The 22q11 PRODH/DGCR6 deletion is frequent in hyperprolinemic subjects but is not a strong risk factor for ASD
Abstract
The proline dehydrogenase (PRODH) gene maps to 22q11.2 in the region deleted in the velo-cardio-facial syndrome (VCFS). A moderate to severe reduction (>50%) in PRODH activity resulting from recessive deletions and/or missense mutations has been shown to cause type 1 hyperprolinemia (HPI). Autistic features have been reported as a common clinical manifestation of HPI. Here we studied the frequency of a recurrent small 22q11.2 deletion encompassing PRODH and the neighboring DGCR6 gene in three case-control studies, one comprising HPI patients (n = 83), and the other two comprising autism spectrum disorder (ASD) patients (total of n = 2800), analyzed with high-resolution microarrays. We found that the PRODH deletion is a strong risk factor for HPI (OR = 50.7; 95%CI = 7.5–2147) but not for ASD (P = 0.4, OR = 0.6–3.3). This result indicates either that the suggested association between ASD and HPI is spurious and results from a bias leading to the preferential inclusion of patients with autistic features in HPI series, or that HPI is present in only a very small subset of ASD patients. In this latter case, a very large sample size would be required to detect an association between the PRODH deletion and ASD in a case-control study. © 2016 Wiley Periodicals, Inc.
INTRODUCTION
The PRODH gene encodes proline dehydrogenase, a mitochondrial enzyme that catalyzes the conversion of proline into Δ1-pyrroline-5-carboxylate (P5C). It is located in the 22q11.2 chromosomal region, near the segmental duplication that defines the proximal boundary of the 3 Mb segment commonly deleted in velo-cardio-facial syndrome (VCFS) [Bittel et al., 2009]. Typical 22q11.2 deletions in VCFS patients include ∼30 genes. We previously identified a smaller 22q11.2 deletion encompassing only PRODH and the neighboring DGCR6 gene in a schizoaffective patient without VCFS clinical features [Jacquet et al., 2002] (Fig. S1). The frequency of this copy number variant (CNV) was subsequently reported to be 1/250 individuals in a Japanese population, making it 10-fold more prevalent than the VCFS-associated 22q11.2 deletion [Ohtsuki et al., 2004]. The reciprocal duplication was even more common, with an estimated frequency of 2.5/100 individuals [Guilmatre et al., 2010]. The PRODH/DGCR6 deletion is frequently observed in patients with type 1 hyperprolinemia (HPI) [Jacquet et al., 2003; Guilmatre et al., 2010]. Since HPI is a recessive condition, the PRODH deletion must be combined with another hit (usually a missense mutation reducing enzymatic activity) on the second allele to cause an increase in plasma proline level [Guilmatre et al., 2010]. The clinical phenotype of HPI is still not well characterized. While some HPI subjects have intellectual disability, seizures, language delay, autistic features, schizophrenia, or bipolar disorder, others are asymptomatic [Afenjar et al., 2007; Di Rosa et al., 2014, 2008; Guilmatre et al., 2010]. In particular, autism spectrum disorder (ASD) has been reported as a prevalent clinical feature of HPI, although the diagnosis often does not rely on formal criteria [Afenjar et al., 2007; Di Rosa et al., 2008; Guilmatre et al., 2010]. The involvement of HPI in ASD determinism is supported by the fact that PRODH deficiency in mice results in increased neurotransmitter release at glutamatergic synapses [Paterlini et al., 2005] and glutamate is possibly involved in ASD pathology [reviewed in Choudhury et al., 2012].
Thus, the question arises whether the PRODH/DGCR6 deletion is associated with ASD. Supporting a putative role as a risk factor, Guilmatre et al. [2009] reported a significant association of this deletion with ASD in a small sample (9/260 ASD cases vs. 1/236 controls, P = 0.02 Fisher exact test, OR = 8.4; 95%CI = 1.15–370.2). Due to their combined frequency >1% in the general population, PRODH/DGCR6 deletions and reciprocal duplications were not reported in most of the recently published large CNV case-control studies in ASD, which focused on rare variants, usually with a frequency <1% [Pinto et al., 2010; Sanders et al., 2011; Gai et al., 2012; Girirajan et al., 2013; Pinto et al., 2014]. Moreover, since these studies used different microarray platforms, this CNV might have been missed due to insufficient probe coverage or differences in CNV-calling algorithms. To further clarify the implication of the PRODH/DGCR6 CNV in ASD risk, we examined its frequency in three case/control studies, one including HPI subjects and two including ASD patients, using data generated by both CGH and SNP array platforms.
