Rapid discrimination of MHC class I and killer cell lectin-like receptor allele variants by high-resolution melt analysis

Abstract

Ly49G and H-2 class I D^k molecules are critical to natural killer cell-mediated viral control. To examine their contributions in greater depth, we established NKC/Ly49 congenic strains and a novel genetic model defined by MHC class I D^k disparity in congenic and transgenic mouse strains. Generation and maintenance of Ly49 and H-2 class I select strains requires efficient and reproducible genotyping assays for highly polygenic and polymorphic sequences. Thus, we coupled gene- and allele-specific PCR with high-resolution melt (HRM) analysis to discriminate Ly49g and H-2 class I D and K alleles in select strains and in the F₂ and backcross hybrid offspring of different genetic crosses. We show that HRM typing for these critical immune response genes is fast, accurate and dependable. We further demonstrate that H-2 class I D HRM typing is competent to detect and quantify transgene copy numbers in different mice with distinct genetic backgrounds. Our findings substantiate the utility and practicality of HRM genotyping for highly related genes and alleles, even those belonging to clustered multigene families. Based on these findings, we envision HRM is capable to interrogate and quantify gene- and allele-specific variations due to differential regulation of gene expression.

Human CMV (HHV-5) is prevalent in human populations worldwide and most individuals (≥80%) display HCMV seropositivity by adulthood (Cook and Trgovcich 2011; Mocarski 2002). Typically, primary HCMV infection stimulates robust innate and adaptive immunity, which favors the establishment of viral latency in immunocompetent hosts. Nevertheless, HCMV is an important opportunistic pathogen and the major infectious cause of congenital birth defects (Nigro and Adler 2011; Schleiss 2008). In addition, HCMV causes substantial morbidity and mortality in immunocompromised and immune deficient human patients and it is becoming increasingly evident that viral reactivation from HCMV latency in non-immunosuppressed patients during a critical illness may worsen the clinical course with increased morbidity, and possibly increased mortality (Cook and Trgovcich 2011). Because HCMV, like cytomegaloviruses in all other species, displays species-specific tropism, host resistance and CMV immunity is modeled in other species (Jackson et al. 2011; Lemmermann et al. 2011).

Extensive use has been made of murine (M)CMV to examine antiviral immunity, natural killer (NK) cell-mediated recognition and control of viral infection and the genetics of viral resistance and susceptibility. Classical and forward genetics approaches have uncovered many critical viral resistance factors, including sensors, signal transducers and effectors of immune function (Moresco and Beutler 2011). Prominent resistance factors which affect NK cell sensing of viral infection includes Ly49 and MHC class I molecules (Adam et al. 2006; Brown et al. 2001a; Daniels et al. 2001; Desrosiers et al. 2005; Dighe et al. 2005; Kielczewska et al. 2009; Lee et al. 2001; Xie et al. 2009; Xie et al. 2010). The Ly49H activation receptor targets NK protection at m157-bearing MCMV infected cells in C57BL/6 (B6) mice (Arase et al. 2002; Smith et al. 2002), whereas MHC I D^k, its cognate inhibitory receptor Ly49G, and the Ly49P activation receptor have been implicated in D^k-dependent MCMV control (Desrosiers et al. 2005; Dighe et al. 2005; Kielczewska et al. 2009; Xie et al. 2009; Xie et al. 2010). Our recent findings suggest that MCMV-altered MHC I D^k in infected cells elicits specific Ly49G recognition of viral targets via release of inhibitory receptor signaling, consequent triggering through activation receptors and NK attack. Because Ly49G+ NK cells are licensed by D^k in mouse strains with a C57L or MA/My Ly49G allele expressed, these data further implicate that licensed NK cells may have a vital role in viral immunity (Stadnisky et al. 2011; Xie et al. 2010).

