Abstract

Identifying molecular drivers of pathology provides potential therapeutic targets. Differentiating between drivers and coincidental molecular alterations presents a major challenge. Variation unrelated to pathology further complicates transcriptomic, proteomic and metabolomic studies which measure large numbers of individual molecules. To overcome these challenges towards the goal of determining drivers of Huntington's disease (HD), we generated an allelic series of HD knock-in mice with graded levels of phenotypic severity for comparison with molecular alterations. RNA-sequencing analysis of this series reveals high numbers of transcripts with level alterations that do not correlate with phenotypic severity. These discorrelated molecular changes are unlikely to be drivers of pathology allowing an exclusion-based strategy to provide a short list of driver candidates. Further analysis of the data shows that a majority of transcript level changes in HD knock-in mice involve alteration of the rate of mRNA processing and/or degradation rather than solely being due to alteration of transcription rate. The overall strategy described can be applied to assess the influence of any molecular change on pathology for diseases where different mutations cause graded phenotypic severity.

Introduction

Identification of remediable pathological molecular processes is a cornerstone of rational therapeutic design. A common first step in this strategy is to search for bio-molecules with levels that differ between normal and pathological states. Powerful techniques involving genomics, proteomics, transcriptomics and metabolomics provide means of simultaneously identifying and reporting the concentrations of thousands of molecules. Such studies reveal pathologic correlates that broadly fall into one of four classes: pathology drivers, beneficial responders, epiphenomenal changes and false-positives arising from chance. Well-correlated molecular changes that are not pathologic drivers may be useful as state or target engagement biomarkers. False positives are likely to be frequent, especially when obtained by screening large numbers of molecules. Accumulation of false-positive candidates makes determination of therapeutically relevant drivers, responders, and biomarkers difficult. An alternate strategy to the correlative approach is to exclude candidates that fail to correlate with phenotypic severity. Identification of a molecular change as not correlating with phenotypic severity argues against it being a driver or responder.

This approach may be particularly useful for diseases such as Huntington's disease (HD) that exhibit different intensities of a single pathologic etiology (1). Caused by inheritance of an expanded CAG/polyglutamine repeat in the Huntingtin gene, this midlife onset neurological disease exhibits wide variation in age of onset. Longer repeat lengths are associated with earlier onset of clinical features (2). Murine genetic models of HD share this relationship between length and phenotypic onset with the exception of some extremely long transgenic CAG repeat mice (3–5).

These models have been used to identify molecular alterations that might drive HD-like abnormalities with the goal of uncovering therapeutic targets (6–8). Transcript profiling, for example, revealed several alterations that are also present in HD patient tissues postmortem (9). These results and the presence of polyglutamine repeats in many other proteins involved in transcription support the hypothesis that the HD mutation acts through transcriptional dysregulation (10). A speculative extension of this idea is that some transcript level changes drive the pathology of HD and therapies aimed at normalizing these levels might prove therapeutic. Despite the promise of such potential drivers, the field is faced with the challenge of having too many such candidates. In fact, approximately one-third of all transcripts analyzed are found to have altered levels in HD (11). Enhanced screening methods are needed to narrow this large field of candidates.

We describe an allelic series of knock-in HD mice that allows rank ordering by phenotypic severity and comparison with transcript profiles from RNA-sequencing analysis of four critical members of the series. The mRNA studied was from the area of the brain most severely affected in HD, the striatum. Data were analyzed to assess the power of the method. Correlating transcript levels with phenotypic severity provided at least 7-fold enrichment of driver/responder/biomarker candidates than would be expected by the standard method of repeatedly comparing one line to a control. Genes with transcript levels alterations that do not correlate with phenotypic severity were at least 20-fold more numerous than those correlating. The nature of the changes was explored by analysis of transcripts with more than one distinguishable isoform which showed that most transcript level alterations involve more than just changes in transcription rate. This result implicates the HD mutation in alterations of mRNA processing and/or degradation.

