Functional effects of dopamine transporter gene genotypes on in vivo dopamine transporter functioning: a meta-analysis

Abstract

Much psychiatric genetic research has focused on a 40-base pair variable number of tandem repeats (VNTR) polymorphism located in the 3’-untranslated region (3’UTR) of the dopamine active transporter (DAT) gene (SLC6A3). This variant produces two common alleles with 9- and 10-repeats (9R and 10R). Studies associating this variant with in vivo DAT activity in humans have had mixed results. We searched for studies using positron emission tomography (PET) or single-photon emission computed tomography (SPECT) to evaluate this association. Random effects meta-analyses assessed the association of the 3’UTR variant with DAT activity. We also evaluated heterogeneity among studies and evidence for publication bias. We found twelve studies comprising 511 subjects, 125 from PET studies and 386 from SPECT studies. The PET studies provided highly significant evidence that the 9R allele was associated with increased DAT activity in human adults. The SPECT studies were highly heterogeneous. As a group, they suggested no association between the 3’UTR polymorphism and DAT activity. When the analysis was limited to the most commonly used ligand, [123I]β-CIT, stratification by affection status dramatically reduced heterogeneity and revealed a significant association of the 9R allele with increased DAT activity for healthy subjects. In humans, the 9R allele of the 3’UTR polymorphism of SLC6A3 regulates dopamine activity in the striatal brain regions independent of the presence of neuropsychiatric illness. Differences in study methodology account for the heterogeneous results across individual studies.

INTRODUCTION

The dopamine active transporter (DAT) is a key regulator of the dopamine system and the gene that encodes it (SLC6A3) has been the focus of much research in biological psychiatry, having been implicated in several disorders including attention deficit hyperactivity disorder (ADHD),^1–3 pediatric bipolar disorder,⁴ Tourette syndrome⁵ and alcoholism,⁶ but not schizophrenia.⁷ Much research in this area has focused on the DAT gene (SLC6A3), especially a 40-base pair variable number of tandem repeats (VNTR) polymorphism located in the 3’ -untranslated region (3’UTR) of the gene, which has a regulatory role during transcription. This variant produces two common alleles with 9- and 10-repeats (9R and 10R). In humans, meta-analyses suggest the 10R allele of this polymorphism is associated with ADHD in youth⁸ whereas the 9R allele is associated with ADHD in adults.⁹ Meta-analysis also associates the 9R allele with alcoholism,⁶ a common comorbidity of ADHD in adults.

The DAT was initially implicated in ADHD by the stimulant drugs, which are efficacious for the disorder and block the DAT, thereby increasing the concentration of dopamine in the synaptic cleft. These effects are most pronounced in the nucleus accumbens and dorsal striatum due to the high density of DATs in these regions.^10,11 Positron emission tomography (PET) studies in humans show that both methylphenidate¹² and amphetamine¹³ increase extracellular dopamine levels in the striatum. Single-photon emission tomography (SPECT) and PET studies also show that methylphenidate treatment blocks the DAT.¹⁴ Consistent with this, methylphenidate normalizes elevated DAT densities in a rat model of ADHD.³ Based on a meta-analysis of nine in vivo SPECT and PET studies, Fusar-Poli et al.¹⁵ concluded that DAT activity was 14% higher in ADHD patients compared with controls and that, among ADHD patients, DAT activity was higher among patients with a history of medication (although this latter conclusion has been questioned due to incorrect coding of medication status.¹⁶)

Functional in vitro studies have shown mixed results as to whether it is the 9R or 10R allele that increases DAT gene expression.^17–21 These results varied in the reporter gene designs and cell types used. A few studies measured in vivo striatal DAT gene expression using postmortem brains, and the results were also inconsistent.^22–25 MRI and magnetic resonance spectroscopy studies have also produced heterogeneous results.^26–28 A review of neuropsychological studies found little evidence supporting the idea that the SLC6A3 3’ UTR is associated with deficits in cognition,²⁹ with the possible exception of functions mediated by the striatum.²⁷

PET and SPECT neuroimaging studies have examined the association of the 3’UTR polymorphism with in vivo striatal DAT binding in humans. Such studies are particularly compelling because they directly measure the protein produced by the gene rather than measuring mRNA level, or downstream effects of brain activation or cognition. DAT binding may be an intermediate phenotype that mediates the effects of DAT gene variants on dopamine-regulated brain functions and, ultimately, a wide array of behavior; including information processing, inhibition, emotion, movement, salience and reward. Advances in molecular imaging and the development of highly specific DAT binding ligands allow for the imaging of the DAT in humans facilitating the direct in vivo examination of the product of the DAT gene and its polymorphisms in humans. Thus, studies of the SLC6A3 3’UTR polymorphism and DAT binding provide evidence as to whether this polymorphism regulates DAT functioning in living humans.