The significance of class I MHC and NK inhibitory receptor polymorphisms in host defense and viral immunity are underscored by genetic studies with the MA/My (H-2^k) strain (Desrosiers et al. 2005; Dighe et al. 2005). We investigated the effect of natural variation on viral resistance in C57L and MA/My mice and then performed additional experiments with MHC I D^k congenic and transgenic strains in C57L and MA/My.L-H2^b (M.H2^b) backgrounds ((Stadnisky et al. 2009), and our unpublished data). These approaches necessitate rapid, high-throughput identification and genotyping methods for class I MHC and NK gene complex (NKC) alleles with only one known SNP to distinguish the relevant Ly49g (Klra7) alleles (Brown et al. 2001b; Xie et al. 2009).

Various strategies amenable to classical genetic analysis and DNA genotyping include RFLP, SSP, microsatellite SSLP and direct DNA sequencing. Each approach has definite advantages in genotyping, precision mapping for DNA crossovers, and positional mapping for genes and alleles of interest. However, these methods are also labor-intensive, time-consuming, microsatellite/DNA sequence dependent, and sometimes error-prone. Further, these methods are not easily adapted to detect gene- and/or allele-specific variations, especially when the target gene or allele resides within highly related multigene clusters. This is an important consideration in large genetic screens, detection of targeted allele replacements, and in routine screening for genetic variations in genetically modified mice when speed, labor and cost are major factors.

A relatively new genotyping strategy is based on PCR amplification of allele variants followed by high-resolution melt (HRM) analysis (Vossen et al. 2009; Wittwer et al. 2003). HRM PCR amplicons incorporate intercalating dyes like EvaGreen that facilitate precision melt analysis during DNA strand dissociation and discernment of short DNA sequences with only a single nucleotide variation. Hence, HRM is directly amenable to gene- and allele-specific PCR genotyping strategies, with little investment in preparatory or analytical time and no requirement for additional probes. Moreover, HRM analysis requires little additional time beyond a typical PCR itself; so altogether, it is a fast, low-cost and reliable genotyping method. Here we examined HRM typing of highly related allele variants belonging to polygenic clusters of genes for decidedly polymorphic class I MHC and NK cell receptor molecules.

High-resolution melt analysis

To examine HRM genotyping for polygenic and polymorphic Ly49 and class I MHC genes, optimal allele sequences (75-200 bp segments) with a single nucleotide polymorphism (SNP) mismatch were first aligned to related gene sequences to ensure primer specificity (Table 1 and Fig 1a). Based on sequence alignments, HRM allele-specific primers (Table 2) were selected to avoid amplification of related genes with high homology. All HRM allele-specific primers were further vetted to avoid additional adverse features (e.g. primer dimers and repetitive sequence elements) and then optimized for PCR on the iCycler CFX96 (Bio-Rad) as described previously (Brown et al. 2001b). Allele-specific PCR was directly followed with HRM analysis and HRM data were then analyzed and clustered using CFX Manager and Precision Melt Analysis software tools (Bio-Rad) and visualized by plotting relative fluorescence units (RFU) and relative difference curves (RFU) against temperature (°C).

Table 1. Ly49 and H-2 class I alleles examined by HRM analysis.

Allele	Intron/Exon	SNP Positiona		SNP
		Mb loca	cM locb
Ly49gc57l	Exon 3	130.168837	63.435	C
Ly49gmamy		130.168837		T
H-2 Kc57l	Intron 3 - Exon 4	34.135617	17.982	A
H-2 Kmamy		34.135627		G
H-2 Kc57l	Intron 7 - Intron 8	34.136642		T
H-2 Kmamy		34.136652		A
H-2 Dc57l	Intron 1	35.39982	18.6093	T
H-2 Dmamy		35.39982		C
H-2 Dc57l	Intron 3	35.40096		G G C T G
H-2 Dmamy		35.400953		A T A C A
H-2 Dc57l	Exon 5	35.402762		A
H-2 Dmamy		35.402744		G
H-2 Dc57l	Intron 8	35.403677		T
H-2 Dmamy		35.403665		G

^aSNP position locations are based on B6 reference strain sequence alignment in build 37. C57L and MA/My SNP positions are estimated relative to the transcriptional start site in B6 based on alignment of the relevant alleles.