Results

The HD knock-in allelic series

We previously reported a mouse model of HD where gene targeting was used to replace the short CAG repeat of the mouse HD gene with 150 CAGs (12). This mouse exhibited mild late onset motor, behavioral and neuroanatomic abnormalities consistent with HD (12,13). A line congenic to C57BL/6 was derived; then germ line instability in repeat number was exploited to selectively breed longer and shorter repeat variants. The resulting lines with 50, 100, 150, 200, 250 and 315 CAGs were maintained (Fig. 1A). HD knock-in mice with fewer than 100 CAGs show no overt HD-like abnormalities, 150 CAGs mild symptoms (12) while longer repeats (200–250) show earlier ages of onset of the same abnormalities (14,15).

Figure 1.

Molecular aspects of the HD knock-in allelic series. (A) PCR across the repeats of genomic DNA from tail biopsies. Lanes 1–8 represent MW ladder, WT and heterozygotes for 50, 100, 150, 200, 250 and 315 CAGs in the mouse HD gene, respectively. (B) Striatal HD mRNA levels of long repeat mRNAs relative to wild type as determined by allele-specific HD QRTPCR. Levels in homozygotes was halved to provide mRNA level per allele. Each assay performed in triplicate and results normalized to β-actin mRNA levels of same sample. White, gray and black shading indicates WT, heterozygous and homozygous for expanded allele, respectively. Error bars indicate standard error of the mean and asterisks show level of statistical certainty versus line ∼100 CAGs shorter as determined by ANOVA with Tukey–Kramer multiple comparisons test (*P < 0.05, **P < 0.01, ***P < 0.001). Number of mice is shown at the base of each bar. (C) Western blot analyses of striatal protein from allelic series. Lanes 1–7 represent WT, 50, 100, 150, 200 homozygotes, 250 and 315 heterozygotes, respectively. Top panel shows detection with the anti-polyglutamine antibody 1C2, middle shows the same blot after stripping and detection with the anti-HD antibody 2166 and lower with anti-α-tubulin antibody. Filled arrowheads indicate the expected migration of wild-type Huntingtin protein (345 kDa) and open arrowhead indicates expected migration of α-tubulin (50 kDa).

Figure 1.

Molecular aspects of the HD knock-in allelic series. (A) PCR across the repeats of genomic DNA from tail biopsies. Lanes 1–8 represent MW ladder, WT and heterozygotes for 50, 100, 150, 200, 250 and 315 CAGs in the mouse HD gene, respectively. (B) Striatal HD mRNA levels of long repeat mRNAs relative to wild type as determined by allele-specific HD QRTPCR. Levels in homozygotes was halved to provide mRNA level per allele. Each assay performed in triplicate and results normalized to β-actin mRNA levels of same sample. White, gray and black shading indicates WT, heterozygous and homozygous for expanded allele, respectively. Error bars indicate standard error of the mean and asterisks show level of statistical certainty versus line ∼100 CAGs shorter as determined by ANOVA with Tukey–Kramer multiple comparisons test (*P < 0.05, **P < 0.01, ***P < 0.001). Number of mice is shown at the base of each bar. (C) Western blot analyses of striatal protein from allelic series. Lanes 1–7 represent WT, 50, 100, 150, 200 homozygotes, 250 and 315 heterozygotes, respectively. Top panel shows detection with the anti-polyglutamine antibody 1C2, middle shows the same blot after stripping and detection with the anti-HD antibody 2166 and lower with anti-α-tubulin antibody. Filled arrowheads indicate the expected migration of wild-type Huntingtin protein (345 kDa) and open arrowhead indicates expected migration of α-tubulin (50 kDa).