PATIENTS AND METHODS
HPI Study
To constitute the HPI cohort, we reviewed all HPI cases (with fasting plasma proline > 400 μmol/L) in which a genetic analysis of the PRODH gene had previously been undertaken by our laboratory. Only cases for whom the phase of PRODH mutations was unambiguously established were retained (n = 83). VCFS patients were not included. The mean age of the patients was 9.6 years (SD = 2.8, range = 2–42). All cases except five were Caucasian. The control cohort consisted of 236 adult healthy Caucasian subjects with normal plasma proline level [Guilmatre et al., 2009]. This study was approved by the Rouen University Hospital ethics committee and all patients or their parents and controls signed a written informed consent. The DNA samples were derived from whole blood. CNV analysis was performed by using the Human High-Resolution Discovery Microarray Kit 1 × 1M (design ID 021529, Agilent Technologies, Santa Clara, CA) following standard protocols. A non-commercial genomic DNA pool of 10 control individuals was used as a reference sample. Hybridization results were extracted with the Feature Extraction software v.10.5.1.1 (Agilent Technologies) and analyzed using the DNA Analytics software v.4.0.81 (Agilent Technologies). The data were processed by the ADM-2 algorithm, with the threshold set at 6.0 SD. CNVs were confirmed by quantitative multiplex PCR of short fluorescent fragments (QMPSF) as previously described [Guilmatre et al., 2009]. Sanger sequencing of the PRODH gene was performed on an ABI prism 3100 automated sequencer (Applied Biosystems, Foster City, CA) using previously described primers [Raux et al., 2007].
ASD Sample 1
The subjects analyzed by the international consortium Autism Genome Project (AGP) included 2,446 ASD unrelated cases (2,114 males and 332 females) and 4,768 controls (2,022 males and 2,746 females) [Pinto et al., 2014]. Affected individuals were assessed using the Autism Diagnostic Interview-Revised (ADI-R) and/or the Autism Diagnostic Observation Schedule (ADOS). Unrelated adult control subjects were assembled from three studies in which individuals had no obvious psychiatric history: the Study of Addiction Genetics and Environment [Bierut et al., 2010], the Ontario Colorectal Cancer Case-Control Study [Figueiredo et al., 2011], and Health, Aging, and Body Composition [Fox et al., 2012]. The ASD cases and comparison subjects were not matched for age. The Autism Genome Project cases included both children and adults, whereas the control subjects where all adults, reflecting the fact that they were not recruited specifically for the ASD study. Given that this is a genetic association study, the age of the subjects should not have any effect on the results.
DNA samples were derived from whole blood or cell lines. All samples were genotyped using Illumina 1M single or duo microarrays (Illumina, San Diego, CA) and analyzed with the same quality-control procedures and CNV analysis pipeline. CNVs were detected using a multiple-algorithm approach including QuantiSNP [Colella et al., 2007], iPattern [Pinto et al., 2011], and PennCNV [Wang et al., 2007]. CNVs spanning less than five consecutive probes or less than 5 kb, or with low confidence score were excluded. For a detailed description of quality control procedures, see Pinto et al. [2010]. Ancestry was inferred from genotyping data; among the cases, 2,147 (87.8%) were of European ancestry; in the control cohort, 2,640 (55.4%) were of European ancestry. Informed consent was obtained from all participants or their parents, and all procedures followed were in accordance with the ethical standards on human experimentation of the participating sites.
ASD Sample 2
This study included 676 unrelated ASD cases and 1005 controls that passed array quality control. The ASD samples (560 males and 116 females) came from three Canadian sites: Hospital for Sick Children in Toronto, Ontario; McMaster University, Hamilton, Ontario; and Memorial University of Newfoundland, St. John's, Newfoundland [Prasad et al., 2012]. All the cases met the criteria for autism on one or both diagnostic measures—ADI-R and ADOS. CNV analysis was performed on genomic DNA derived from blood or lymphoblastoid cell lines. Data for the control cohort of 1005 DNA samples were provided by Population Diagnostics, Inc. (Melville, NY) [Prasad et al., 2012]. Briefly, the cohort (referred to as PDx controls) comprised reportedly healthy donors of European ancestry (500 females and 505 males, >45 years of age) with no known history of disease, including psychiatric disorders. All donors gave informed consent and were deidentified via a protocol approved by the institutional review board. The DNA samples were all derived from whole blood. All case and control samples were run on the Agilent 1M CGH array (Agilent Technologies, Santa Clara, CA) and used CNV-calling methods as previously described [Prasad et al., 2012]. The reference DNA for cases was a pool of 50 sex-matched European ancestry control samples; for controls, the same sex-matched sample (one female or one male) was used. Ancestry analysis in 615 out of 676 cases with SNP microarray data available indicated that 505 cases were of European ancestry. Of the 676 ASD cases studied with Agilent 1M CGH arrays, 322 were also included in the AGP and were studied with SNP arrays [Pinto et al., 2010; Pinto et al., 2014], thus allowing a direct comparison between the two microarray platforms.