Results from such in vivo studies could advance our understanding of the genetic control of dopamine as an important neuromodulator of brain function. However, they have produced mixed results. Only two PET studies addressed this issue. These studies of striatal structures reported that the 9R allele was associated with increased DAT activity.^30,31 In contrast, ten SPECT studies produced conflicting results. A meta-analysis of eight of these SPECT studies concluded that these studies did not provide support for the putative association between the SLC6A3 3’UTR polymorphism and DAT activity in the brain.³²

In vivo imaging of DAT is particularly relevant for ADHD given that DAT is the target of stimulant medications and, subsequently, a target protein for studies of pathophysiology. Clarifying the nature of the association of SLC6A3 variants would provide a key step towards identifying part of ADHD’s pathophysiology. Depending upon the strength of the relationship, it could point toward a means of parsing ADHD’s heterogeneity, which could have implications for treatment development.

The goal of the present study was to clarify the PET and SPECT imaging studies by updating Costa et al.’s³² meta-analysis in several ways. Costa et al.’s³² meta-analysis is limited in several ways. It ignored two available SPECT studies,^33,34 which could potentially add to our understanding of the magnitude of effects and sources of heterogeneity and it was conducted prior to publication of the two recent PET studies described above.^30,31 Costa et al. did not (or could not) assess the degree to which heterogeneity of results could be accounted for by sample characteristics (affected vs. healthy), imaging method (PET vs. SPECT) and, for SPECT studies, type of ligand. As we show in this manuscript, attending to these key issues provides a better understanding of the association between the SLC6A3 3’UTR polymorphism with in vivo striatal DAT binding in humans.

METHODS

A PubMed literature search identified studies that met the following criteria: (1) use of SPECT or PET to assess DAT availability in the brains of human subjects; (2) genotyping of the DAT gene (SLC6A3) 3’UTR VNTR. (3) reporting of means and standard deviations of DAT availability stratified by genotype; (4) reporting of the numbers of subjects in each genotype group. We used the following search algorithm in PubMed: ‘dopamine transporter’ [TIAB] OR DAT1[TIAB] OR SLC6A3[TIAB] AND (imaging [TIAB] OR single-photon [TIAB] OR SPECT [TIAB] OR PET [TIAB] OR ‘positron emission tomography’[TIAB]) AND (genetic [TIAB] OR genotype [TIAB] OR genotypes [TIAB] OR allele [TIAB] OR alleles [TIAB] OR polymorphism [TIAB]). If the reference sections of any of these articles suggested additional articles, these were also examined. The preferred reporting items for systematic reviews and meta-analyses (PRISMA) diagram in Figure 1 describes the number of articles identified and their disposition. If the required data were not available in relevant articles, we contacted authors for that data or extracted it from a prior summary of a subset of relevant studies.³²

Figure 1.

PRISMA flow diagram. Note: PRISMA = Preferred reporting items for systematic reviews and meta-analyses (http://www.prisma-statement.org/).

We computed separate meta-analyses for the PET and SPECT studies. Our meta-analysis used the random effects model of DerSimonian and Laird,³⁵ which computes a pooled standardized mean difference weighted by sample size. We use the I² index to assess the heterogeneity of effect sizes.³⁶ Its value lies between 0 and 100 and estimates the percentage of variation among effect sizes that can be attributed to heterogeneity. A significant I² suggests that the effect sizes analyzed are not estimating the same population effect size. We used meta-analytic regression to assess the degree to which effect sizes varied with methodological features.^37,38 The meta-analyses and meta-analytic regressions were weighted by the reciprocal of the variance of the effect size. We used Egger’s³⁹ method to assess for publication biases.

RESULTS

Table 1 gives the characteristics of each study’s sample. Ten studies used SPECT and these used four different ligands. Two studies used PET, one using ¹¹Altropane as the ligand and the other ¹¹Cocaine. Six studies were of Caucasian samples, one used a Korean sample and the others were of mixed ethnicity. As reported by Costa et al.,³² data from van Dyck et al.⁴⁰ overlapped with data reported by Jacobsen et al.⁴¹ and Martinez et al.⁴² For our meta-analyses, we used the non-overlapping van Dyck⁴⁰ data that had been presented by Costa et al.³² The twelve studies comprised 511 subjects, 125 from PET studies and 386 from SPECT studies.

Table 1.

Description of studies providing data

Study	Method	Ligand	Ethnicity
Cheon34	SPECT	[123I]IPT	Korean
Contin87	SPECT	[123I]FP-CIT	C
Heinz88	SPECT	[123I] β-CIT	C
Jacobsen41	SPECT	[123I] β-CIT	AA, C
Krause75	SPECT	89MTC-TRODAT–1	C
Lafuente90	SPECT	[123I]FP-CIT	C
Lynch43	SPECT	89MTC-TRODAT–1	C, AA, other
Martinez42	SPECT	[123I]β-CIT	AA, C
Shumay30	PET	[11Cocaine]PET	AA, C, Other
Spencer31	PET	[11Altropane]	AA, C
Van Dyck40	SPECT	[123I]β-CIT	C
Van de Giessen91	SPECT	[123I] β-CIT	C

Abbreviatons: AA, African–American; C, Caucasian; PET, positron emission tomography; SPECT, speeded photon emission tomography.

