OBJECTIVES. Despite epidemiologic evidence to the contrary, claims of an association between measles-mumps-rubella vaccination and the development of autism have persisted. Such claims are based primarily on the identification of measles virus nucleic acids in tissues and body fluids by polymerase chain reaction. We sought to determine whether measles virus nucleic acids persist in children with autism spectrum disorder compared with control children.
PATIENTS AND METHODS. Peripheral blood mononuclear cells were isolated from 54 children with autism spectrum disorder and 34 developmentally normal children, and up to 4 real-time polymerase chain reaction assays and 2 nested polymerase chain reaction assays were performed. These assays targeted the nucleoprotein, fusion, and hemagglutinin genes of measles virus using previously published primer pairs with detection by SYBR green I. Our own real-time assay targeted the fusion gene using novel primers and an internal fluorescent probe. Positive reactions were evaluated rigorously, and amplicons were sequenced. Finally, anti-measles antibody titers were measured by enzyme immunoassay.
RESULTS. The real-time assays based on previously published primers gave rise to a large number of positive reactions in both autism spectrum disorder and control samples. Almost all of the positive reactions in these assays were eliminated by evaluation of melting curves and amplicon band size. The amplicons for the remaining positive reactions were cloned and sequenced. No sample from either autism spectrum disorder or control groups was found to contain nucleic acids from any measles virus gene. In the nested polymerase chain reaction and in-house assays, none of the samples yielded positive results. Furthermore, there was no difference in anti-measles antibody titers between the autism and control groups.
INTERPRETATION. There is no evidence of measles virus persistence in the peripheral blood mononuclear cells of children with autism spectrum disorder.
Controversy over measles-mumps-rubella (MMR) vaccine erupted in 1998 when it was suggested that the measles virus (MV) component of the vaccine was responsible for autistic enterocolitis, a new form of autism spectrum disorder (ASD) characterized by the presence of ileo-colonic lymphonodular hyperplasia, chronic inflammatory colonic disease, and loss of acquired cognitive skills after a period of normal development.1 To address this concern, several epidemiologic studies were performed that found no association between MMR vaccine and ASD.2–14 This controversy has been the subject of several authoritative reviews and statements by, among others, the American Academy of Pediatrics and the Institute of Medicine.15–17
Despite the rising tide of epidemiologic evidence against any MMR-ASD link, several polymerase chain reaction (PCR)-based investigations have been published that seem to support the association.18–20 Using nested reverse-transcription (RT)-PCR targeting the MV hemagglutinin and fusion genes to study peripheral blood mononuclear cells (PBMCs), Kawashima et al19 reported the presence of ≥1 MV gene in 2 (18%) of 11 samples from subjects with inflammatory bowel disease (IBD) and 3 (33%) of 9 children with ASD compared with 0 of 8 control samples. By direct sequencing, these authors implicated wild-type virus in 1 of the positive IBD cases and vaccine-strain virus in the remainder. In 2002, both Uhlmann et al18 and Martin et al20 published studies claiming the detection of MV nucleic acids in intestinal samples from children with the alleged new variant ASD. Using real-time TaqMan PCR, Uhlmann et al18 reported the detection of MV fusion and hemagglutinin genes in biopsies from 75 of 91 children with ASD (vs 5 of 70 developmentally normal children). Similarly, Martin et al20 described the detection of MV fusion and hemagglutinin genes in biopsies from 62 of 68 ASD children (vs 4 of 39 biopsies from controls). Both groups reported the detection of MV nucleoprotein gene by in situ PCR (42 of 57 and 25 of 28 ASD biopsies in the Uhlmann et al18 and Martin et al20 studies, respectively, vs 1 of 5 control samples in both cases). No sequence information was provided in either study, despite the apparent ease with which amplicons were generated and the high copy numbers reported (range: 1–300000 copies/ng RNA).
These molecular data have been used to implicate MMR vaccination in the development of ASD in at least a subset of affected children. In effect, it has been argued that these molecular data trump the epidemiologic evidence at the level of the individual child. Unfortunately, the media attention that accompanied the initial claims and the reporting of the ongoing controversy have led to a loss of confidence in MMR vaccine. MMR coverage in the United Kingdom fell from 92% in 1995–1996 to ∼80% in 2004–2005, resulting in several major measles outbreaks and ≥3 deaths.21,22
Although concerns have been expressed regarding the molecular data generated by proponents of the MMR-ASD hypothesis23 and ethical concerns have been raised about one of the reports,24 no one has attempted specifically to confirm or refute the molecular data18–20 to resolve the controversy. We sought to replicate the real-time PCR and nested PCR results reported by Uhlmann et al18 and Kawashima et al,19 respectively, by applying their primer pairs to PBMCs isolated from children with ASD and developmentally normal controls. The principal objective of this study was to determine whether or not the amplicons produced in these assays were MV- specific by sequencing. In addition, we developed an in-house, probe-based PCR assay to detect the MV fusion gene in these same samples. We found no evidence to support the contention that MV persists in any of the PBMC samples from ASD patients or developmentally normal children. Although many samples yielded positive results in the real-time PCR assays based on the Uhlmann et al18 primers, none of the amplified products were of MV origin by sequencing. No sample was found to be positive using the Kawashima et al19 assays or our probe-based fusion gene real-time RT-PCR assay. Finally, we found no significant difference in anti-MV antibody titers in ASD subjects and age-matched controls.