Reduction in HD mRNA levels with increased repeat size

Longer repeat alleles have lower HD mRNA levels (Fig. 1B) and express expanded polyglutamine HD protein at levels readily detectable by western analysis (Fig. 1C). Reduction in HD mRNA levels with increasing CAG repeat length is a property shared by the longer repeat length versions of an analogous allelic series derived from the R6/2 transgenic line of mice (Fig. 2). The R6/2 series was created by random insertion of a portion of the human genomic region containing the promoter and exon 1 of HD with an expanded repeat (16) and breeding of germline repeat size changes to create a series of different repeat lengths each with the same albeit unknown genomic location (5).

Figure 2.

Transgene mRNA levels from the R6 HD allelic series. (A) R6 transgene mRNA copies per nanogram of whole brain RNA at 6 weeks of age using a transgene-specific QRTPCR assay. All R6 mice were hemizygous for transgene which is inserted in the same unknown location of the genome for all lines in the series. (B) HD mRNA copies per nanogram of whole brain RNA at 6 weeks of age weeks for HDQ50 and HDQ100 knock-in homozygotes by QRTPCR across exon2,3 junction. Absolute numbers of cDNAs present in each sample were calculated by interpolation to a standard curve of known amounts of DNA template for R6 or endogenous mouse HD. Number of mice is shown at the base of each bar. Error bars indicate standard error of the mean. Unless annotated asterisks reflect comparison with measures two bars to left (typically 100 CAGs shorter) as determined by ANOVA with Tukey–Kramer multiple comparisons test (*P < 0.05, **P < 0.01, ***P < 0.001). acompared with all other transgenic R6 lines shown including 50s. bcompared with HDQ50s.

Figure 2.

Transgene mRNA levels from the R6 HD allelic series. (A) R6 transgene mRNA copies per nanogram of whole brain RNA at 6 weeks of age using a transgene-specific QRTPCR assay. All R6 mice were hemizygous for transgene which is inserted in the same unknown location of the genome for all lines in the series. (B) HD mRNA copies per nanogram of whole brain RNA at 6 weeks of age weeks for HDQ50 and HDQ100 knock-in homozygotes by QRTPCR across exon2,3 junction. Absolute numbers of cDNAs present in each sample were calculated by interpolation to a standard curve of known amounts of DNA template for R6 or endogenous mouse HD. Number of mice is shown at the base of each bar. Error bars indicate standard error of the mean. Unless annotated asterisks reflect comparison with measures two bars to left (typically 100 CAGs shorter) as determined by ANOVA with Tukey–Kramer multiple comparisons test (*P < 0.05, **P < 0.01, ***P < 0.001). acompared with all other transgenic R6 lines shown including 50s. bcompared with HDQ50s.

The expression decrease in longer length repeats might explain the diminishing phenotypic abnormalities seen in the R6/2 series with repeats above ∼300 CAGs in length (3–5). These data are consistent with the view that longer repeats produce a more toxic gene product on a per molecule basis but reduced transgene expression of super long repeat lengths diminishes phenotypic effects (3–5).

HD-like abnormalities in the HDQ315/+ line

Previous studies on the R6/2 allelic series show decrements in HD-like abnormalities for mice with ∼300 CAGs when compared with 150 CAGs (3–5,17). Unlike the R6/2 series, our knock-in series does not share the reduction in phenotype shown in the transgenic R6/2 series at 300 repeats. The toxicity of the HDQ315 allele is significantly greater than HDQ150 and HDQ200 as suggested by inability to breed HDQ315 allele homozygotes (Table 1). Longitudinal behavioral analyses of the HDQ315/+ line show deficits as early as 30 weeks which worsen in degree through 65 weeks (Fig. 3). Neuroanatomic analyses show a decrease in D1 and D2 dopamine receptor binding by 35 weeks of age and a significant decrease in brain weight at 70 weeks, not accompanied by striatal neuronal loss (Fig. 4). Striatal aggregates immunoreactive to N-terminal HD antibodies were also prevalent at 70 weeks (Fig. 4).

Table 1.