Table 2 describes the data used in the meta-analysis. Some studies provided data for both healthy and affected groups. Several disorders were studied: ADHD, schizophrenia, alcoholism and Parkinson’s disease. Most studies imaged the striatum or subareas of the striatum. The only exception was Cheon et al.,³⁴ who studied the left and right basal ganglia. If a study provided data on a structure (for example, putamen) and subdivisions of the structure (for example, left and right putamen), we only analyzed the substructures. Each study compared a 10 repeat (10R) genotype group with a 9 repeat (9R) genotype group. As Table 2 shows, nearly all studies defined the 10R group as those carrying two 10R alleles (10R/10R). The only exception was Lynch et al.,⁴³ who also included 10R heterozygotes carrying a rare allele other than the 9R allele. The 9R genotype groups primarily comprised 9R homozygotes and 9R/10R heterozygotes. However, some rare genotypes that did not include a 10R allele were also included in some samples.

Table 2.

Data used in the meta-analysis

Study	Sample	Mean age and range	Brain region	9R genotype group	N with 9R genotypes	Mean of 9R group	SD of 9R group	10R genotype group	N with 10R genotypes	Mean of 10R group	SD of 10R group
Shumay30	Healthy	34.5 AA: 21.5–45.5 CE: 20.5–49.5 Others: 22.2–40.3	Caudate	9R/9R 9R/10R 9R/11R	47	0.79	0.14	10R/1 OR	44	0.76	0.15
Shumay30	Healthy	34.5 AA: 21.5–45.5 CE: 20.5–49.5 Others: 22.2–40.3	Putamen	9R/9R 9R/10R 9R/11R	47	0.96	0.13	10R/10R	44	0.9	0.13
Shumay30	Healthy	34.5 AA: 21.5–45.5 CE: 20.5–49.5 Others: 22.2–40.3	Ventral striatum	9R/9R 9R/10R 9R/11R	47	0.81	.2	10R/10R	44	0.8	0.19
Spencer31	Healthy	27.8 18–49	Left caudate	9R/9R 9R/10R 9R/11R	18	3.589	0.847	10R/10R	16	3.336	0.6202
Spencer31	ADHD	32.8 18–52.3	Left caudate	9R/9R 9R/10R 9R/11R	15	3.401	0.7424	10R/10R	19	2.988	0.3448
Spencer31	Healthy	27.8 18–49	Light putamen	9R/9R 9R/10R 9R/11R	15	3.349	0.643	10R/10R	19	3.174	0.3701
Spencer31	ADHD	32.8 18–52.3	Left putamen	9R/9R 9R/10R 9R/11R	18	3.386	0.6246	10R/10R	16	3.312	0.4996
Spencer31	Healthy	27.8 18–49	Right caudate	9R/9R 9R/10R 9R/11R	15	3.266	0.5557	10R/10R	19	3.051	0.343
Spencer31	ADHD	32.8 18–52.3	Right caudate	9R/9R 9R/10R 9R/11R	18	3.664	0.62	10R/10R	16	3.247	0.538
Spencer31	Healthy	27.8 18–49	Right putamen	9R/9R 9R/10R 9R/11R	15	3.296	0.5265	10R/10R	19	3.19	0.335
Spencer31	ADHD	32.8 18–52.3	Right putamen	9R/9R 9R/10R 9R/11R	18	3.355	0.7085	10R/10R	16	3.245	0.6071
SPECT [123I]β-CIT studies
Martinez42	Schizophrenia	39.2 ± 9	Striatum	9R/9R 9R/10R	7	7.9	2.1	10R/10R	14	7.8	1.5
Martinez42	Healthy	40.0 ± 9	Striatum	9R/9R 9R/10R 9R/11R	6	8.2	1.5	10R/10R	15	8.2	1.3
Van Dyck40	Healthy	49.9 18–88	Striatum	9R/9R 9R/10R	30	7.5	1.9	10R/10R	35	6.6	1.6
Van de Giessen91	Healthy/prior Ecstasy use	22.0 18–35	Putamen	9R/9R 9R/10R	32	11.4	2.72	10R/10R	45	9.6	2.06
Van de Giessen91	Healthy	22.0 18–35	Caudate	9R/9R 9R/10R	32	13.5	3.25	10R/10R	45	11.4	2.42
Heinz88	Alcoholics/controls	36.7 ± 7	Caudate	9R/10R	10	2399	750	10R/10R	15	2895	724
Heinz88	Alcoholics/controls	36.7 ± 7	Putamen	9R/10R	10	2226	635	10R/10R	15	2840	743
Jacobsen41	Healthy	37 ± 9.3	Striatum	9R/9R 9R/10R	9	8.2	1	10R/10R	18	7.1	1
SPECT [123I]FP-CIT studies
Contin87	Parkinson’s	60.4 9–10 [N = 14): 63 ± 11
9–9 (N = 6): 58 ± 9 10–10 (N = 16)59 ± 7	Putamen	9R/9R 9R/10R	20	0.92	0.54	10R/10R	16	0.91	0.63
Lafuente90	Schizophrenia	24.0 9–9: 26 ± 0 9–10: 22.6 ±3 10–10: 24.8 ± 6.5	Striatum	9R/9R 9R/10R	8	4.6	0.5	10R/10R	6	4.4	0.5
SPECT89MTC-TRODAT-1 studies
Krause75	ADHD	37.6 19–54	Striatum	9R/9R 9R/10R	12	1.31	.27	10R/10R	17	1.28	.34
Lynch43	Healthy	46.5 18.3–83.3	Striatum	9R/1 OR 9R/*R	49	1.175	.26	10R/10R 10R/*R	32	1.25	.22
Lynch43	Parkinson’s	60.8 38.3–84.2	Striatum	9R/10R 9R/*R	58	.65	.215	10R/10R 10R/*R	42	.65	.2
SPECT [123I]IPT studies
Cheon34	ADHD	9.8 6–12	Left basal ganglia	10R/11R 9R/10R	4	2.22	1.39	10R/10R	7	6.91	2.5
Cheon34	ADHD	9.8 6–12	Right basal ganglia	10R/11R 9R/10R	4	2	1.2	10R/10R	7	7.1	1.95

