Several methods exist for genotyping class II DLA gene polymorphisms in the dog. The most accurate method is sequence-based typing, which involves direct sequencing of polymerase chain reaction products. However, this method is expensive and unsuitable for large-scale studies. Recently, reference strand-mediated conformation analysis (RSCA) has been shown to be effective for characterizing major histocompatibility complex genes in humans, sheep, horse, and cats. RSCA is a cheap and rapid method, ideal for large epidemiological studies. We have developed RSCA for typing DLA-DRB1 in the dog. Control panels including dogs typed by sequence-based typing and cloned major histocompatibility complex class II alleles in plasmids were used to establish migration patterns for each allele using 20 different fluorescent labeled references, of which 5 were selected to allow for clear identification and discrimination of all known DLA-DRB1 alleles. We have compared 168 dogs typed by RSCA for DLA-DRB1 and characterized by sequence-based typing, with less than 1% discrepancy. These differences were due to missing alleles because of a weak polymerase chain reaction. To date, we have RSCA-typed 1,394 dogs. RSCA is likely to become the method of choice for characterizing DLA genes in the dog and will prove a useful tool for dissecting the immune response of dogs in clinical studies.
Major histocompatibility complex gene products are critical for the regulation of effective immune responses in all higher animals. They provide the foundation for self-recognition and, as such, are fundamental elements for allotransplantation and rejection and for explaining variation in immune response to vaccines, susceptibility to infections, and the development of autoimmunity. They also have a role in immune surveillance and tumor immunology.
As evidence accumulates that the major histocompatibility complex is implicated in susceptibility to canine diseases such as polyarthritis (Ollier et al. 2001), leishmaniasis (Quinnell et al. 2003), diabetes (Kennedy LJ, unpublished data), and hypothyroid disease (Happ GM, personal communication; Kennedy LJ, unpublished data), the need for a low-cost, rapid DLA-typing method becomes more acute so that further studies can be conducted.
DLA-typing methods include restriction fragment length polymorphism, which is labor intensive and low resolution (Sarmiento and Storb 1988); sequence-based typing (SBT), which is definitive but expensive (Kennedy et al. 1998); sequence-specific oligonucleotide probing (SSOP), which is low resolution, high throughput, and inexpensive but laborious and time-consuming to analyze (Kennedy et al. 1999); and single-stranded conformational polymorphism followed by DNA sequencing, which is labor intensive, time-consuming, and expensive (Wagner et al. 1998). None of these methods are ideal for typing the large numbers of samples and controls required for disease association and epidemiological studies.
Recently, a novel method, reference strand-mediated conformation analysis (RSCA), has been used to characterize major histocompatibility complex alleles in a variety of species, including human (Arguello et al. 1998a,b; Arguello and Madrigal 1999; Ramon et al. 1998; Turner et al. 1999, 2001), sheep (Feichtlbauer-Huber et al. 2000), cat (Drake et al. 2004; Kennedy et al. 2003), and horse (Brown et al. 2004). This method depends on automated fluorescence-based detection of polymerase chain reaction (PCR) amplified alleles hybridized as heteroduplexes with fluorescently labeled reference (FLR) DNA sequences (Arguello et al. 1998c). RSCA can detect one–base pair differences between two sequences, unlike all the published methods with the exception of SBT. RSCA is a high-resolution, high-throughput, inexpensive method that can be automated, making it less laborious than other methods. RSCA also has the advantage of being able to identify new as well as known alleles.
We have developed RSCA to characterize DLA-DRB1 alleles in the dog and have now RSCA-typed over 1,300 dogs.
Materials and Methods
In total, 1,394 dogs were used in this study. These included dogs from 81 different breeds with a range of 1 to 133 dogs per breed. All EDTA blood samples used were the residual of that taken for clinical purposes only.
DNA was extracted using a variety of methods, including standard phenol-chloroform extraction and several commercial kits, such as DNAeasy (Qiagen). All DNA samples were normalized to a standard concentration at 5 ng/μl.
Some of the samples in the study had been DLA-DRB1 typed using other methods. These were used to form a panel to establish the mobilities of known alleles in RSCA. All together, 168 samples had been typed by SBT (Kennedy et al. 1998) and 78 by SSOP (Kennedy et al. 1999). Many DLA-DRB1 alleles were available as cloned plasmids. These were used to establish allele mobilities and to provide the base material for generating the FLRs.