Anti-coagulated blood (2–3 mL with ethylenediaminetetraacetic acid) was collected from 54 sequential children with ASD (autism, 41; pervasive developmental disorder not otherwise specified, 13; mean: 4 years old; 48 boys) referred to the Autism Spectrum Disorders Program and 34 control children (mean: 4 years old; 26 boys) seen in the outpatient pediatric clinic of the Montreal Children’s Hospital. Children with ASD were assessed by a standardized protocol involving the administration of the Autism Diagnostic Interview-Revised25 to the main caregiver and a direct examination of the child with the Autism Diagnostic Observational Schedule Generic.26 Both measures were administered by trained staff. The children were all seen by a clinical expert in autism (E.F.) who formulated the final diagnosis using Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, criteria.27 Controls with normal development were selected among children attending an outpatient clinic of the same hospital who were to undergo blood testing for other reasons. The absence of ASD in the controls was verified by parental interview, review of medical charts, and the administration of the Autism Screening Questionnaire.28 Parents of children from both groups were requested to provide the vaccination booklet of the child that was used to ascertain exposure to MMR and date of immunization(s). Intestinal tissue was not sought for this study, because it was felt that endoscopy with biopsies could not be ethically justified in this patient population. Furthermore, the observation by Kawashima et al19 that MV nucleic acids could be identified in the blood of 33% of children with autistic enterocolitis suggested that the use of PBMCs would be adequate to address the principal hypothesis. The operator (Y.D.) was blinded to study group (ASD or control) while performing RNA manipulations and the in-house fusion gene and Uhlmann assays.
PBMCs were isolated by Ficoll-Plus (GE Healthcare, Buckinghamshire, United Kingdom) density gradient centrifugation as described previously.29 Briefly, whole blood was centrifuged (300g for 10 minutes), and plasma was stored at −20°C until used for ELISA testing. The remaining blood cells were resuspended in Hank’s buffered salt solution (HBSS, Wisent, St-Bruno, Quebec, Canada) to a total of 10 mL and layered onto 4 mL of Ficoll-Plus. After centrifugation (1020g for 30 minutes), the PBMC interface was removed and washed twice in 10 mL of HBSS (300g for 10 minutes). The PBMCs were counted by hemocytometer before the second wash. The final cell pellet was resuspended in 2 mL of bovine serum albumin (Sigma, Oakville, Ontario, Canada) containing 7.5% dimethyl sulfoxide (American Chemicals Ltd, St Laurent, Quebec, Canada) and frozen in liquid nitrogen until it was used for RNA extraction.
Positive and Negative Control Samples
PBMCs from 2 laboratory volunteers were cultured in RPMI 1640 (Wisent) containing 15% fetal bovine serum (Wisent) and supplemented with Hepes (0.01M final concentration: Wisent), glutamine (2 mM final concentration: Wisent), gentamicin (0.1 mg/mL final concentration: Wisent), and penicillin/streptomycin (100 IU/100 μg/mL final concentrations, respectively: Wisent). Cells (1 × 106) were stimulated with 1 μg/mL phytohemagglutinin (Sigma) for 1 hour at 37°C, followed by infection with the Edmonston strain of MV (gift from S. Ratnam, Newfoundland Public Health Laboratory, Newfoundland, Canada) at a multiplicity of infection of 1. Uninfected PBMCs isolated from the same volunteers served as negative controls. These samples were used to optimize assay conditions.
RNA Extraction and Measurement
PBMCs were thawed and washed in 10 mL of HBSS (300g for 10 minutes), and subsequent RNA isolation was conducted with the RNeasy kit according to the manufacturer’s instructions (Qiagen, Valencia, CA), with terminal digestion of residual DNA using the ribonuclease-free deoxyribonuclease kit (Qiagen). Total RNA was eluted in 30 μL of distilled water, and its concentration was determined by fluorescence (Ribogreen, Invitrogen, Burlington, Ontario, Canada). Integrity of the harvested RNA was assessed on 1% ribonuclease-free agarose gels. It is noteworthy that Kawashima et al19 extracted total RNA from PBMCs by a similar method, also based on guanidine salts but using a specific RNA-binding resin instead of a silica-based membrane.
Name and direction, GC content, melting temperature, position, sequence, gene-amplified and expected size of band for the in-house primer-probe set, as well as the Uhlmann and Kawashima primers are shown in Table 1. The housekeeping gene used by Uhlmann et al18 and in our in-house assay was human glyceraldehyde-3-phosphate dehydrogenase (GAPDH). This constitutively expressed gene serves as an internal control for RNA adequacy, because it is subject to the same errors in RNA extraction and preparation as the gene of interest. We had originally planned to include in-house probe-based assays for both the nucleoprotein and fusion genes. However, during optimization studies, our nucleoprotein assay proved to have a lower sensitivity limit of 10E4 copies per reaction and was abandoned. To some extent, this decision was dictated by the number of assays that we wished to perform and the limited amount of RNA available.
Probe-Based, In-House, Real-Time Fusion Gene Assay
Oligonucleotide primers and probes were designed and synthesized according to the Qiagen guidelines for custom real-time PCR primer and probe design (https://customassays.qiagen.com/design/inputsequences.asp).30 The oligonucleotide probe was labeled with fluoro-amidocarboxymethyldethia coenzyme A, a fluorescent reporter dye. All of the positions in the primers and probe correspond with the Edmonston, vaccine-strain MV sequence reported in the National Center for Biotechnology Information nucleotide sequence database, accession No. AF266288. The in-house primers and probe contain superbases, which are proprietary, modified nucleotides that permit highly specific binding to GC-rich DNA templates. The sequences for the GAPDH primers and probe are proprietary information of Qiagen. The cycling conditions for controls, samples, and standards were as follows: reverse transcription: 50°C for 20 minutes; activation: 95°C for 15 minutes; PCR conditions: 42 cycles of 94°C for 0 seconds; 56°C for 30 seconds; and 76°C for 30 seconds at a ramp rate of 2°C/second (Lightcycler, Roche Applied Science, Laval, Quebec, Canada). Each reaction tube contained: 10 μL of master mix, 0.2 μL of RT, 1 μL of probe, 1 μL of combined forward and reverse primers, 4.8 μL of distilled water, 1 μL of uracyl DNA-glycosylase (UNG) and 2 μL of RNA (2–332 ng). All of the sample PCR reactions were performed in duplicate. A third aliquot of each sample RNA was spiked with a known amount of MV RNA and run in parallel to guard against false-negative results (QuantiTect Probe RT-PCR kit, Qiagen). The expected amplicon size was 111 base pairs (bp).