Offspring of heterozygous by heterozygous crosses

Offspring CAG repeat length
 
50 100 150 200 250 315 
WT 26 64 192 79 67 18 
Heterozygote 39 101 317 163 109 35 
Homozygote 29 52 167 30* 1* 0* 
Total 94 217 676 272 177 53 
Offspring CAG repeat length
 
50 100 150 200 250 315 
WT 26 64 192 79 67 18 
Heterozygote 39 101 317 163 109 35 
Homozygote 29 52 167 30* 1* 0* 
Total 94 217 676 272 177 53 

*Significant decrease from 1:2:1 Mendelian ratio (P < 0.0005 χ2).

Figure 3.

Longitudinal evaluation of behavioral abnormalities of the HDQ315 line at different ages. (A) Accelerating rotarod. (B) Time to traverse horizontal ladder in automated foot misplacement apparatus. (C) In cage activity (lower beam breaks per 24 h period). (D) Open field distance traveled per 4-min trial. (E) Voluntary wheel cage distance traveled per 24 h period. (F) Four paw grip strength. Open and gray bars represent mean of WT and HDQ315/+ mice, respectively. Error bars indicate standard error of the mean and number of mice in each group shown at base of bar. Groups contain approximately equal numbers of males and females. Statistical differences between WT and HDQ315/+ groups at same age as determined by t-test shown above bar (*P < 0.05, **P < 0.01 and ***P < 0.001). Statistically significant differences within the HDQ315/+ group between 30-week-old and both 50 and 65 weeks old were found for all six tests shown (P < 0.05 by Kruskal–Wallis non-parametric ANOVA with Dunn's multiple comparison test) showing progression of abnormality with age.

Figure 3.

Longitudinal evaluation of behavioral abnormalities of the HDQ315 line at different ages. (A) Accelerating rotarod. (B) Time to traverse horizontal ladder in automated foot misplacement apparatus. (C) In cage activity (lower beam breaks per 24 h period). (D) Open field distance traveled per 4-min trial. (E) Voluntary wheel cage distance traveled per 24 h period. (F) Four paw grip strength. Open and gray bars represent mean of WT and HDQ315/+ mice, respectively. Error bars indicate standard error of the mean and number of mice in each group shown at base of bar. Groups contain approximately equal numbers of males and females. Statistical differences between WT and HDQ315/+ groups at same age as determined by t-test shown above bar (*P < 0.05, **P < 0.01 and ***P < 0.001). Statistically significant differences within the HDQ315/+ group between 30-week-old and both 50 and 65 weeks old were found for all six tests shown (P < 0.05 by Kruskal–Wallis non-parametric ANOVA with Dunn's multiple comparison test) showing progression of abnormality with age.

Figure 4.

Neuroanatomic and neurochemical features of the HDQ315 line. (A) HDQ315/+ mice have reduced brain weight at 70 weeks of age. (B–D) HDQ315/+ mice show reduction in striatal volume, increase in striatal neuronal density and no loss of NeuN+ striatal neurons at 70 weeks of age. (E and F) HDQ315/+ mice show reductions in D1 and D2 receptors binding at 35 and 70 weeks of age. Open bars indicate WT mice and gray bars HDQ315/+ mice. Number of mice in each group is shown at base of bar. Asterisks indicate significant difference from wild type (* P < 0.05, **P < 0.01 and ***P < 0.001) by t-test. (H) Aggregates immunoreactive to anti-HD antibody in 70-week-old HDQ315/+ striatum that are not present in age-matched WT controls (G). Bar indicates 20 µm.

Figure 4.