Amplification of Exon 2 from DLA-DRB1
The primers used were intronic and locus specific: forward 5′-3′ GATCCCCCCGTCCCCACAG and reverse 5′-3′ CGCCCGCTGCGCTCA. The product size was 303 bp.
All PCR reactions were performed with 25 ng DNA in a 25 μl reaction containing 1× PCR buffer as supplied by Qiagen, Q solution (Qiagen), final concentrations of 0.1 μM for each primer, 200 μM for each dNTP, with 2 units of Taq polymerase (Qiagen HotStarTaq).
A standard PCR program was used for all amplifications, which consisted of an initial 15 min at 95°C, 14 touch-down cycles of 95°C for 30 s, followed by 1 min annealing, starting at 62°C and reducing by 0.5°C each cycle and 72°C for 1 min, then 20 cycles of 95°C for 30 s, 55°C for 1 min, 72°C for 1 min plus a final extension at 72°C for 10 min.
The annealing temperature was adjusted for use in a 96-well plate using adhesive plate sealers with a heated lid on the thermal cycler, avoiding the need for an oil overlay. All PCR reactions were performed on a DNA Tetrad Engine (MJ Research). A negative control containing no DNA template was included in each run of amplifications to identify any contamination.
The same PCR protocol was used to amplify DNA from plasmids containing cloned alleles for use as FLRs and as controls in RSCA.
Generation of FLRs
FLRs were generated using a range of DLA-DRB1 alleles from the domestic dog and gray wolf. The FLRs were produced by PCR using cloned alleles as templates and a 5′-FAM22-labeled forward primer. To increase the proportion of the labeled reference strand in the reaction, the primer proportions were altered to 0.5 μM FAM22-labeled forward primer and 0.1 μM reverse unlabeled primer. All other aspects of the PCR reaction remained the same. This single-stranded biased FLR was used to increase the heights of the FLR-allele heteroduplex peaks relative to the homoduplex peaks in subsequent RSCA (data not presented). All the resulting FLRs were diluted 1:30 in water before use in the hybridization reactions.
To form duplexes between test samples and FLRs, 2 μl diluted FLR and 2 μl test sample PCR product were mixed in a 96-well plate and incubated in a thermal cycler at 95°C for 10 min, ramped down to 55°C at 1°C per second, 55°C for 15 min and 4°C for 15 min. The plate was stored at 4°C until required. Subsequently, 8 μl distilled water was added to each hybridization reaction, and then 2 μl was mixed with 4.8 μl water and 0.2 μl Genescan Rox-500 size standards (Applied Biosystems) in a 384-well plate. These samples were run on an ABI 3100 DNA analyzer, using 50 cm capillary arrays, 4% Genescan nondenaturing polymer (Applied Biosystems), and data collected using matrix Dye set D. The conditions were as follows: injection voltage 15 kV, injection time 15 s, run voltage 15 kV, run temperature 30°C. Each run took 35 min. The data were analyzed using software programs Genescan and Genotyper (Applied Biosystems). Genescan was used to assign sizes to each peak, based on the ROX-500 standards. Through Genotyper, allele peaks formed by the control samples were assigned to “bins” for each FLR used. The bins were exported to an in-house program (Martin A, unpublished) that assigned the alleles for each sample.
The latest DLA nomenclature report (Kennedy et al. 2001) lists 52 DLA-DRB1 alleles, with a further 23 having been reported since then, making a total of 75 DLA-DRB1 alleles. Of these, 26 have been found in the gray wolf only; 18 are alleles that are rare in the domestic dog, of which 12 are alleles that we have identified in single dogs only; we have no examples of the other 6 in our DLA-DRB1-characterized population of 2,300 dogs (including the 1,394 typed by RSCA and a further 906 characterized by SBT).
Thus, our target was to identify 31 common DLA-DRB1 alleles using this RSCA system.
Selection of FLRs
Initially, 20 different FLRs were generated from plasmids containing DLA-DRB1 alleles. These included DLA-DRB1 alleles both rare and common in the dog population, alleles with the greatest genetic distance from the majority of alleles (following analysis using phylogenetic trees), and alleles from the gray wolf that we have not as yet seen in any dogs. One drawback of using only dog alleles as FLRs is that for those dogs that carried the alleles used to make the FLRs, comigration of alleles with the reference strand homoduplex resulted in an inability to characterize these alleles. Therefore, we included gray wolf alleles as FLRs because doing so should have avoided comigration of reference strand homoduplex and alleles in the test samples.