Primers and Cycling Conditions for Uhlmann PCR Assays
Nucleoprotein 1, fusion 1, and hemagglutinin 1 primer sequences (Integrated DNA Technologies, Coralville, IA) were as described by Uhlmann et al18 (Table 1). Cycling conditions were as follows: RT: 50°C for 20 minutes; activation: 95°C for 15 minutes; and PCR conditions: 94°C for 15 seconds, 56°C for 30 seconds, and 72°C for 30 seconds, and a fourth segment of 5 seconds was added to each cycle depending on the targeted gene (nucleoprotein: 78°C; hemagglutinin: 79°C; fusion: 74°C) to minimize the amplification of nonspecific products (Lightcycler, Roche Applied Science). Each reaction tube contained 10 μL of master mix, 0.2 μL of RT, 2 μL (0.5 μM) each of forward and reverse primers, 3.8 μL of distilled water (or 2.8 μL of distilled water and 1 μL of UNG for the Uhlmann fusion and hemagglutinin assays), and 2 μL of RNA. In all but 1 case, >5 ng of sample RNA were used (range: 5–332 ng) per reaction. All of the sample reactions were performed in duplicate. An MV-spiked aliquot was run in parallel for each sample (QuantiTect SYBR green RT-PCR kit, Qiagen). Expected amplicon sizes were 150 bp for each of the Uhlmann MV assays and 226 bp for the Uhlmann GAPDH primers. The Uhlmann et al18 study reported the use of 2 primer sets for each fusion, nucleoprotein, and hemagglutinin assay (ie, 6 primer sets total). However, it is unclear which and how many of their samples were positive using which primer sets. The F1/F2 and N1/N2 primer sets partially overlap one another; we, therefore, arbitrarily chose the first sets for inclusion in our study (F1 and N1). The H2 primer set performed poorly during optimization in our hands and was omitted from further study.
Primers and Cycling Conditions for Kawashima Nested PCR Assays
Primer sequences (Integrated DNA Technologies) were as described by Kawashima et al19 (Table 1). We modified their protocol slightly by omitting the initial independent complementary DNA (cDNA) synthesis step. In addition, instead of conventional PCR, we used real-time PCR with detection by SYBR green I for both PCR rounds. Cycling conditions for the first round of nested RT-PCR were as follows: RT: 50°C for 20 minutes; activation: 95°C for 15 minutes; PCR conditions: 94°C for 15 seconds, 56°C for 30 seconds, and 72°C for 30 seconds; melting conditions: 95°C for 0 seconds, 65°C for 30 seconds, and 95°C for 0 seconds at a ramp rate of 0.1°C/second (Lightcycler, Roche Applied Science). Each reaction tube contained 10 μL of master mix, 0.2 μL of RT, 2 μL (0.5 μM) of forward and reverse outer primers, 4.2 μL of distilled water, 0.5 μL of UNG, and 2 μL of RNA. In all but 1 ASD and 1 control case, >3 ng of sample RNA were used (range: 3–200 ng) per reaction. All sample reactions were performed in duplicate (QuantiTect SYBR green RT-PCR kit, Qiagen). Expected amplicon sizes were 834 bp for the fusion assay and 595 bp for the hemagglutinin assay. Cycling conditions for the second round were as follows: activation: 50°C for 15 minutes; PCR conditions: 94°C for 15 seconds, 56°C for 30 seconds, and 72°C for 30 seconds; and melting conditions: 95°C for 0 seconds, 65°C for 30 seconds, and 95°C for 0 seconds at a ramp rate of 0.1°C/second. Each reaction tube contained 10 μL of master mix, 2 μL (0.5 μM) of forward and reverse inner primers, 4 μL of distilled water, and 2 μL of DNA from the first round (QuantiTect SYBR green PCR kit, Qiagen). Expected amplicon sizes were 181 bp for the fusion assay and 332 bp for the hemagglutinin assay.
External Standard Curves
The standards were synthesized in vitro from DNA segments amplified by PCR from plasmids containing full-length, Edmonston vaccine-strain MV nucleoprotein, fusion, and hemagglutinin genes kindly provided by Dr Alex Valsamakis (the Johns Hopkins Hospital, Baltimore, MD). Briefly, DNA transcripts containing a T3 polymerase site were generated for each targeted sequence. Using real-time PCR amplification (Lightcycler FastStart DNA Master SYBR green I kit, Roche Applied Science), the in-house forward primer and all 3 of the Uhlmann forward primers were modified to contain the T3 polymerase recognition sequence AATTAACCCTCACTAAAGGGACT. The resulting PCR products were pooled and purified on 3% agarose gels. Bands of the correct size were excised, and the products were extracted using the QIAquick gel extraction kit according to the manufacturer’s instructions (Qiagen). The extracted DNA products were then in vitro transcribed using the Megascript kit (Ambion, Austin, TX) to generate the corresponding RNA standard templates. Transcription reactions were conducted for 14 to 16 hours at 37°C. Any residual DNA was removed using 2 μL of Turbo deoxyribonuclease (Ambion) for 30 minutes at 37°C, followed by a further 2-minute incubation at room temperature with 4 μL of deoxyribonuclease inactivation reagent (Ambion). The supernatant was removed and washed using the MegaClear kit according to the manufacturer’s instructions (Ambion). Final standard template RNA concentrations were determined by fluorescence (Ribogreen, Invitrogen) using the Lightcycler (Roche Applied Science). An aliquot of each of the final purified RNA templates was separated by electrophoresis on 1% ribonuclease-free agarose gels to verify purity and size. Standard curves were generated for each assay using 10-fold serially diluted RNA standards to quantify MV gene copy numbers detected over a range of 0 to 107 copies. Carrier transfer RNA (Roche Applied Science) was added to dilutions E4 to E0 at a concentration of 10 ng/μL to prevent the standard RNA from adhering to nonspecific sites. Results of all of the assays are reported in both qualitative (eg, positive/negative based on the presence of any fluorescence peak at <41 cycles) and quantitative terms (eg, estimated copy number per microliter based on extrapolation from the standard curve).