RESULTS
HPI Cohort
Among the 83 HPI subjects, 15 (18%) carried the PRODH/DGCR6 deletion (Table I). In four patients, the deletion was homozygous. CGH array analysis indicated that PRODH deletions and duplications had similar breakpoints, with an approximate size of 120 kb (Fig. S2). The CNV is immediately preceded by a 237 kb-long segmental duplication (also termed low-copy repeat, LCR) region, referred to as LCR22-2, which corresponds to the proximal breakpoint of the classical VCFS deletions (Fig. S1). Of note, the presence of this LCR is characterized by a gap in probe coverage on the CGH array, which makes the exact determination of the breakpoint uncertain. The clinical phenotype of HPI subjects included ASD in 7/15 subjects (Table I). In the control group (Table II), only one subject carried the PRODH/DGCR6 deletion. His plasma proline level (337 μmol/L) was at the upper limit of normal values and his complete PRODH genotype was Del/Q19P + T275N. The frequency of deletion carriers (either at the homozygous or heterozygous state) significantly differed between cases and controls (P = 8.6 × 10−9, Fisher exact test), and carrying at least one deleted allele was a strong risk factor for HPI (OR = 50.7; 95%CI = 7.5–2,147).
| Subject | Age (y) | Sex | Clinical features | Plasma proline (μmol/l) | Genotype |
|---|---|---|---|---|---|
| 144 | 42 | F | ID, SZ AFF | 538 | del/R453C |
| 45435 | 31 | F | ASD | 487 | del/R185W + Q19P |
| TF1 | 4 | M | ID, Epi | 2246 | del/del |
| BP1 | 39 | F | BP | 470 | del/P406L |
| 001 | 2 | F | ID, ASD, Epi | 1515–2186 | del/del |
| 117 | 11 | F | ASD | 488 | del/T275N + V427M |
| Si30 | 5 | M | ID, ASD, Epi | 422–1883 | del/R453C |
| 12363 | 10 | M | ID | 512 | del/R185W |
| 107 | 14 | F | LD | 692–878 | del/del |
| 105 | 7 | M | ID, ASD | 599–899 | del/T275N |
| 201a | 13 | M | ID, ASD, Epi | 1008 | del/del |
| 202a | 7 | F | Behav D, Epi | 1700 | del/R453C + T466M |
| 203a | 13 | F | ID, Behav D | 411 | del/R185W |
| 204a | 15 | F | ID, Epi | 556 | del/Q19P + R431H |
| 205a | 11 | M | ASD | 481 | del/R453C + R185W |
- a Previously unreported. ASD, autism spectrum disorder; Behav D, behavioral disorder (e.g., aggression, hyperactivity); BP, bipolar disorder; Epi, epilepsy; F, female; ID, intellectual disability; LD, learning disability; M, male; SZ AFF, schizoaffective disorder.
| Study | Ancestry | Cases | Controls | P-valuea | Odds ratio (95%CI) | Array platform |
|---|---|---|---|---|---|---|
| HPI | All ancestries | 15/83 (18%) | 1/236 (0.4%) | P = 8.6 × 10-9 | OR = 50.7 (7.5–2,147) | Agilent 1M CGH array |
| European | 12/78 (15%) | 1/236 (0.4%) | P = 2.9 × 10-7 | OR = 42.2 (6.0–1,815) | ||
| Other | 3/5 | — | ||||
| ASD 1 (AGP) | All ancestries | 11/2,446 (0.4%) | 15/4,768 (0.3%) | P = 0.4 | OR = 1.4 (0.6–3.3) | Illumina 1M SNP array |
| European | 9/2,147 (0.4%) | 12/2,640 (0.5%) | P = 1 | OR = 0.92 (0.3–2.39) | ||
| Other | 2/299 (0.7%) | 3/2,128 (0.1%) | ||||
| ASD 2 | All ancestries | 6/676b, c (354) (0.9%) | 9/1005 (0.9%) | P = 1 | OR = 0.99 (0.3–3.2) | Agilent 1M CGH array |
| European | 4/505 (0.8%) | 9/1005 (0.9%) | P = 1 | OR = 0.88 (0.2–3.2) | ||
| Other | 1/110 (0.9%) | — |
- a Fisher exact test.