Abbreviations: 9R, 9 repeat allele; 10R, 10 repeat allele; *R, other allele; xR/yR = x repeat allele/y repeat allele genotype.

‘Mean of 9R group’ refers to the mean DAT activity for the 9R group and the ‘s.d.’ column gives the standard deviations.

In the analyses that follow, differences between genotype groups are expressed as the standardized mean difference (SMD) effect size. Positive SMDs favor the 9R allele as being associated with increased DAT activity; negative SMDs favor the 10R allele. Figure 2 shows the results for PET studies. In Figure 2, the dot gives the relative risk and the horizontal line gives the 95% confidence interval. The diamonds give the weighted SMD across studies and the width of the diamond gives its 95% confidence interval. The first two diamonds give pooled results for the affected and healthy subgroups. The last diamond gives results for all studies with the left and right ends of the diamond marking the 95% confidence interval.

Figure 2.

Meta-analysis of the dopamine transporter gene 3’UTR VNTR polymorphism with in vivo dopamine transporter activity assessed by PET. Note: For each comparison, the dot gives the relative risk and the horizontal line gives the 95% confidence interval; the diamonds give the weighted SMD across studies and the width of the diamond gives its 95% confidence interval. The first two diamonds give pooled results for the affected and healthy subgroups. The last diamond gives results for all studies with the left and right ends of the diamond marking the 95% confidence interval. I-squared is a measure of heterogeneity among studies.

Across all PET study observations, the SMD of 0.31 was statistically significant (z = 3.61, P < 0.0009). The SMD was significant for the healthy group (0.31, z = 3.1, P = 0.002). Although the magnitude of the SMD was the same for the affected (all ADHD) group, it was only marginally significant, perhaps due to the smaller sample size (0.33, z = 1.9, P = 0.056). The low and non-significant I² statistic of 0.0% indicates very low heterogeneity of results across all observations. The test for publication bias was not significant (t = 1.3, P = 0.2). Consistent with the finding of no heterogeneity, meta-analysis regression found that the SMDs were not associated with age (F(1, 9) = 1.9, P = 0.22), sex (F(1, 9) = 1.4, P = 0.26) brain region (F(6, 4) = 2.80, P = 0.17), affection status (F(1, 9) = 0.02, P = 0.88) or ligand (F(1, 9) = 1.38, P = 0.27). The non-significant PET findings should be viewed cautiously given that there were only two PET studies.

Figure 3 shows the results for SPECT studies. Across all studies, the SMD of 0.00 was not statistically significant (z = 0.00, P = 0.99). The SMD was positive and significant for the healthy group (0.46, z = 2.1, P = 0.035) and negative but not significant for the affected group (−0.40, z = 1.7, P = 0.093). The test for publication bias was not significant (t = −2.0, P = 0.06). The I² statistic was high and significant for the entire SPECT group (I² = 73.8%, P < 0.0009). Although meta-analysis regression confirmed that one source of heterogeneity was affection status (F(1, 13) = 5.04, P = 0.04), heterogeneity remained high when analyzing the affected (I² = 71.2%, P = 0.004) and healthy groups (I² = 63.1%, P = 0.006) separately. Neither sex (F(1, 13) = 0.1, P = 0.8) nor age were significant predictors of the SMDs (F(1, 13) = 0.5, P = 0.48).

Figure 3.

Meta-analysis of the dopamine transporter Gene 3’UTR VNTR polymorphism with in vivo dopamine transporter activity assessed by SPECT. Note: For each comparison, the dot gives the relative risk and the horizontal line gives the 95% confidence interval; The diamonds give the weighted standardized mean difference across studies and the width of the diamond gives its 95% confidence interval. The first two diamonds give pooled results for the affected and healthy subgroups. The last diamond gives results for all studies with the left and right ends of the diamond marking the 95% confidence interval. I-squared is a measure of heterogeneity among studies.