Neuroanatomic and neurochemical features of the HDQ315 line. (A) HDQ315/+ mice have reduced brain weight at 70 weeks of age. (B–D) HDQ315/+ mice show reduction in striatal volume, increase in striatal neuronal density and no loss of NeuN+ striatal neurons at 70 weeks of age. (E and F) HDQ315/+ mice show reductions in D1 and D2 receptors binding at 35 and 70 weeks of age. Open bars indicate WT mice and gray bars HDQ315/+ mice. Number of mice in each group is shown at base of bar. Asterisks indicate significant difference from wild type (* P < 0.05, **P < 0.01 and ***P < 0.001) by t-test. (H) Aggregates immunoreactive to anti-HD antibody in 70-week-old HDQ315/+ striatum that are not present in age-matched WT controls (G). Bar indicates 20 µm.

Ordering by phenotypic severity

These behavioral and neurochemical analyses place the age of onset of mice heterozygous for the HDQ315 allele and the wild-type allele (HDQ315/+) somewhere between the later onset HDQ150/150 and the earlier onset HDQ200/200 lines. This rank ordering is illustrated in Figure 5 which compares body weight loss, rotarod performance and general activity levels for the three lines. Comparison with other data supports this ordering. Between 35 and 40 weeks of age, HDQ315/+ mice show D1 dopamine receptor binding reduced by 25–30% compared with wild type (WT), midway between HDQ150/150 mice (11% reduction) and HDQ200/200 mice (50–60% reduction) (13). The utility of ranking a heterozygote between two homozygotes is that it results in different orders for phenotypic severity (WT < HDQ150/150 < HDQ315/+ < HDQ200/200) and mutant gene dosage (WT < HDQ315/+ < HDQ150/150 and HDQ200/200). Differential ordering dissociates molecular changes more influenced by expression levels of mutant Huntingtin from molecular changes related to the overall toxicity of the mutation.

Figure 5.

Comparison of phenotypic severity of HDQ150/150, HDQ315/+ and HDQ200/200 lines. (A) Body weight normalized to wild-type mice of same age and sex. Black triangles represent HDQ150/150 mice (each point n = 13–22 except 80 weeks n = 9). Gray circles represent HDQ315/+ mice (n = 13 to 29) and black squares represent 200/200 mice (n = 3–7). (B) Rotarod performance normalized to wild-type mice of same age. Black triangles represent HDQ150/150 mice (each point n = 10 except 26 weeks n = 31). Gray circles represent HDQ315/+ mice (n = 29–31) and black squares represent HDQ200/200 mice (n = 3). (C) In cage mouse activity normalized to wild-type mice of same age. Black triangles represent HDQ150/150 mice (each point n = 7–15 except 26 weeks n = 20). Gray circles represent HDQ315/+ mice (n = 29–31) and black squares represent HDQ200/200 mice (n = 3). Asterisks show statistical significance compared with HDQ315/+ mice of most similar age by t-test (*P < 0.05, **P < 0.01 and ***P < 0.001). Approximately equal numbers of male and female mice were used for each point. ‘a’ indicates our previously published data adapted from (13). Lines determined by method of least squares. In each panel, upper, middle and lower lines represent HDQ150/150, HDQ315/+ and HDQ200/200 mice, respectively.

Figure 5.

Comparison of phenotypic severity of HDQ150/150, HDQ315/+ and HDQ200/200 lines. (A) Body weight normalized to wild-type mice of same age and sex. Black triangles represent HDQ150/150 mice (each point n = 13–22 except 80 weeks n = 9). Gray circles represent HDQ315/+ mice (n = 13 to 29) and black squares represent 200/200 mice (n = 3–7). (B) Rotarod performance normalized to wild-type mice of same age. Black triangles represent HDQ150/150 mice (each point n = 10 except 26 weeks n = 31). Gray circles represent HDQ315/+ mice (n = 29–31) and black squares represent HDQ200/200 mice (n = 3). (C) In cage mouse activity normalized to wild-type mice of same age. Black triangles represent HDQ150/150 mice (each point n = 7–15 except 26 weeks n = 20). Gray circles represent HDQ315/+ mice (n = 29–31) and black squares represent HDQ200/200 mice (n = 3). Asterisks show statistical significance compared with HDQ315/+ mice of most similar age by t-test (*P < 0.05, **P < 0.01 and ***P < 0.001). Approximately equal numbers of male and female mice were used for each point. ‘a’ indicates our previously published data adapted from (13). Lines determined by method of least squares. In each panel, upper, middle and lower lines represent HDQ150/150, HDQ315/+ and HDQ200/200 mice, respectively.