The known susceptibility of PCR to contamination and false-positive results31 was limited by the partial substitution of deoxythimidine triphosphate (dTTP) and deoxyuridine triphosphate (dUTP) in all of the amplifications and pretreatment of samples with heat-labile uracil-DNA glycosylase ([UNG] Roche Applied Science). UNG catalyzes the hydrolysis of the N-glycosylic bond between the uracil and sugar, leaving an apyrimidinic site in uracil-containing single- or double-stranded DNA. This approach can markedly reduce carry-over contamination in PCR by removing uracil incorporated into amplicons, thereby blocking reamplification. Although Uhlmann et al18 used 0.01 U of UNG per reaction, it has subsequently been shown that even 0.5 U per 25 μL PCR mix can sometimes fail to degrade ≤250 copies of contaminating DNA.32 We, therefore, incubated each test sample with between 0.5 and 1 U of UNG per 20 μL of PCR mix for 10 minutes at 20°C before PCR for the in-house fusion, Uhlmann (fusion and hemagglutinin), and Kawashima (fusion and hemagglutinin) assays. UNG was inactivated during the RT step (50°C for 20 minutes). Also, DNA Zap (Ambion) was applied to the flow hood, capillary holder, all of the pipettes, tip boxes, and tweezers before each PCR work session.
We closely monitored positive reactions in our nontemplate and uninfected PBMC controls for any sign of contamination. In addition, all of the samples that flagged positive were routinely rerun in triplicate or quadruplicate to confirm the result before proceeding with further analysis (ie, gel electrophoresis, melt curve inspection, and, ultimately, sequencing). As noted below, we modified the Kawashima assays during optimization studies to eliminate the independent first-strand cDNA synthesis. This change both increased efficiency and reduced the potential for contamination. In addition, we monitored positive reactions by melting curve analysis after the first and second rounds of nested PCR to screen for contamination.
Protocol for Samples That Flagged Positive Using Uhlmann Assays
Samples that flagged positive were analyzed by melting curve analysis. In the event that the melting peak of a sample reaction corresponded with the melting peak of the positive control, PCR product was pooled and separated on a 3% agarose gel at 55 V for 3 hours. In the case of a band of the expected amplicon size, this band was excised and extracted using the QIAquick gel extraction kit (Qiagen). This product was then subject to 2 sets of analyses. First, we attempted direct di-deoxy sequencing of the amplification products, using the primers that generated the products. Second, PCR products with the correct melting temperature and band size were also subcloned into pCR4Blunt-TOPO overnight at 16°C and used to transform Escherichia coli TOP 10 competent cells using the Zero Blunt TOPO PCR Cloning Kit for Sequencing (Invitrogen). Transformants were detected according to the manufacturer’s protocol. Plasmids were prepared using the PureLink HiPure plasmid DNA purification kit (Invitrogen). Each resulting plasmid preparation was digested with 1 unit EcoRI (New England Biolabs, Ipswich, MA) at 37°C for 1 hour then separated on 1% agarose gels at 110 V for 1 hour. Digests containing a band of 4000 bp (expected plasmid vector size) were sent for sequencing using T3 primer (McGill University Genome Center, Montreal, Quebec, Canada). Mega3 software33 was used for computer-based analyses of nucleotide sequences. The Basic Local Alignment Search Tool (BLAST; National Center for Biotechnology Information)34 was used to find regions of similarity between amplified sequences and all sequences deposited in databases. The program compares these nucleotide sequences and calculates the statistical significance of matches.
Protocol for Samples That Flagged Positive Using Kawashima Assays
Positive samples were first analyzed by melting curve analysis after both the first and second round of nested PCR. Next, 5 μL of first- and second-round PCR product were separated by electrophoresis on 1% agarose gels for 2 hours at 80 to 90 V. In the event that a second-round product had the expected band size, this sample product was pooled and separated on a 1% agarose gel for 3 hours at 55 V, then excised and gel extracted using the QIAquick gel extraction kit (Qiagen). Samples were sent for direct di-deoxy sequencing using the primers that generated the products (McGill University Genome Center). Mega3 software33 was used for computer-based analyses of nucleotide sequences. BLAST34 was used to search the nucleotide database for sequence similarities.
Anti-Measles Immunoglobulin G ELISA
Singh and colleagues35–37 have reported higher anti-MV antibodies in ASD children (vs normal children), arguing that this difference supports the presence of a persistent virus. Briefly, 96-well microtiter plates were coated with 1 μg/mL of MV antigen in carbonate-bicarbonate buffer (pH 9.6) overnight at 4°C. Plates were washed 3 times with phosphate-buffered saline-Tween 0.05% (PBS-T) buffer (pH 7.4) and blocked with phosphate-buffered saline-5% nonfat milk at 37°C for 1 hour (300 μL/well). The blocking buffer was removed and standard (NIBSC [National Institute for Biological Standards and Control, Potters Bar, United Kingdom] 5 IU/mL: twofold serial dilutions) or test plasma (diluted 1:200) were added (100 μL/well) and incubated at 37°C for 2 hours. After washing 3 times with PBS-T, assays were completed with goat anti-human-immunoglobulin G-alkaline phosphatase (100 μL/well at 1:5000: Sigma) at 37°C for 30 minutes, further PBS-T washes, and 100 μL/well of TM Blue solution (1 mg/mL p-nitrophenylphosphate in 50 mM sodium bicarbonate buffer [pH 9.6] containing 1 mM magnesium chloride; Serologicals Corp, Norcross, GA). The color reaction was stopped with 50 μL/well of 0.5 mol/L sulfuric acid (Sigma) and was read at 405 nm (Microplate Reader model 3550: Bio-Rad, Richmond, CA). Antibody concentrations in test plasma were estimated by extrapolation from the standards included on each plate and are reported as international units.
Approval for this study was obtained from the Montreal Children’s Hospital Research Ethics Board. Fully informed, written consent was obtained from parents or legal guardians of all of the patients, including controls.