We explored two additional sources of heterogeneity. As Figure 3 shows, the findings of Cheon et al.³⁴ are markedly discrepant from the other SPECT studies. Their study was the only one to study children (seven 10/10R homozygotes, two 9/10 heterozygotes and two 10/11 heterozygotes). They were the only investigators to use [123I]IPT as a ligand and basal ganglia as the region of interest. Consistent with this, meta-analysis regression found significant effects of brain region F(4, 10) = 3.86, P = 0.038) and ligand F(3, 11) = 5.9, P = 0.01). After excluding Cheon et al.³⁴ from the analysis, the overall I² statistic remained high and significant for the entire SPECT group (I² = 64.7%, P < 0.001) but was reduced to non-significance in the affected group (I² = 10.9%, P = 0.35). The SMDs for the overall group and the affected group did not achieve significance (P’s>0.25).

As suggested by the meta-analysis regression, another potential source of heterogeneity among SPECT studies was the choice of ligand. However, only one ligand, [123I]b-CIT, was used with sufficient frequency to be analyzed separately. In this analysis, the overall I² statistic remained high and significant for the entire SPECT group (I² = 70.7%, P < 0.001) but was low and nonsignificant when separately considering the affected (I² = 15.9%, P = 0.35) and healthy groups (I² = 0.0%, P = 0.47). The SMD was significant and positive for the healthy group (0.67, z = 5.2, P < 0.009). In contrast, it was nearly significant and negative for the affected group (−0.52, z = 1.9, P = 0.057).

DISCUSSION

The prior meta-analysis of the association between the SLC6A3 3’UTR polymorphism with in vivo striatal DAT binding in humans concluded that there was no evidence to support the hypothesized association. In contrast, by analyzing a larger sample and incorporating relevant covariates, our meta-analyses have yielded several firm conclusions based on 12 studies comprising 511 subjects. Although limited by the existence of only two studies, the PET studies provided highly significant evidence indicating that the 9R allele is associated with increased DAT binding in the striatal brain regions. Although the effect size was similar for the healthy and affected samples, the latter effect size did not achieve statistical significance, probably due to the smaller number of observations. Notably, all observations from PET studies, albeit individually not significant, were consistent with the 9R allele predicting greater DAT binding in humans. This consistency of results was reflected in finding zero heterogeneity across these observations.

In contrast to the consistency of findings across PET study observations, the data from SPECT studies was highly and significantly heterogeneous. Meta-analysis regression suggested that heterogeneity among SPECT studies could be due to the brain region, affection status or choice of ligand. Due to the distribution of these features across samples, it was not possible to assess their joint effects. However, when the analysis was limited to [123I]β-CIT, the most commonly used ligand, which was used in 5 of the 10 SPECT studies, stratifying by affection status dramatically reduced the heterogeneity of results. This sub-analysis was consistent with the PET studies in finding a significant association of the 9R allele with increased DAT activity for the healthy group.

Unlike the PET studies, which showed a trend association of the 9R allele and DAT activity for the affected group, for SPECT studies, the effect for the affected group favored the 10R allele and was nearly significant. This difference is likely due to the difference in the composition of the affected sub-samples. For the PET studies, these were all ADHD patients. For SPECT studies, a variety of disorders were studied: ADHD, alcoholism, schizophrenia and Parkinson’s disease.

Our results suggest that heterogeneity of findings across studies can be explained by methodological differences. DAT binding varies considerably in different brain regions, and within the striatum, density gradients are detected from superior to inferior, medial to lateral and anterior to posterior regions, particularly in the caudate nucleus.⁴⁴ Several studies report differences between DAT binding in caudate and putamen.³¹ In a study of ADHD, Jucaite et al.⁴⁵ reported decreased DAT binding in the midbrain but increased DAT binding in the caudate. Due to poor spatial discrimination in SPECT, most of the SPECT studies report a combined striatal value that may obscure apparent effects, if discrete regional expression of SLC6A3 is modified by the VNTR (Table 2, Figure 3).

One must also consider variability among DAT radioligands when interpreting the heterogeneity of findings across studies. Although the core structures of these potent DAT radioligands are derived from CFT (or WIN 35,428), only Altropane contains a 4-fluoro substituent on the phenyl ring, rendering it relatively selective, 28-fold, for the DAT over the serotonin transporter and displaying favorable kinetic properties.^46,47 Among all the studies included in our meta-analysis, only Spencer et al.³¹ used altropane. The more commonly used ligands, CIT, IPT, FP-CIT and TRODAT, contain 4-chloro or 4-iodo substituents, which markedly reduce DAT:serotonin transporter specificity (TRODAT 3:1, CIT: 1:1, FP-CIT 3:1, IPT: 5:1) and require varying lengths of time for the radioligand to wash out from serotonin transporter sites and other non-specific sites.^48–50 In addition, differences in lipophilic properties may affect the ability of the ligands to detect intracellular vs. membrane-bound DAT.