Rationale for molecular comparisons within the series

The comparison of molecular levels typically involves finding differences between two groups (e.g. mutant versus WT control). In the HD field, where several animal models are available, consensus as to the validity of a changed molecular level is formed by the compilation of varied mutant versus WT comparisons. This approach is based on the assumption that molecular drivers of each mutant's abnormalities are the same. This assumption is also important for the approach described here in which molecular levels are compared simultaneously across several lines. There is support for the view that the drivers in the HD knock-in allelic series are similar. First, the mutation is the same except for repeat length. Second, use of congenics reduces genetic background effects. Third and most important, the abnormalities seen are strikingly similar but with different ages of onset suggesting longer repeat lengths accelerate the same toxic processes. Phenotypic similarity also holds true when a specific repeat length is made homozygous. Both HDQ150 and HDQ200 allele homozygotes exhibit strikingly similar behavioral and neuroanatomic features which present at an earlier age than their corresponding heterozygous lines (12–14, 18). The HDQ315/+ mice share this similarity in type of abnormality (Figs. 3–5). Further supporting comparisons between homozygous and heterozygous lines is evidence that loss of wild-type HD in homozygotes has little effect on striatal transcript profiles. Ninety four percent concordance is seen between transcript-level alterations HDQ150/150 (which lack WT HD) and R6/2-Q150 mice (which have wild-type endogenous HD) (9). These genetic and phenotypic similarities support the approach of simultaneous comparisons across the allelic series.

Screen for expression correlates and discorrelates

RNA-sequencing analysis was used to assess early changes in striatal transcript levels between WT, HDQ150/150, HDQ315/+ and HDQ200/200 lines. We chose 21 weeks of age given results that some early markers of HD show significantly altered transcript levels at this age and since it was an age early in the course of disease. At this age, stereological results (13,14, Fig. 4) show no decrease in striatal neuronal number, eliminating skewing of transcript levels secondary to neurodegeneration. The depth of sequencing allowed analyses of transcript levels from 15 543 genes based on the inclusion criteria presented in Materials and Methods. Table 2 shows that the number of transcript levels changing in the same direction versus WT for HDQ150/150, HDQ315/+ and HDQ200/200 mice allowing for different levels of variation. This table shows that a transcript level alteration in one direction versus wild type for all three mutants were at least 14-fold more than expected if the changes were independent in each line. Such concordance supports a multiple comparison approach for transcript profiling.

Table 2.

Concordance and enhanced discrimination supports use of ordering by phenotypic severity