Fifty-one of 54 children from the ASD group and 31 of 34 from the control group had received MMR vaccine. From the ASD group, 2 children had not been vaccinated, and 1 child’s status was unknown. From the control group, vaccination status was unknown for 3 children. The majority of children with ASD (n = 43; 84%) and the developmentally normal children (n = 28; 90%) had received 2 doses. The Quebec immunization schedule recommends a 2-dose MMR schedule at ages 12 and 18 months.14 Only 1 (control) child had received 3 doses. Gastrointestinal symptoms, including constipation, cramping, diarrhea, and abnormal stools, were reported in 43 (79.6%) of 54 children from the ASD group and 11 (32.3%) of 34 children from the control group. Regression was reported in 20 (37%) of 54 subjects in the ASD group. The time between vaccination and blood analysis was similar between the 2 groups. The mean time (±SD) between vaccination and blood sampling for the ASD group was 30.4 (±15.9) months and 26.6 (±15.3) months for the control group (2-tailed t test, t = 1.056, not significant).
Extracted RNA was intact in all of the samples as shown by the presence of discrete 28S and 18S bands on 1% ribonuclease-free agarose gels. The mean amounts of RNA (±SD) for the ASD and control groups were 47.4 (±38.2) ng/μL and 18.70 (± 30.4) ng/μL, respectively. The difference in extracted RNA amounts is most likely attributable to lower PBMC numbers in the control group. The mean cell counts (±SD) for the ASD and control groups were 3.2 × 106 (±2.5) and 1.5 × 106 (±1.2) cells/mL, respectively. GAPDH was detected in all of the samples by both the Uhlmann and the in-house GAPDH primers. Parallel aliquots spiked with MV RNA amplified in all of the samples, demonstrating the absence of PCR inhibitors.
Optimization of Standard Curves and Assays
Reagent optimization was relatively straightforward, because the SYBR green and probe kits have been preoptimized for use with the Lightcycler. Nonetheless, we verified the band sizes of all of the standards and positive controls on gels. All of the amplified products were of the appropriate size. Based on amplification of the specific RNA templates produced for each real-time PCR assay, the detection limits were 36.5, 35.5, 24.0, and 17.5 copies/μL for the Uhlmann nucleoprotein, fusion, and hemagglutinin gene assays and our in-house fusion gene assay, respectively. Slopes for the corresponding standard curves of the optimized assays ranged from −3.505 to −3.201 with r values between 0.99 and 1.00, indicating high efficiency of the PCR reactions. All of the assays yielded strong positive results using the positive control RNA harvested from infected PBMCs, and amplicons from the positive controls generated sequences that were 100% identical to the deposited Edmonston MV sequence (accession AF266288). Although there was considerable nonspecific amplification from the negative control RNA (ie, uninfected PBMCs) in all 3 of the Uhlmann primer-based assays, the melting peaks of these amplification products were generally distinct from those from the positive control RNA. Independent first-strand cDNA synthesis was omitted from the Kawashima protocol, because MV RNA from the positive control was not consistently detected during optimization studies. Instead, we combined first-strand cDNA synthesis and the first round of PCR to exploit the highly sensitive RTs contained in the SYBR green kit. In addition, in our hands, the Kawashima fusion primers did not produce visible amplification products of the correct size from the positive control RNA after the first round of PCR. However a product was detected by the Lightcycler with an appropriate melting curve. Both the fusion and hemagglutinin Kawashima assays yielded strong bands on gels after second-round PCR using the positive control RNA.
Uhlmann Primer-Based Assays
Almost all of the PBMC samples yielded positive signals in all of the assays based on the Uhlmann primer pairs (see Fig 1). Because of the high number of positive results using the Uhlmann primers, PCR was not routinely repeated. Using the Uhlmann nucleoprotein primers, 37 (100%) of 37 of the ASD samples and 18 (100%) of 18 of the samples from developmentally normal controls were positive. The Uhlmann fusion gene primers yielded 93% positive results in the ASD samples (39 of 42) and 100% of controls (17 of 17). The Uhlmann hemagglutinin assay yielded 100% positive results in the ASD samples (40 of 40) and 100% of controls (13 of 13). On inspection of the melting curves, however, it was clear that only a proportion of PCR products had melting peaks that overlapped with the melting peak of the positive control. For example, only 17 (46%) of 37 of the positive ASD results in the nucleoprotein gene assay had the appropriate melting peak, whereas 6 (33%) of 18 of controls had the appropriate melting peak. Melting peaks were not always reproduced in duplicate samples. Nevertheless, these assays generated copy numbers. The mean copy numbers (±SD) using the Uhlmann nucleoprotein assay were 5.66E4 (±4.41E4) for the ASD group and 2.03E4 (±1.28E4) for the control group. For the Uhlmann fusion assay, 8 (20.5%) of 39 of the positive ASD results and 3 (17.6%) of 17 controls had the appropriate melting peak (mean assigned copy number [±SD]: 1.47E3 [±1.17E3] for the ASD group and 5.66E2 [±5.66E2] for the control group). Finally, 11 (27.5%) of 40 of the positive ASD results and 1 of 13 (7.7%) of the controls had the appropriate melting peak for the Uhlmann hemagglutinin assay (mean assigned copy number [±SD]: 3.81E4 [±5.10E4] for the ASD group and 2.16E3 [±7.25E3] for the control group). Almost all of the remaining positive results were eliminated by analysis of amplicon size on 3% agarose gels. The Uhlmann assays gave rise to PCR products of ∼100 bp or less in most cases, demonstrating their lack of specificity (Fig 2). However, 2 of the nucleoprotein gene assay amplicons, 3 of the fusion gene assay amplicons, and 4 of the hemagglutinin assay amplicons (all from ASD subjects) were of the appropriate size (150 bp; Fig 3). On direct sequencing, none of these amplification products generated interpretable sequence data. Therefore, we attempted to clone these amplicons for sequencing. Seven of the 9 appropriately sized amplicons were successfully cloned and sequenced, but none of these revealed MV RNA. BLAST searches instead revealed a number of different mammalian genes, suggesting nonspecific amplification of PBMC RNA.34 It is unclear why 2 of the 9 positive samples could not be cloned; cloning can be particularly difficult with nonspecific amplification products.