Combining disorders may also obscure findings, as different pathophysiologies may be associated with greater or lesser DAT binding. These could be due to either other genetic and environmental risks or medications that alter membrane SLC6A3 expression. Other variables that are hard to control for include age and smoking status, which are known to affect DAT dramatically.^40,51 DAT densities also decline as a function of age,³⁰ and it remains unclear whether the expression levels of 9R and 10R DAT alleles are equally affected by age. In our analyses, age was not predictive of SMDs, which suggests that age effects do not moderate the effects of SLC6A3 alleles. This finding is, however, tempered by the fact that we only had access to mean ages from each study, which ranged from 9.8 to 60.8. Shumay et al.’s³⁰ PET study found that the association of 3’UTR genotypes and DAT binding was significant across all regions (caudate, putamen, ventral striatum) for younger subjects but not in older subjects despite the use of comparable sample sizes for both groups. Their study also found that the age-related decline in DAT levels was greatest for carriers of the 9R/9R 3’UTR genotype.

The strength of the PET-derived data together with consistency with SPECT imaging data that used CIT as a probe, insinuates the 9R allele in the 3’UTR as a regulator of SLC6A3 expression. The association of the 9R allele with higher DAT binding site density could result from a number of possible pathways: interaction of the 3’ UTR with regulatory proteins or microRNAs, shunting of mRNA to distinct compartments in the neuron, regulation of mRNA stability, turnover, increases in translational efficiency,^52,53 or even remote interaction with regulatory elements of other genes that may affect SLC6A3 expression, stability and trafficking.⁵⁴ A parsimonious interpretation of the functional consequences of elevated DAT is more efficient in clearing extracellular dopamine, yielding lower extracellular levels and reduced dopamine signaling. It may be feasible to test this hypothesis by monitoring extracellular dopamine indirectly, using displacement of D2 receptor occupancy with [11C]raclopride as a surrogate for direct measures of dopamine, in 9R and 10R carriers.

The role of the DAT in pathophysiology and therapeutic response has catalyzed research into the association of DAT density with pathology, and the interrogation of whether SLC6A3 alleles are relevant to SLC6A3 expression, regulation and membrane transporter density. Based on a meta-analysis of 9 SPECT and PET studies, Fusar-Poli et al.¹⁵ concluded that DAT density was 14% higher in ADHD patients compared with controls. Mill et al.²⁵ measured dopamine transporter mRNA levels in the cerebellum, temporal lobe and lymphocytes and reported that dopamine transporter mRNA expression increased with the number of 10R alleles. Brookes et al.²² also found that the 10R allele increased levels of SLC6A3 mRNA in human postmortem midbrain tissue. However, Zhou et al.²⁴ and Pinsonneault et al.²³ failed to find a differential effect of 9R and 10R alleles on SLC6A3 expression in postmortem brains.

In vitro studies of the functional effects of the 3’UTR have produced conflicting results. A study of HEK–293 cells reported that cells containing the 10R allele had a SLC6A3 expression that was 50% greater than cells with the 9R allele.¹⁸ Similarly, using a luciferase reporter system in COS–7 cells Fuke et al.¹⁷ found greater SLC6A3 mRNA expression for constructs containing the 10R allele compared with other alleles. In contrast to these findings suggesting that the 10R allele is associated with greater transcription, using the human neuroblastoma cell line, Inoue-Murayama, et al.²¹ reported that the 9R allele led to more SLC6A3 mRNA expression than the 10R allele. Increased DAT gene expression associated with the 9R allele was also reported by Miller and Madras⁵⁵ using HEK–293 cells transfected with 3’UTR variants using two different promoters. Greenwood and Kelsoe²⁰ found that the 9R allele led to non-significantly increased transcriptional regulation in dopaminergic substantia nigra (SN4741) cell lines. Mill et al.¹⁹ also reported non-significantly greater mRNA expression for 9R compared with 10R constructs evaluated in SH-SY5Y and HEK–293 cell lines. In agreement with these findings, if the 9-repeat allele was subcloned upstream of the viral promoter coupled to a green fluorescent protein reporter, the construct enhanced transcription in an immortalized dopaminergic cell line derived from mouse substantia nigra.⁵⁶

These differences among studies could be due to using different experimental constructs to introduce the mutation into cell lines, the amount of flanking sequence included in the construct, choice of reporter gene, along with variable presence of other SLC6A3 transcription regulators across different cell lines.^19,57 Clarifying this issue will require detailed analyses of polymorphisms of length or of single nucleotides in the DAT gene that conceivably contribute to the dynamic processes regulating DAT density in the brain. The grouping together of multiple alleles by the number/length of repeat sequences of the 3’UTR could mask the relevance of other sequence variations, which contribute to DAT gene regulation, SLC6A3 expression levels and phenotype, either in conjunction with, or independently of the 3’UTR.