 Allowed variation υ (%) 12 15 18 21 
Decreasing mRNA level WT±υ>150±υ 7955 6680 5693 4917 4186 3679 3272 2948 
WT±υ>315±υ 6609 5644 4874 4222 3728 3341 2894 2635 
WT±υ>200±υ 9818 8924 7992 7039 6072 5170 4386 3694 
WT±υ>150±υ,315±υand200±υ 
 Expecteda 214 139 92 60 39 26 17 12 
 Observed 3306 2565 2026 1626 1318 1128 959 810 
 Observed/expected 15 18 22 27 34 43 56 68 
WT±υ>150±υ>315±υ>200±υ 469 247 117 84 32 20 15 
 Discrimination ratiob 7 10 17 19 41 56 64 90 
Increasing mRNA level WT±υ<150±υ 7505 6215 4921 3861 3095 2544 2108 1813 
WT±υ<315±υ 8874 7704 6423 5173 3981 3083 2479 1927 
WT±υ<200±υ 5661 4927 4279 3761 3353 3002 2762 2531 
WT±υ<150±υ,315±υand200±υ 
 Expected concordantsa 156 98 56 31 17 10 
 Observed concordants 2252 1680 1214 945 748 616 526 379 
 Observed/expected 14 17 22 30 44 63 88 104 
WT±υ<150±υ<315±υ<200±υ 272 137 83 60 44 31 27 23 
 Discrimination ratiob 8 12 15 16 17 20 20 17 
 Allowed variation υ (%) 12 15 18 21 
Decreasing mRNA level WT±υ>150±υ 7955 6680 5693 4917 4186 3679 3272 2948 
WT±υ>315±υ 6609 5644 4874 4222 3728 3341 2894 2635 
WT±υ>200±υ 9818 8924 7992 7039 6072 5170 4386 3694 
WT±υ>150±υ,315±υand200±υ 
 Expecteda 214 139 92 60 39 26 17 12 
 Observed 3306 2565 2026 1626 1318 1128 959 810 
 Observed/expected 15 18 22 27 34 43 56 68 
WT±υ>150±υ>315±υ>200±υ 469 247 117 84 32 20 15 
 Discrimination ratiob 7 10 17 19 41 56 64 90 
Increasing mRNA level WT±υ<150±υ 7505 6215 4921 3861 3095 2544 2108 1813 
WT±υ<315±υ 8874 7704 6423 5173 3981 3083 2479 1927 
WT±υ<200±υ 5661 4927 4279 3761 3353 3002 2762 2531 
WT±υ<150±υ,315±υand200±υ 
 Expected concordantsa 156 98 56 31 17 10 
 Observed concordants 2252 1680 1214 945 748 616 526 379 
 Observed/expected 14 17 22 30 44 63 88 104 
WT±υ<150±υ<315±υ<200±υ 272 137 83 60 44 31 27 23 
 Discrimination ratiob 8 12 15 16 17 20 20 17 

Shown is the number of genes out of 15 543 genes analyzed by RNA sequencing. All transcripts from a single gene combined and classified by relative FPKM between WT, 150 (HDQ150/150), 315 (HDQ315/+) and 200 (HDQ200/200). aExpected is calculated as the product of the proportion of total altered genes of the three mutant versus WT comparisons times total number of genes. bDiscrimination ratio is calculated as the observed concordants divided by the step-wise phenotypic correlates shown in the next row above. Bold numbers indicate calculated ratios. Statistical differences by χ2 (P < 0.0001) were found for all observed and expected concordants.

Graphical examples of transcript levels across the series are presented in Figure 6. The relationships of transcript levels across the series are classified as described in Materials and Methods. Panels A–C show a class termed ‘step-wise correlates’ which represent driver/responder/biomarker candidates. Panels D–F show a new class that is not possible to identify for individual mutant versus WT comparison studies. This class is termed ‘discorrelate’ because it indicates there can be no correlation between transcript level and phenotypic severity. It arises when two or more of the members of the series show a trend which is then broken by one or both other members of the series. As such, the discorrelate class consists of transcripts with a high probability of being negative when addressing the hypothesis that a transcript level alteration is driving phenotype. A look at the discorrelate class reveals a shortcoming in a single mutant versus WT analysis that the simultaneous assessment of the several lines of the allelic series overcomes. For example, Figure 6E shows that an experiment assessing Tbr1 mRNA levels in the single HDQ150/150 versus WT comparison would show a large increase with the HD mutation while an experiment that compared HDQ315/+ versus WT would show no change. The great number of such discorrelates suggests single mutant versus WT comparisons often lead to false-positive driver candidates. The potential power to rule out false positives using multiple line comparisons with an ordered grading of phenotypic severity as opposed to treating them as three independent mutant versus WT comparisons is shown in Table 2. These data show that the multiple comparison approach was at least seven times better at ruling out false driver/responder/biomarker candidates. Furthermore, in Table 3, a complete list of mRNAs where such classifications have statistical support, discorrelates vastly outnumber step-wise correlates. A wider view of phenotypic classes is presented in Table 4. In this table, every gene's total mRNA level is classified without regard to statistical significance by allowing different levels of variation. These data reveal that step-wise phenotypic correlates, an indicator of a potential driver/responder/biomarker of pathology, are rare, at least 20-fold fewer were found than discorrelates. Partial correlates, a less stringent classification allowing for some equivalent values, are also typically less prevalent than phenotypic discorrelates. For transcripts with altered levels, phenotypic discorrelates are the largest class, representing thousands of genes.