Results for the Kawashima Nested PCR Assays
Sufficient sample RNA was available from 23 ASD subjects and 16 control subjects to complete nested PCR testing using the Kawashima primer pairs. No samples were positive using the Kawashima hemagglutinin assay. One ASD sample generated a band of the correct size (181 bp) using the Kawashima fusion primers but was negative by melting curve analysis after first-round PCR, leading us to suspect contamination. This sample was negative when rerun in quadruplicate.
Probe-Based Fusion Gene Assay
The in-house fusion gene probe-based assay generated positive results for all of the positive control samples and internal standards but did not yield any positive results in test samples from either ASD or control PBMCs.
Anti-Measles Antibody Titers
The mean titer (±SD) for the ASD group was 4.3 (±3.9) IU/mL vs 4.6 (±4.0) IU/mL for the control group (2-tailed t test, P = 0.722, not significant). Three outlier samples were removed from this analysis (1 sample from the ASD group at 120.3 IU and 2 samples from the control group at 266.6 and 86.7 IU). Although unvaccinated children (n = 2) and children of unknown status (n = 4) were included in the antibody analysis, their inclusion did not skew the mean titers (data not shown). Anti-measles antibodies were detectable in all of the subjects, regardless of vaccination history.
The main objective of this study was to compare the real-time PCR primers and nested PCR primers used by Uhlmann et al18 and Kawashima et al19 with our own real-time PCR primers to determine whether or not MV nucleic acids are present in the PBMCs of children with ASD and age-matched, developmentally normal controls. The majority of samples generated positive results in the Uhlmann assays. However, melting curve analyses revealed that only a proportion of these PCR products had the correct melting temperature (ie, identical to the positive MV control). Furthermore, only 2 products generated by the Uhlmann nucleoprotein assay, 3 by the Uhlmann fusion assay, and 4 by the Uhlmann hemagglutinin assay had the correct amplicon size when separated by gel electrophoresis. We successfully obtained sequences for 7 of these 9 products, but none were MV RNA. No specimen was positive using the Kawashima primer-based assays or our own probe-based fusion gene assay. We also found no significant difference in anti-MV antibody titers in ASD subjects and age-matched controls. These data do not support the purported association between MMR vaccination and ASD, including the hypothesis that MMR induces regression in a subset of ASD children. These data are consistent with the recently published findings of Afzal et al,38 who found no evidence of persistent MV in leukocyte preparations from MMR-vaccinated children with ASD using TaqMan and nested PCR. Finally, our data reinforce the epidemiologic evidence against such an association.15,39
In the course of this work, we experienced many of the vulnerabilities of using PCR technologies to support a claim of association between an organism and an event. Without meticulous attention to these vulnerabilities, errors in interpretation can occur, as we believe have occurred in the MMR-ASD controversy. First, real-time PCR software will provide copy number information for almost any amplification product whether or not the product is specific. As a result, this technology can provide what seems to be a highly objective measure of the presence of an organism when such an inference is not justified. Uhlmann et al18 claimed that copy number in their positive samples was generally low (ie, typically between 1 and 3×105 MV gene copies/ng total RNA)20 but did not carry their analyses any further. Using the Uhlmann primer pairs, we also obtained copy number estimates between 3.6 and 3.5×103 MV gene copies/ng total RNA (100 and 10000 copies/μL) but demonstrated that these amplicons were not of MV origin by rigorous application of a protocol for melt-curve analysis, amplicon size determination by electrophoresis, and, ultimately, sequencing. There is no evidence that Uhlmann et al18 used a similar protocol for dealing with positive results. A detailed review of the Uhlmann reagents and protocol provides several plausible explanations for their reported results. First, as noted above, Uhlmann et al18 likely used suboptimal concentrations of UNG in their assays, reducing the efficacy of this contamination control strategy. Second, a ClustalW40 search for sequence similarity between the human sequences we amplified and the Uhlmann primers and probes revealed several clustered regions of homology, ranging from 3 to 9 nucleotides (Fig 4, which is published as supporting information on www.pediatrics.org/content/full/118/4/1664). It is plausible that these regions of homology permitted primers, probes, or both to bind, leading to false-positive results. Our data reinforce the principle that great caution must be exercised in interpreting copy number data in the absence of further verification and validation of the amplification products.
Several limitations of our own study deserve mention. Although real-time PCR is regarded by many as the gold standard for the detection of microorganisms in human disease, this technique has significant disadvantages as well. When the hypothesized target nucleic acid is likely to be present in small quantities, the key issue in real-time PCR is guarding against almost inevitable contamination results. Amplification of rare nucleic acids requires rigorous adherence to a set of standard operating procedures, including separate work stations for RNA and DNA handling, decontamination of work materials before every PCR setup, and the optimal use of reagents such as UNG. A single PCR run has the capacity to generate sufficient material to contaminate large areas and all of the subsequent amplifications of the same target sequence. Despite our best efforts, we, like others,41 experienced several episodes of contamination in the course of this work. For example, whenever the nontemplate and negative PBMC controls flagged positive, all of the data from that PCR run were discarded and, if sufficient sample RNA was available, the run was repeated. Had we not anticipated such occurrences, we could easily have been misled. If there is reason to question an experimental PCR result, it is always wisest to repeat the experiment.41 In this context, unintentional contamination is a plausible explanation for Kawashima et al’s19 observation that MV nucleic acids are present in a variety of conditions, including IBD,19 autoimmune hepatitis,42 and epilepsy.43 Moreover, both Uhlmann et al18 and Martin et al20 have reported the presence of MV nucleic acids in a proportion of control subjects.