The inconsistencies from in vitro and postmortem brain expression studies support in vivo imaging as a direct method Standardized Mean Difference to monitor the relevance of 9R and 10R genotypes on SLC6A3 regulation and function. In a functional magnetic resonance imaging study, ADHD patients homozygous for the 10R allele showed significant hypoactivation in the left dorsal anterior cingulate cortex compared with 9R carriers.²⁶ In another functional magnetic resonance imaging study, carriers of a haplotype including the 10R allele showed differentially modulated neural activation to reward-predicting cues in the caudate nucleus.⁵⁸ In a sample of adults, Hoogman et al.⁵⁹ found that the 9–6 haplotype of SLC6A3 was associated with ADHD and that ADHD adults showed striatal hypoactivation during reward anticipation. There was not, however, an association between SLC6A3 genotypes and striatal hypoactivation. In a magnetic resonance spectroscopy study, Sherk et al²⁸ reported that the 10R allele was associated with higher ratios of NAA/Cho and NAA/Cr in the left putamen. They concluded that the 3’UTR VNTR polymorphism modulates dopaminergic activity, and neuronal function in putamen.

Our meta-analyses suggest that, in humans, the 9R allele of the 3’UTR polymorphism leads to increased DAT activity in the striatal brain regions. These results imply that some of the DAT (and therefore dopamine) regulation could be due to the presence (or absence) of the 9R allele. The relationships among the 3’UTR polymorphism, DAT binding and pathophysiology remain complex. This complexity is well illustrated by the ADHD literature for which meta-analyses show: (a) significantly increased DAT density in ADHD, which was greatest for those having had prior stimulant treatment;¹⁵ (b) an increased prevalence of the 10R allele in ADHD youth;⁸ and (c) an increased prevalence of the 9R allele in ADHD adults.⁶⁰ Because the positive studies in our meta-analysis all used adult samples, our results are consistent with the adult association studies. We cannot, however, explain why the 10R allele has been associated with ADHD in youth, which is either a false-positive finding or reflects the complex regulation of SLC6A3, which we discuss below. Given current sample sizes, reconciling apparent genetic differences between childhood and persistent ADHD has not been feasible. It is, however, notable that the only in vivo DAT imaging study to find a significant association of the 10R allele with DAT binding was also the only study of children. This pattern is consistent with the pattern of genetic association seen in adult and child studies.

None of the studies analyzed addressed other mechanisms that might influence DAT binding. Additional mechanisms were implicated in a study by our group, which reported that, while the 3’UTR polymorphism increased DAT binding regardless of ADHD status, ADHD made an additional, independent contribution to DAT binding.³¹ This suggests that there are additional ADHD-associated genetic or non-genetic mechanisms that influence DAT binding. For example, DAT is constitutively recycled through the endosome.⁶¹ Although majority of DAT is sequestered intracellularly in the recycling endosome, only membrane-associated DAT is functionally available for the reuptake of dopamine. An endosome sodium–hydrogen exchanger protein, encoded by SLC9A9, has been implicated in ADHD^62–66 and disruptive genetic variants of SLC9A9 have been found in ADHD and autism cases.^67,68 A meta-analysis of ADHD GWAS data sets implicated the CHMP7 gene (charged multivesicular body protein 7), another protein involved in endosomal sorting/recyling pathway.⁶⁹ These findings suggest that endosomal pathway genes may be compromised in ADHD and that DAT membrane density could be altered due to mutations in these genes.^70,71

The variability of our DAT binding findings in affected vs. healthy subjects, in child vs. adult samples and in various different brain regions could be influenced by both genetic and non-genetic factors, as well as their interactions, about which we do not currently have a full understanding. Non-genetic mechanisms that alter the striatal DAT density include caffeine, cigarette smoking and alcohol consumption.^51,72–74 The effect of stimulant medications on DAT density is of considerable interest given Fusar-Poli’s¹⁵ report that increased DAT density in ADHD could be accounted for by stimulant treatment. Their conclusion, however, has been questioned due to incorrect coding of medication status.¹⁶ Our study cannot shed much light on this issue because only three of the studies examined ADHD patients and these all used treatment-naïve samples.^31,34,75

There are several potential genetic mechanisms underlying DAT variability. Some studies suggest that the effects of the SLC6A3 3’UTR VNTR are limited to specific haplotypes formed with an intron 8 VNTR.^{2,9,29,76–78} In their study of DAT density using [11Cocaine]PET, Shumay et al.,³⁰ showed that the intron 8 VNTR was associated with DAT levels in caudate and putamen. In contrast, Guindalini et al.⁷⁹ using SPECT with TRODAT–1 did not detect this association. Haplotype analyses by Shumay et al.,³⁰ suggested that higher DAT levels were associated with the 5R intron 8 allele. Thus, it may be the 10R–5R haplotype formed by the 3’UTR/intron8 haplotype is worthy of further study as the variant increases DAT density.