^bGenetic (cM) positions are estimated based on a conversion from the physical (Mb) location http://cgd.jax.org/mousemapconverter/

Figure 1.

Establishment of Ly49 and MHC class I gene specific high-resolution melt (HRM) analysis. a) Sequencer (v4.10.1) alignments of Ly49g (Top) and H-2 class I D, K and Qa-1 (Bottom) alleles spanning a single nucleotide polymorphism (SNP) are highlighted (*). Primer target sites (underlined sequences) which are complimentary to both relevant alleles are indicated. Importantly, these target sites were selected to maximize primer disparity, especially at the 3′-most nucleotide, with the related genes and alleles shown in each sequence alignment. All published sequences were obtained from the NCBI. b) and c) Spleen gDNA (MA/My, C57L and MA/My × C57L-F₁) samples were isolated using a kit (Gentra Puregene, QIAgen). Mouse toe and tail gDNA samples were similary prepared. However, toe gDNA samples prepared by rapid alkaline treatment were incompatible with HRM analysis (data not shown). Ly49g- and H-2 class I exon 5-specific PCR were performed essentially as described previously (Brown et al. 1997; Brown et al. 2001b) in 25-μl reaction volumes: 0.875 U Taq polymerase (Promega GoTaq), 2.5 mM MgCl₂, 0.2 mM dNTPs, specific primers (0.4 μM each), 1 × Colorless GoTaq® Flexi Buffer, 1× Biotium EvaGreen Dye (31000) and 50 ng gDNA. Using these conditions, HRM successfully types alleles in the range 3-50 ng gDNA (data not shown). Allele-specific PCR was directly followed with HRM analysis (standard melt range from 65 to 95 °C in 0.2 °C increments every 12 sec). HRM analysis was performed using CFX Manager and Precision Melt Analysis software tools (Bio-Rad) and visualized by plotting relative fluorescence units (RFU) and relative difference curves (RFU) against temperature (°C) according to manufacturer specifications and protocol. The graphs display representative HRM difference RFU curves for (b) Ly49g and (c) H-2 D exon 5 alleles. All samples were performed in triplicate.

Table 2. Ly49 and H-2 class I allele primers, product size and HRM features.

Gene/SNP	Primer Name	Primer Sequence	Product Size	Optimal Annealing Temperature (°C)	Allelea Melt Temperature (°C)
Ly49G2	G2 E3 F G2 E3 R	TTCAACAAAAACATGAACTACAGGAAAC TAGTTTCACTGTACCATCTT	69	56	M = 81.2, F1 = 81.0, L = 81.4
H-2 K	H2K I3E4 F H2K I3E4 R	GTTTGTCTTGTTAATTGTGTG CCTCAGGGTGACTTTATCTTC	101	56.7	M = 80.2, F1 = 79.8, L = 79.6
	H2K I3I4 F H2K I3I4 R	GACATGATTGGGTTTCAGGA CTTCACGCTAGAGAATGAGG	108	60	All at 80.8
H-2 D	H2D I1 F H2D I1 R	TCGCCCACCGGACCCTCCGCCCCT CAGGCCGGGAGGGGATCTGGGCGC	63	66	M & F1 = 87.0, L = 87.2
	H2D I3 F H2D I3 R	CAGGCCGGGTTCTCTGCCCA CATGTTTCCAATCAGGTAAGGC	113	60	M = 84.4, F1 = 84.6, L = 85.2
	H2D E5 F H2D E5 R	CGTCCACTGACTCTTACATGGTGATCG AAAAGCCACCACAGCTCCAATGATGGC	82	60	M = 82.0, F1 = 81.8, L = 81.6
	H2D I8 F H2D I8 R	CTGTAAGCTCCATGCTACCCTGAG CTGAGCCATCTCTCCAGCCACCC	90	60	M = 80.8, F1 = 80.6, L = 80.4