Second, we took some liberties with the PCR strategies used by Uhlmann et al18 and Kawashima et al.19 For example, Uhlmann et al18 performed TaqMan RT-PCR, which is a probe-based, fluorescent reporting chemistry, using the EZ TaqMan kit and the ABI 7700 Sequence detector from Applied Biosystems. We chose to use SYBR green I, a nonspecific detection chemistry, because we were interested in assessing the performance of the Uhlmann primers without the use of probes. The SYBR green approach allows the user to carry out initial, exploratory screens of different primer sets before using a probe-based protocol. Interestingly, there is a report suggesting that SYBR green I detection is more precise and produces a more linear decay plot than TaqMan detection.44 The use of an internal probe, as in TaqMan, is supposed to provide a higher level of specificity. In theory, nonspecific amplification because of mispriming or primer dimers should not generate a signal in TaqMan assays, because such amplification is ignored by the fluorescence detector. However, absence of detection is not the same as absence of artifacts. Nonspecific artifact amplification can certainly affect amplification efficiency and subsequent quantification. Because of the theoretical specificity of the TaqMan approach, artifacts that interfere with amplification efficiency typically cannot be detected. As a result, many authorities believe that intercalating dyes should be used to optimize primers and reaction conditions before any quantification experiments to ensure the absence of amplification artifacts.45 Our data suggest that the Uhlmann primers are probably not sufficiently efficient to detect MV genes, because they performed suboptimally during our SYBR green I screening. The only other major difference between the TaqMan and the SYBR green approaches is the choice of RT: TaqMan RT requires MnOAc2 to function as a probe-cleaving nuclease and a DNA polymerase, whereas the SYBR green kit contains a combination of Omniscript and Sensiscript RTs optimized to detect a wide range of RNA template amounts. Another difference is the fact that Uhlmann et al18 used their nucleoprotein gene primers to detect the MV nucleoprotein gene by in situ PCR, whereas we used the same primers to detect MV nucleoprotein gene by real-time RT-PCR. Although these primers were not used in precisely the same way, the fact that they resulted in nonspecific amplification in our hands should certainly raise concerns about the in situ data.
A final and obvious difference between our study and those of Uhlmann et al18 and Martin et al20 is the fact that we targeted PBMCs instead of gut biopsy material. As noted above, we did not feel that it was ethically justifiable to subject the children in this study to endoscopy and biopsy despite the fact that almost 80% of the children with ASD had gastrointestinal complaints (vs 32% of our control population). Although there is good evidence that natural MV accumulates mutations when it persists in the central nervous system of subjects with subacute sclerosing panencephalitis,46–49 there is no reason to postulate that the mutation pattern in a virus persisting in the gut would be qualitatively different from that of a virus persisting in the PBMCs. Although such tissue-specific changes in mutation pattern are theoretically possible, it seems highly unlikely that the lack of amplification of MV genes in our study using 6 different primer pairs and several different PCR strategies can be explained in this way.
In the case of the Kawashima assays, we attempted to replicate their results using the same RNA source (PBMCs), a similar method for RNA extraction (guanidine salts with an ethanol wash) and a potentially more sensitive PCR method that combines the strengths of real-time and nested PCR. In addition, by eliminating the first-strand cDNA synthesis step, we were able to eliminate a step where contamination could potentially occur. Despite the similarities between the 2 protocols, we were still unable to identify MV nucleic acids from 23 ASD samples and 16 control samples. These observations stand in sharp contrast to the 33% yield from PBMCs isolated from ASD children reported by Kawashima et al.19 There is no reason to believe that the minor technical differences between our approaches explain these discrepant results.
In light of our findings, it is worth noting that some of the authors who contributed to the Kawashima et al19, Uhlmann et al18, and Martin et al20 articles had earlier reported the detection of MV proteins and nucleic acids in the bowel tissues of subjects with IBD using a variety of techniques.50–52 This claim also provoked considerable media attention and undermined public confidence in the safety of measles-containing vaccine.53 In a series of carefully designed experiments, Iizuka and colleagues54,55 demonstrated that the monoclonal antibody used in many of the early studies supporting the MMR-IBD link (MAS 182r) cross-reacted with an unidentified human protein. A novel monoclonal antibody raised against the positive clone (4F12) was found to react to Crohn disease, ulcerative colitis, and non-IBD colitis control tissues, providing evidence of its nonspecificity. Furthermore by sequencing, Iizuka et al55 demonstrated that the RT-PCR amplicons generated using primers targeting the MV nucleoprotein gene were actually derived from a human rather than a viral gene. The host-derived protein has not yet been identified but may play a role in intestinal inflammation, because neither MAS 182r nor 4F12 antibodies seem to react with control tissues. Our observations raise the possibility that detection of ≥1 host-derived genes may again have been mistakenly interpreted as persisting MV, this time in children with ASD.
Wild-type MV has broad tissue tropism, including the gut, the central nervous system, and the immune system.53 Both wild-type and vaccine-strain MV can persist for prolonged periods of time in some individuals (eg, subacute sclerosing panencephalitis and HIV infected). However, a large burden of proof is required to move from the biological plausibility of an association to establishing that such an association truly exists and is causal. To date, the epidemiologic burden of evidence against such an association in the case of MMR and autism is overwhelming.15,39 We now provide evidence that the PCR data published by Uhlmann et al,18 Martin et al,20 and Kawashima et al19 in support of the more limited claim of an association between MMR and a subset of children with ASD is also unlikely to be true.
Our data, together with the epidemiologic evidence, demonstrate that arguments against vaccinating children with MMR because of fear of ASD are not defensible on scientific grounds. The risk of death and disability from MV infection has been unequivocally demonstrated. The hypothesized link between MMR and ASD is spurious and undermines the success of measles control programs.56
The technical work for this study was funded by the Crohn’s and Colitis Foundation of Canada. Clinical data collection was supported by a grant to Dr Fombonne from the Fonds de Recherche en Santé du Québec.
We thank Rita Zakarian for blood collection and compilation of patient information, Nathalie Martel and Angela Brewer for technical assistance, and Christine Turenne and Marcel Behr for their help with evaluating sequence data and reviewing an earlier version of this article.