SLC6A3 has not been highly conserved during evolution, especially as regards the 3’UTR VNTR region.⁸⁰ This would be expected to cause variability in gene expression and functionality. Such effects could contribute to the inconsistencies among previous studies. Moreover, DAT is expressed in a region-specific manner in the brain⁸¹ and demonstrates an age-dependent profile,⁸² with multiple alternative transcription initiation and splice isoforms existing. The SLC6A3 gene region is enriched for transcription factor and miRNA-binding sites and DNA methylation sites. Many of these regulatory sites are co-localized with known variants such as sequence repeats, single-nucleotide polymorphisms (SNPs) and copy number variations.⁸⁰

For example, the number of the 3’UTR VNTR repeats can change the length of the transcribed mRNA, which may alter the secondary structure and degradation rate of the mRNA. This may also alter the efficiency of the miRNA-binding sites due to changes of sequence and of secondary structure. This cascade of signaling could continue with changes in transcription and degradation rates. Furthermore, miRNA expression itself is often tissue specific and developmentally regulated.⁸³ Some miRNAs simultaneously interact with both the 3’ and 5’UTR regions.⁸⁴ Shumay et al⁸⁰ showed the presence of such sites in the 3’UTR region of SLC6A3, indicating that variations in the 5’ regulatory region of the gene may influence the function of the 3’UTR VNTR via long-range interactions. Consistent with this, Brookes et al.¹ reported replicated associations between ADHD and SNPs in the 5’ regulatory region of SLC6A3. Drgon et al.⁸⁵ reported that haplotypes of two SNPs in the 5’ regulatory region were associated with in vivo DAT activity measured by [11Cocaine]PET and also with striatal DAT activity in postmortem brain samples.

Signals from the ENCODE chromatin interaction analysis with paired-end tag sequencing (ChIA-PET) also indicate various long-range physical interactions of the 3’UTR VNTR. By changing the length and sequence of the mRNA 3’UTR, the VNTR may alter these long-range interactions in synergy with other 5’-haplotypes that together regulate SLC6A3 transcription and splicing. These long-range regulatory interactions, mediated through complex transcription factor interactions, are often sensitive to cell-type specificity and developmental stage.⁸⁶ These interactions offer another explanation for some of the discrepancies seen between adults and children in genetic association studies and among the in vitro studies.

Our conclusions are limited by methodological issues. Because only two PET studies were available, the power for these studies, which comprised 125 subjects, was lower than the power for the SPECT studies, which comprised 386 subjects. Although this low power does not vitiate the statistically significant results, the lack of significance for the affected group should be viewed cautiously. A range of disorders in which dopamine dysregulation has been implicated (Parkinson’s schizophrenia, ADHD, alcoholism) were represented in our analyses. The variability in data from these cohorts may conceivably reflect DAT regulatory processes, which compensate for dysfunctional dopamine signaling. Cheon et al.’s study,³⁴ the most prominent outlier, was the only study conducted in children and used IPT as a SPECT probe. IPT shows an unusually high sensitivity to age-dependent DAT decline.⁴⁹ Information on the age of each 9R or 10R subject would improve interpretation of these data. Like all meta-analyses, our analyses of covariates were limited by the information provided in the papers we reviewed. Notably, we did not have sufficient information about any single disorder to draw firm conclusions about their potential moderating effects on SLC6A3 DAT associations. As is apparent from the Tables, our analyses were also limited by the variability of brain regions across studies. None of the studies in the meta-analysis adjusted their analyses for genomic background using ancestrally informative SNPs. We could not correct the meta-analysis for ethnicity (a coarse measure of genomic background), as there were too many ‘mixed’ ethnic samples. If the SLC6A3 polymorphism we studied is associated with genomic background and if genomic background is associated with DAT availability, these results could be spurious. Caution about ethnic effects is suggested by the work of Shumay et al.,³⁰ which found significantly different SLC6A3 genotype distributions between Caucasians and African–Americans. In their study, the association between SLC6A3 genotypes and DAT density was significant for Caucasians but not African–Americans.

Moreover, heterogeneity of genomic background or of environmental exposures relevant to DAT binding might have accounted for the heterogeneity of findings if they had been measured. We also could not adjust our analyses for smoking, which was not consistently reported. Taken together, these problems limited our ability to find significant covariate effects, but they would not have created spurious results.

Our results along with the limitations of our work highlight directions for future functional studies of genetic variants using in vivo studies of DAT density. PET imaging has clearly been most effective in producing replicable results. Thus, attempting to ‘replicate’ these findings with other imaging methods would not be appropriate. Future studies should genotype ancestrally informative SNPs to assure that differences in genomic background do not affect results. Careful documentation of medications, smoking history and alcohol use is also essential. And, given the complex regulation of SLC6A3, genotyping tag SNPs in all regulatory regions would be a real advance.

Despite these limitations, our meta-analyses suggest that in human adults, the 9R allele of the 3’UTR polymorphism of the DAT gene regulates dopamine activity in the striatal brain regions independent of the presence of ADHD or other disorders. Future in vivo neuroimaging studies of the DAT should attend to the methodological features we highlighted as contributing to the heterogeneity of findings across studies.