^aM, MA/My; L, C57L; F₁, MA/My × C57L

Ly49g (Klra7) allele discrimination

Efficient viral control in MCMV infected MHC I D^k mice requires Ly49G+ NK cells (Xie et al. 2009). Ongoing research into NK inhibitory receptor recognition of viral infection includes genetic imaging of NK responses to viral infection in intercross and backcross offspring as well as in Ly49 congenic and transgenic mice (our unpublished data). The approach necessitates a fast, reproducible, selective and specific method to discriminate Ly49 alleles. Thus, we tested Ly49g allele-specific PCR followed by HRM analysis using cloned and sequence-verified Ly49g allele target sequences (Xie et al. 2009) to establish HRM genotyping. As shown in Fig 1b, Ly49g HRM readily detected MA/My, C57L and F₁ DNA types (Fig 1b). To validate the methodology, 233 genomic DNA (gDNA) samples of genome-wide genotyped F₂ and backcross offspring were HRM typed for Ly49g. Of 210 non-recombinants between SNPs rs13479014 and rs6339546 on chromosome 6, Ly49g (63.435 cM) HRM types displayed 100% concordance with both markers (Table 3), indicating that HRM typing for Ly49g alleles is highly consistent and reliable. In accord with this, Ly49g HRM readily detected C57L and congenic C57L.M-Nkc (data not shown) DNA types. These data demonstrate that HRM typing very effectively discriminates Ly49 alleles even in mixed genetic backgrounds (e.g. in Ly49 and/or NKC hybrid haplotypes) where their identification, tracking and confirmation are critical. These data are consistent with the findings of Gonzalez et al. demonstrating HRM discrimination of exceedingly polymorphic killer cell immunoglobulin-like receptor (KIR) alleles (Gonzalez et al. 2009). We conclude that HRM is an extremely practical method to quickly and accurately genotype polygenic and polymorphic NK cell receptor alleles in mice.

Table 3. Genetic confirmation of Ly49 and H-2 class I HRM genotyping.

SNP	Chr	Test Intervala (cM)	# Non-recombinants/233 hybrid offspringb	Concordancec
Ly49g	6	rs13479014 (59.17) - rs6339546 (64.03)	210	210/210
H-2 Dk exon 5	17	rs3693494 (15.69) - rs6298471 (19.16)	118	118/118
H-2 Kk intron 3 - exon 4	17	H2-Dk Exon 5	233	233/233

^aSNP markers (with cM position) flanking the relevant genes in the test genomic interval are given. All Ly49 and H-2 SNP typings were performed in 96-well plates with relevant genotype controls included in every PCR plate. Random gDNA samples selected for confirmatory HRM analysis agreed with initial HRM typings in every case (data not shown).

^bCohorts of backcross and F₂ offspring (n = 233) issued in genetic crosses of MA/My and C57L mice were genome-wide genotyped by DartMouse (Dartmouth University, Lebanon, New Hampshire) using the Illumina 1449 SNP array. Shown are the number offspring without DNA crossovers in the test interval.

^cConcordance is equal to the number mice with HRM types identical to both flanking SNP markers/the number of non-recombinant genotypes for the test genomic interval.

H-2 class I allele discrimination and gene copy number determination

Due to its vital role in NK cell-mediated viral control (Stadnisky et al. 2011; Xie et al. 2010), we sought to establish HRM typing for MHC I D allele variants. Although we had already developed genotyping for class I D alleles, SSLP proved challenging due to frequent amplification failures, low tolerance for DNA concentration variation and inability to amplify D^b in the presence of D^k or to assess allele copy number variation (data not shown). In pioneering studies, Wittwer and colleagues established HRM typing for HLA A and B alleles, thereby demonstrating its enormous potential and clinical applicability in determining HLA allele identities (Seipp et al. 2005; Zhou et al. 2004). However, HRM typing for murine H-2 class I alleles has yet to be reported. To pursue this, we tested and optimized MHC I D^k exon 5 allele-specific PCR using a previously cloned and sequence verified MHC I D^k gDNA fragment (Xie et al. 2010). As with Ly49g typing, exon 5 HRM consistently and reliably detected MA/My, C57L and F₁ DNA types (Fig 1c). We further verified exon 5 marker linkage with gDNA samples described above. Of 118 non-recombinants between SNPs rs3693494 and rs6298471 on chromosome 17, exon 5 types displayed 100% concordance (Table 3), indicating that exon 5 HRM is also highly consistent and reliable. For the purpose of precision genetic mapping in hybrids with crossovers in the critical MHC interval, we designed and tested several additional allele-specific HRM markers for MHC I D (Table 1). In accord with exon 5 types, introns 1, 3 and 8 HRM types were identical in all 118 non-recombinants. Thus, MHC I D allele variants can be efficiently screened and identified using HRM genotyping. Using a similar approach, two novel MHC class I K markers were generated (Table 1). Importantly, MHC I K (17.982 cM) i3e4 and i7i8 HRM genotypes were fully concordant with MHC I D (i.e. there were no crossovers between the class I genes in 233 gDNA samples; Table 3). These data therefore demonstrate that HRM typing of H-2 class I alleles is consistent and very reliable.