- Accepted July 11, 2006.
- Address correspondence to Brian J. Ward, MDCM, Montreal General Hospital, 1650 Cedar Ave, Room D7-153, Montreal, Quebec, Canada H3G 1A4. E-mail:
Financial Disclosure: In the United Kingdom, Dr Fombonne has provided advice on the epidemiology and clinical aspects of autism to scientists advising parents, to MMR vaccine manufacturers (for a fee), and to several government committees between 1998 and 2001. He has been consulted by ad hoc US committees from the Institute of Medicine and the American Academy of Pediatrics reviewing MMR safety. Since June 2004, Dr Fombonne has been an expert witness for vaccine manufacturers in the US thimerosal litigation. None of his research has ever been funded by industry. Dr Ward served on a number of Canadian and US government advisory committees addressing the issues of vaccine use and safety between 1994 and 2006. He has provided expert testimony for both the US and Quebec vaccine injury compensation programs. Dr Ward has also provided advice and teaching to Canadian government and industry groups in the area of vaccine immunology. He has conducted and participated in several studies of measles vaccine safety sponsored by Canadian and US government funding agencies. He has also conducted a small number of phase I and phase II industry-sponsored clinical trials of nonlicensed vaccines for smaller biotechnology companies. He has conducted a single, company-sponsored, immunologic study of a licensed, acellular pertussis vaccine.
Editor’s note: Please read the commentary by Dr Samuel Katz in this issue of Pediatrics.
- Kaye JA, del Mar Melero-Montes M, Jick H. Mumps, measles, and rubella vaccine and the incidence of autism recorded by general practitioners: a time trend analysis. BMJ.2001;322 :460– 463
- DeStefano F, Bhasin TK, Thompson WW, Yeargin-Allsopp M, Boyle C. Age at first mumps-rubella vaccination in children with autism and school-matched control subjects: a population-based study in metropolitan Atlanta. Pediatrics.2004;113 :259– 266
- Fombonne E, Chakrabarti S. No evidence for a new variant of measles-mumps-rubella-induced autism. Pediatrics. 2001;108(4). Available at: www.pediatrics.org/cgi/content/full/108/4/e58
- Makela A, Nuorti JP, Peltola H. Neurologic disorders after measles-mumps-rubella vaccination. Pediatrics.2002;110 :957– 963
- Taylor B, Miller E, Lingam R, et al. Measles, mumps, and rubella vaccination and bowel problems or developmental regression in children with autism: Population study. BMJ.2002;324 :393– 396
- ↵Fombonne E, Zakarian R, Bennett A, et al. Pervasive developmental disorders in Montreal, Quebec: prevalence and links with immunizations. Pediatrics 2006;118(1). Available at: www.pediatrics.org/cgi/content/full/118/1/e139
- ↵Halsey NA, Hyman SL. Conference Writing Panel. Measles-mumps-rubella vaccine and autistic spectrum disorder: report from the New Challenges in Childhood Immunizations Conference convened in Oak Brook, Illinois, June 12–13, 2000. Pediatrics. 2001;107(5). Available at: www.pediatrics.org/cgi/content/full/107/5/e84
- Institute of Medicine of the National Academies. Immunization Safety Review: Vaccines and Autism. Washington, DC: The National Academies Press; 2004.
- ↵Demicheli V, Jefferson T, Rivetti A, et al. Vaccines for measles, mumps and rubella in children. Cochrane Database Syst Rev.2005;(19) :CD004407
- ↵Uhlmann V, Martin CM, Sheils O, et al. Potential viral pathogenic mechanism for new variant inflammatory bowel disease. Mol Pathol.2002;55 :84– 90
- ↵Fitzpatrick M. MMR: Risk, choice, chance. Br Med Bull.2004;69 :143– 153
- ↵Department of Health. Statistical Bulletin: NHS Immunisation Statistics 2004–05. London, United Kingdom: Department of Health; 2005
- ↵Horton R. A statement by the editors of the Lancet. Lancet.1998;363 :820– 821
- ↵American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders - DSM IV. Washington, DC: American Psychiatric Association; 1994
- ↵Kazak Berument SK, Rutter M, Lord C, et al. Autism screening questionnaire: Diagnostic validity. Br J Psychiatry.1999;175 :444– 451
- ↵Qiagen. QuantiTect Gene Expression Assays. Available at www1.qiagen.com/Products/Pcr/QuantiTect/GeneExpressionAssays.aspx. Accessed July 28, 2005
- ↵Kumar S, Tamura K, Nei M. MEGA3: Integrated software for molecular evolutionary genetics analysis and sequence alignment. Brief Bioinform.2004;5 :150– 163
- ↵Altschul SF, Madden TL, Schäffer AA, et al. Gapped BLAST and PSI-BLAST: A new generation of protein database search programs. Nucleic Acids Res.1997;25 :3389– 3402
- ↵Thompson JD, Higgins DG, Gibson TJ. CLUSTALW: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice. Nucleic Acids Res.1994;22 :4673– 4680
- ↵Ghosh S, Armitage E, Wilson D, et al. Detection of persistent measles virus infection in Crohn’s disease: current status of experimental work. Gut.2001;48 :748– 752
- Lewin J, Dhillon AP, Sim R, et al. Persistent measles virus infection of the intestine: confirmation by immunogold electron microscopy. Gut.1995;36 :564– 569
- ↵Daszak P, Purcell M, Lewin J, et al. Detection and comparative analysis of persistent measles virus infection in Crohn’s disease by immunogold electron microscopy. J Clin Pathol.1997;50 :299– 304
- ↵Iizuka M, Chiba M, Yukawa M, et al. Immunohistochemical analysis of the distribution of measles related antigen in the intestinal mucosa in inflammatory bowel disease. Gut.2000;46 :163– 169
- ↵Meissner HC, Strebel PM, Orenstein WA. Measles vaccines and the potential for worldwide eradication of measles. Pediatrics.2004;114 :1065– 1069
- Copyright © 2006 by the American Academy of Pediatrics