performed to identify genetic regions containing few polymorphisms between the B6 and A/J inbred strains. The names of candidate genes are shown at the top of the black bars. The black bars only illustrate the approximate locations, rather than the precise sizes, of the candidate genes. Candidate genes located in the nonpolymorphic regions are less likely to give rise to the QTL.
Download figure to PowerPoint
Next, a pair-wise sequence comparison between the B6 and A/J strains was performed to analyze polymorphisms in the VRA QTL interval. A survey of the SNP Database in the Wellcome Trust Sanger Institute, which has sequenced common inbred strains to a 20× coverage, revealed regions with few polymorphisms between the two strains (Fig. 6). Such SNP-desert regions are less likely to contain causative polymorphisms, whereas SNP-rich regions in which two strains contain alleles from different ancestral sources are more likely to encode the causal quantitative trait gene (QTG) (Cervino et al. 2006). From this perspective, U1, Ofcc1, Gm9979, Tcfap2a, AC123834.1, Gcnt2, AC133496.1, Pak1ip1, Tmem14c, Mak, Gcm2, Sycp2l, Elovl2 and Nedd9 are located in the high-density SNP regions, thus are more likely to be the QTG. Taken together, these surveys provided a sound and unbiased evaluation of all the genes within the VRA QTL for potential causative QTG based on expression and sequence variation criteria.
As the studies of cocaine-induced locomotor activity and expression of dopamine-related genes supported our hypothesis that the dopaminergic signaling pathway, especially the Drd1a, contributes to the regulation of wheel running activity in the ISCS, we further examined possible interactions of the QTL candidate genes with the dopamine-related genes. IPA was employed to explore the functional relationship between the 25 QTL candidate genes and the 6 dopamine-related genes. As shown in Fig. 7a, the transcription factor AP-2α (Tcfap2a) was found to have a direct interaction with the Drd1a. The Tcfap2a-Drd1a interaction was previously identified using a yeast one-hybrid screen and supported by gel mobility shift assays. It was found that Tcfap2a could bind the AP2 consensus sequences located in the first intron of Drd1a, which contains a promoter activator region (Yang et al. 2000).
Figure 7. Identification of QTL candidate genes interacting with the dopamine-related genes. (a) Pathway analysis was employed to explore the functional relationship between the QTL candidate genes and dopamine-related genes. Proteins are depicted as nodes in different shapes representing the functional classes of the gene products, and the biological relationship between genes are depicted by edges. Solid and dashed edges indicate direct and indirect molecular interactions, respectively. Among 25 candidate genes, Tcfap2a has a direct connection with Drd1a, whereas Gcm2, Hivep1, Edn1 and Pak1ip1 have indirect connections with the dopamine-related genes. These five QTL candidate genes are highlighted by orange bars, and six dopamine-related genes are marked by green triangles. (b) Luciferase reporter assay was performed to confirm a direct interaction between Tcfap2a and Drd1a. The first intron of Drd1a containing the AP-2 consensus binding site was cloned into the pGL4 luciferase reporter vector using the restriction enzymes KpnI and BglII. Cotransfection of the pGL4 construct (40 ng/reaction) and four different amounts of the Tcfap2a cDNA clone (10, 20, 40 and 80 ng/reaction) in the HEK293 cells generated significantly higher luciferase signal compared with the single transfection of the pGL4 construct containing Drd1a intron without Tcfap2a.
Download figure to PowerPoint
To further confirm the role of Tcfap2a in the transcriptional regulation of Drd1a, we cloned the first intron of Drd1a into the pGL4 luciferase reporter vector using the restriction enzymes KpnI and BglII (New England Biolabs, Ipswich, MA, USA). Cotransfection of the murine Tcfap2a cDNA clone (10, 20, 40 and 80 ng) and the pGL4 luciferase construct (40 ng) containing AP2 consensus sequences of Drd1a significantly stimulated luciferase expression in the HEK 293 cells (Fig. 7b), suggesting that Tcfap2a can regulate the promoter activity of Drd1a and supporting the hypothesis that Tcfap2a may regulate voluntary wheel running activity via its interaction with the dopaminergic signaling pathway through Drd1a.
As shown in Fig. 7a, Gcm2, Hivep1, Edn1 and Pak1ip1 have indirect connections with the dopamine-related genes. Among these, however, Hivep1 and Edn1 are located in the genetic region having few polymorphisms, and the expression of Gcm2 is undetectable in brain tissues, making them less likely to be the causative QTG. However, we cannot exclude the possibility that additional candidate genes in the QTL interval may also influence wheel running activity and warrant further attention in future studies.