Successful genotyping for class I D alleles suggested that HRM may be readily adapted to identify and type MHC I D^k transgenes in mice of different genetic backgrounds. To assess this, we examined two D^k transgenic lines (i.e. Tg1-D^k and Tg3-D^k) previously bred into two different genetic backgrounds (Xie et al. 2010). As expected, C57L.Tg3-D^k and M.H2^b-Tg1-D^k HRM types differed from both non-transgenic breeding partners, C57L and M.H2^b (not shown), directly confirming the utility of this detection method (Fig 2a). However, Tg-D^k HRM types also differed by comparison with MA/My and F₁ (Fig. 2a). Similar results were obtained with C57L.Tg1-D^k (data not shown). These data suggested that the ratio of D^k and D^b amplicons present in transgenic and non-transgenic PCR reaction mixtures may differ. Nonetheless, HRM genotyping is highly amenable to efficient genetic screening and reliable identification of MHC I D transgenes in C57L and M.H2^b genetic backgrounds.

Figure 2.

HRM detection and analysis of H-2 class I D transgenes in two different H-2 class I D^k transgenic mouse lines. a) The graph shows representative H-2 D exon 5 HRM difference curves for hemizygous H-2 class I D^k transgenic lines, Tg1 and Tg3 and the indicated control gDNAs. b) The graph shows representative H-2 D exon 5 HRM difference curves for hemi- and homozygous Tg1, in addition to control gDNA HRM types. Gene-specific PCR and HRM analysis were performed as described for Figure 1. Graphs in a and b are representative of two and four experiments, respectively.

Distinctive HRM profiles obtained for Tg-D^k and non-transgenic genomes suggested that D^k/D^b ratio disparity in the different gDNA samples may have contributed to amplicon melting variations. This outcome further hinted that Tg-D^k mice may harbor multiple transgene integrants. Slightly higher cell surface D^k expression in Tg-D^k peripheral blood cells than in heterozygous cells with only one endogenous D^k allele expressed (Xie et al. 2010) and disparate HRM profiles for homozygous (verified in breeding experiments) and hemizygous Tg-D^k mice (Fig. 2b) support this notion. To investigate possible D^k gene copy number differences, we generated a panel of gDNA samples with D^k/D^b ratios ranging from 1 to 10 by establishing C57L and MA/My DNA admixtures and carried out HRM analysis which could then be compared with both Tg-D^k HRM types. As expected, D^k and D^b were amplified with exon 5 primers at similar rates (i.e. C_t values were not significantly different) in each of the gDNA samples (data not shown). However, the relative difference curves based on variance in melting rates of the different admixtures appeared distinct (Fig. 3a). To further test this, melting point mean difference curve RFU values were plotted against D^k/D^b ratios of each admixture, which revealed remarkable association (R² = 0.99) (Fig. 3b). Thus, we extrapolated gDNA ratio values for both transgenic lines (Tg1-D^k and Tg3-D^k) based on observed mean difference RFU values. In repeated experiments, gDNA ratios of ~3 and ~2 were obtained for Tg1 and Tg3 lines, respectively (Table 4). In accord with this, Tg1 and Tg3 mean difference RFU values were not significantly different from those obtained for admixtures with D^k/D^b ratio values of 3 and 2, respectively (not shown). On the other hand, Tg1 and Tg3 mean difference RFU values differed by comparison to all other admixtures (not shown). Because both strains are homozygous for D^b, we inferred from these data that Tg1 likely integrated 6 D^k transgenes, whereas Tg3 most likely integrated 4 D^k transgenes. Based on these data, we predicted a D^k/D^b ratio of 6 for Tg1^+/+. However, extrapolated D^k/D^b ratios for Tg1^+/+ (~5) suggest that there may be an upper limit to the resolving power in HRM discrimination of allele variants which was likely surpassed under these conditions in the case of Tg1^+/+. The addition of KCl prior to HRM analysis may improve resolution (Vossen et al. 2009), however we did not assess the effect of salt on resolving allele ratios. As a further test, comparison of Tg-D^k plus C57L and MA/My plus C57L admixtures by HRM analysis in the same experiment revealed extrapolated D^k/D^b ratio values of 1.58 and 0.92 (Table 4), which corroborates probable D^k/D^b ratio values and estimated transgene copy numbers given above for both transgenic lines.