The 20 FLRs were tested on a panel of 92 DLA-DRB1-typed dogs. The results were assessed for separation of alleles by each FLR. Figure 1 shows a composite summary for 92 samples tested with one FLR. All the results have been overlaid, and clear segregation of allele peaks can be seen. A set of five FLRs was selected such that each individual FLR gave good segregation of allele peaks, and by combining the results from all five FLRs, it was possible to distinguish 29 of the 31 common dog DLA-DRB1 alleles.
The final set of FLRs chosen included cloned alleles from the gray wolf (DRB1*03101, DRB1*03202) and the dog (DRB1*01201, DRB1*00101, DRB1*024x). All the plasmids containing these alleles were DNA sequenced and shown to be exact matches with the reference sequences in the DLA database (see http://www.ebi.ac.uk/ipd/mhc/dla/index.html), except for DRB1*024x, which has a single mismatch at nucleotide 266. Figure 2 shows the increase in resolution obtained by using multiple FLRs (the total of 31 includes two new alleles identified during this study).
Figure 3 summarizes the RSCA results using the five FLRs for a dog that is homozygous for DRB1*01501 and one that is heterozygous for DRB1*00601/01501. The DRB1*01501 pattern of peaks can be seen in both dogs.
Separation by RSCA of Known Alleles and Newly Identified Alleles
Mobilities were established for all the alleles used in the DLA-DRB1 typed panel for each of the five selected FLRs. RSCA was then carried out on a further 806 samples, and alleles were assigned using in-house software. The data generated from 898 samples comprising 17 separate experiments were analyzed. Each experiment included control samples, both previously tested dog DNA samples and known plasmids, to provide reference data points within each experiment. Analysis showed that the variation between experiments was no greater than the variation within an experiment, and therefore an overall analysis was done on the complete data set. Over 100 samples have been tested more than once in different RSCA experiments, and no discrepant results have been detected.
Initially, we did not have dog DNA samples covering all 31 common DLA-DRB1 alleles, so when new patterns were identified by RSCA, those samples were sequenced. Sometimes the allele identified was revealed to be one of the reported alleles, and sometimes we identified new alleles. Within this set of 898 samples, 31 common alleles were found, plus 10 new alleles (DLA-DRB1*033012, *06801, *07101, *07201, *07301, *07401, *07501, *07601, *07701, *00103), which have been DNA sequenced, submitted to EMBL, and assigned official names by the DLA nomenclature committee (accession numbers AM076483, AM076470, AM076475, AM076476, AM076477, AM076478, AM076479, AM076480, AM076481, AM084832, respectively.)
Table 1 shows a list of alleles identified by two of the FLRs used, plus the number of times that an allele was identified and the mean and standard deviation (SD) for each allele mobility. The SDs varied from 0.06 to 0.93, with the majority being less than 0.50. The SDs for DRB1*015, which is known to have at least two subtypes (currently indistinguishable by RSCA), were higher than the SDs for many other alleles. This suggests that other alleles with generally higher SDs may actually be two closely related alleles that have not yet been identified. The data summarized in this table represent typical variation observed for all FLRs tested (data for all FLRs can be viewed as supplementary data on the journal's website.)
|01901||13||483.29||0.28||Runs with FLR|
|01901||13||483.29||0.28||Runs with FLR|
Dashes (—) indicate not tested; blank cells indicate not available. Boldface indicate examples of the some of the new alleles identified.
None of the dog alleles shown run with this FLR because it is a wolf allele.
For the second FLR shown, DRB1*01901 runs at the same position as the FLR because DRB1*01901 is only 1 bp different from DRB1*01201.
Cannot separate DRB1*00201 and *00202.
It is as yet unknown whether DRB1*010011 and *010012 can be discriminated from each other, as we have no example of the latter allele.
Cannot separate DRB1*01501 and *01502.
Figure 4 shows a plot of migration values for 31 alleles identified within this study with five FLRs. These 31 alleles include 29 of the 31 common dog alleles (because there are two pairs of alleles that cannot currently be distinguished in this assay), plus 2 of the frequent new alleles identified in this study (DRB1*07601 and DRB1*00103). The other 8 new alleles are not shown, because they were identified in only a few animals. Each allele has different patterns of mobility, allowing it to be distinguished from the other alleles shown. New alleles are easily identified by their unique patterns—for example, DRB1*00103, which has a pattern similar to that of DRB1*00102.
In total, 1,394 dogs have been tested by RSCA and assigned alleles using the dedicated software. At present, it is possible for one person to genotype 192 samples for three DLA class II loci per week.