- b Among the 676 ASD subjects from Canada tested with CGH array, 322 were also tested with Illumina 1M SNP array as part of the AGP study, leaving 354 unique patients. Among these 354 patients, there were 4 deletion carriers (1.1%).
- c Ancestry was determined in 615 ASD cases that were genotyped previously with SNP microarrays; no ancestry information was available for the remaining 61 subjects.
ASD Samples
In ASD study 1 (AGP), the SNP array revealed that 11/2,446 ASD subjects carried the deletion (10 males, 1 female), compared to 15/4,768 control individuals (8 males, 7 females) (P = 0.4, OR = 0.6–3.3) (Table II). Two ASD cases with a larger deletion corresponding to VCFS were not included in these counts. In ASD study 2, array CGH analysis identified 6 PRODH/DGCR6 deletion carriers among the 676 ASD individuals (4 deletions in 354 unique individuals not included in the AGP sample), and 9 deletion carriers among the 1005 control individuals (P = 1; OR = 0.3–3.2, Fisher exact test). Analyses restricted to European cases and controls yielded similar negative findings (Table II). Genotype calls were concordant among the 322 ASD patients assessed on both Agilent 1M CGH and SNP array platforms. The two deletion carriers present in this subsample were found in both analyses, indicating that the two microarray platforms seem equally suitable to detect the PRODH/DGCR6 deletion.
There were 45 duplications in the AGP cases (1.8%, 41 males, 4 females; 41 in Europeans)—excluding 5 duplications of the VCFS region—and 102 in controls (2.1%, 45 males, 57 females; 57 in Europeans, 11 mixed ethnicity, 34 other ethnicity). The overall frequency of the duplication in controls and ASD patients was consistent with that previously reported (2.5%) [Guilmatre et al., 2010], and is also comparable to what was observed in the PDx control cohort (18 duplications corresponding to 1.8% frequency, 7 males, 11 females; all European).
DISCUSSION
Despite a possible role of hyperprolinemia in ASD, and the strong involvement of the PRODH/DGCR6 deletion in the genetic determinism of hyperprolinemia, we did not find any significant association between this deletion and ASD in two large case/control samples.
This negative finding could indicate that the suggested enrichment of ASD cases in HPI patients is spurious and reflects an inclusion bias (i.e., clinical geneticists preferentially refer HPI patients with autistic features for genetic analysis). A limitation of this study concerning HPI patients is that we relied on the clinical description of patients made by the clinician who referred blood samples to our laboratory and that structured interviews allowing a formal diagnosis were rarely used. Alternatively, our results could be explained by the low prevalence of HPI among ASD patients. For example, we recently measured plasma proline level in a consecutive series of 230 ASD patients (unpublished data) and found only one HPI case (95%CI [0.028–0.0002]). Thus, only approximately one HPI in 230 ASD cases might contribute to increase the load of PRODH/DGCR6 deletions compared to the baseline. Based on an estimated frequency of HPI among ASD subjects of 1/230 and assuming that only 18% of these HPI subjects will be expected to harbor the PRODH/DGCR6 deletion (this study), the deletion frequency would be expected to increase from 0.4% (i.e., the average frequency observed in the two control samples) to 0.48%. A sample size of about 107,000 patients and a similar number of controls would be required to reach an 80% chance of detecting such an association between ASD and the PRODH/DGCR6 deletion at the 5% level (http://www.stat.ubc.ca/∼rollin/stats/ssize/b2.html). At best, if we consider the upper limit of the 95% confidence interval for the frequency of HPI among ASD, 2.8%, the deletion frequency would be expected to increase from 0.4% in controls to 0.9% in ASD cases, which would require a sample size of about 4,000 each. We conclude that the involvement of HPI in the determinism of ASD remains controversial and is not supported by this CNV study. Given the extreme genetic heterogeneity of ASD and the fact that all disease-causing variants identified to date are very rare [Betancur, 2011], the risk conferred by HPI, if any, is likely to be small. To definitely settle this issue there is an urgent need to measure fasting proline levels in large unbiased cohorts of ASD subjects and controls.
ACKNOWLEDGMENTS
The authors thank the study participants, the physicians who referred patients and the Autism Genome Project (AGP) Consortium.