Figure 3.

HRM detection and analysis of H-2 class I D allele copy number differences in two different H-2 class I D^k transgenic mouse lines. MA/My and C57L admixture gDNA samples were produced by combining spleen DNA samples after meticulous sampling and spectrophotometric (ABS₂₆₀) measurement of DNA concentration. To verify gDNA samples, all sample measurements were performed in triplicate for multiple gDNA dilutions, including 1:10, 1:20, 1:50 and 1:100 dilutions in 1X TE. Subsequently, gDNA samples (12.5 ng/μl) were mixed to generate the designated D^k:D^b admixture ratios. HRM types were determined as decribed for Figure 1, except only 25 ng gDNA/reaction was used. a) The graph shows representative single H-2 D exon 5 HRM difference curves for gDNA admixtures containing D^k:D^b ratios 1:1 to 5:1 and the indicated control gDNAs. All admixture and control gDNA HRM types were performed in triplicate. Data is representative of four independent experiments. b) The graph shows mean HRM difference RFU values at 81.9 °C for the indicated D^k:D^b ratios obtained in (a) when plotted against the D^k:D^b ratio. Regression analysis was used to determine a logarithmic best fit line and equation to extrapolate D^k:D^b ratios from mean difference RFU values of unknown gDNA samples. Data is representative of four independent experiments.

Table 4. HRM analysis of transgene copy number in H-2 class I D^k transgenic lines.

Strain (Genotype)	M.Tg1(+/−)	L.Tg1(+/−)	L.Tg1(+/+)	L.Tg1(+/−) + L	L.Tg3(+/−)	L.Tg3(+/−)	L.Tg3(+/−) + L
Tissue (Admixture)	Spleen	Toe	Toe	Spleen (50:50)	Spleen	Toe	Spleen (50:50)
Mean extrapolated Dk:Db ± S.D.	2.78 ± 0.16	3.22 ± 0.39	5.21 ± 0.26	1.58	2.03 ± 0.40	1.98 ± 0.06	0.92
Number of Triplicate Measurements	2	8	3	1	4	3	1

We conclude that HRM typing is a straightforward and dependable approach to rapidly discriminate highly polymorphic and polygenic murine NK receptor and MHC class I allele variants and determine gene dosage, even in highly related gene families. We have shown that HRM genotyping for congenic, transgenic and other gene modifications in genetically selected and/or targeted mice is feasible. As an identification strategy, HRM is a powerful method that can easily be harnessed for the purpose of following genetic manipulations in targeted mouse strains. Likewise, HRM provides another approach to identify, localize and pinpoint novel DNA markers to resolve informative breakpoints and loci, which define a critical genetic interval. We further envision that HRM typing is well suited and likely easily adapted to quantitative expression analyses in situations where other traditional methodologies may fail due to inability to discriminate amongst related gene transcripts with high homology and in situations when it may be advantageous to examine allele-specific expression and regulation of a single or multiple genes.