Comparison of RSCA to Other Typing Methods
A comparison of results generated by RSCA, SSOP, and SBT was undertaken. SBT can be considered the gold standard for DLA-DRB1 typing, as it gives a definitive result. In sum, 78 dogs had been typed by both SSOP and SBT, with no discrepancies, although SSOP is a low-resolution typing method and cannot distinguish between the following common allele combinations: DRB1*00101 and *00102; DRB1*00201 and *00202; DRB1*00301 and *00901; DRB1*00401, *010011 and *010012; DRB1*00501 and *02301; DRB1*00801, *00802 and *02901; DRB1*01201 and *01901; DRB1*01501 and *01502; DRB1*01601 and *01801.
In all, 168 dogs were tested by both SBT and RSCA with three observed discrepancies, all due to poor PCR amplification of alleles such that one method indicated a homozygote whereas the other indicated a heterozygote. Missing alleles occurred in both methods and were therefore related to poor PCRs rather than to the typing method. When new alleles occurred in combination with known alleles, RSCA was better at identifying the known allele and could be used for confirmation.
Overall, 96 dogs were tested by both SSOP and RSCA with seven discrepancies in the results; but again, all were due to weak PCRs resulting in one allele being undetected (usually by SSOP, rarely by RSCA) because weak peaks can fairly easily be seen.
However, all of these undetected alleles were suspected to be missing because we usually type for three class II loci—DLA-DRB1, DQA1, and DQB1—and the likelihood of a missing allele at one locus is usually suggested by the presence of two alleles at each of the other loci. Generally, the suspect result was repeated, and the missing allele was identified the second time around. This occasional loss of alleles occurred at all three class II loci, but there was no case when all three alleles on one haplotype were missing in the same individual.
Several families were included in the data, and inheritance of DLA-DRB1 alleles could be followed and confirmed through the pedigrees.
Previous methods for characterizing the alleles of DLA-DRB1 have been either high resolution but expensive (such as SBT) or low resolution and high throughput but labor intensive (such as SSOP). We have developed RSCA and have been able to demonstrate that it is a high-throughput, high-resolution, easily automatable, and relatively cheap method for DLA-DRB1 typing.
RSCA can quickly identify both known and new alleles. In DNA cloning the new alleles, the clones can be screened against the original PCR by RSCA to prevent unnecessary sequencing of identical clones and to eliminate clones that contain PCR errors.
Of the 31 most common DLA-DRB1 alleles identified in dogs, 29 of them were clearly distinguishable. Two pairs of alleles (DRB1*00201 and *00202; DRB1*01501 and *01502) were more difficult to separate. These two alleles are only one nucleotide different from each other near the end of the amplicon. Other pairs of alleles with one–base pair differences, such as DRB1*00101 and *00102, could be distinguished (see Figure 1), but that difference is in the middle of the amplicon. It is possible that DRB1*00202 is actually very rare, and as yet, we do not have many examples of this allele; the SDs for DRB1*002 range from 0.14 to 0.53, suggesting a single allele. However, the SDs for DRB1*015 range from 0.29 to 0.92, suggesting the presence of at least two different alleles. Given that DRB1*015 is one of the most common DRB1 alleles, with an overall frequency of 20.9% (and as high as 65.1% in some breeds), and that it has two common subtypes (DRB1*01501 and *01502), two rare subtypes (DRB1*01503 and *01504), plus at least three more subtypes identified in our own data, it seems likely that this group represents several alleles.
We have identified 10 new alleles by RSCA, which have been DNA sequenced. We have also applied this RSCA method to typing DLA-DRB1 alleles in the gray wolf. It was possible to assign unique patterns to all the wolf alleles that we had identified by SBT. Further, we have shown the efficacy of RSCA as an inexpensive, high-resolution, high-throughput DLA-DRB1-typing method. It is likely that RSCA will become the method of choice for characterizing DLA genes in the dog. This method should prove invaluable for dissecting the regulation of immune response in the dog and for characterizing disease susceptibility to conditions with immune-based etiopathogenesis or variation in response to vaccines. Further, it may have clinical applications in allotransplantation, and it may be useful in breed and phylogenetic studies.
A supplementary table is available at Journal of Heredity online (www.jhered.oxfordjournals.org).
This article was presented at the 2nd International Conference on the “Advances in Canine and Feline Genomics: Comparative Genome Anatomy and Genetic Disease,” Universiteit Utrecht, Utrecht, The Netherlands, October 14–16, 2004.