Table 3.

Significant correlates and discorrelates

 Gene Locus WT FPKM 150 FPKM 315 FPKM 200 FPKM WT versus 150 WT versus 315 WT versus 200 150 versus 315 150 versus 200 315 versus 200 
Phenotypic correlates Alas2 chrX:150547416–150570622 3.86 2.73 1.16 0.89 NS ** ** NS 
Hba-a1 chr11:32283671–32284493 99.48 82.94 35.18 25.72 NS *** *** *** *** NS 
 Hbb-bs chr7:103826522–103827928 211.5 200.8 86.12 55.43 NS *** *** *** *** NS 
 Hbb-b1 chr7:103812523–103813923 29.89 28.42 10.21 6.98 NS ** *** ** *** NS 
 Igfbp2 chr1:72824479–72852471 26.53 13.43 12.87 6.57 NS *** NS NS 
 Oprm1 chr10:6788600–7038209 0.09 0.14 0.16 0.80 ** NS NS *** NS ** 
 Sod3 chr5:52363803–52369738 8.14 7.31 3.73 3.27 NS ** NS NS 
Phenotypic discorrelates Abhd12b chr12:70154161–70183343 0.19 0.20 6.52 0.32 NS ** NS NS 
Adam33 chr2:131050816–131063814 0.35 1.11 0.14 0.50 NS NS NS NS 
 Adcyap1 chr17:93199421–93205489 1.34 4.15 0.89 2.03 ** NS NS ** NS NS 
 Adh1 chr3:138277644–138290691 0.79 0.28 3.75 0.22 NS ** NS ** NS 
 Agt chr8:124556586–124569707 5.59 3.96 12.65 4.09 NS ** NS *** NS 
 Aldob chr4:49535976–49549483 0.58 1.25 4.64 0.57 NS *** NS ** NS ** 
 Apoc1 chr7:19689480–19692658 2.80 1.73 29.69 1.09 NS *** NS ** NS 
 Arhgap25 chr6:87459384–87533235 1.31 2.64 1.19 1.63 NS NS NS NS 
 Bean1 chr8:104170512–104219097 0.98 2.90 1.82 1.40 ** NS NS NS NS 
 Bex2 chrX:136066564–136068236 166.4 261.7 397.8 161.8 NS *** NS NS NS 
 Bok chr1:93685693–93695762 9.81 18.67 25.33 8.60 NS *** NS NS 
 C1ql2 chr1:120340581–120343174 0.84 0.88 6.15 0.36 NS *** NS *** NS ** 
 Cabyr chr18:12741354–12755142 1.44 3.54 1.81 1.73 NS NS NS NS 
 Calb2 chr8:110142537–110168206 2.64 3.38 20.26 2.36 NS *** NS *** NS *** 
 Camk2n2 chr16:20611600–20646462 25.84 69.20 50.19 33.05 ** NS NS NS 
 Car3 chr3:14863537–14872373 0.62 0.83 5.61 0.36 NS ** NS ** NS 
 Cbln1 chr8:87468852–87472592 1.78 2.79 6.02 1.47 NS ** NS NS 
 Ccdc135 chr8:95055102–95078141 3.61 0.42 1.58 0.88 *** ** NS NS 
 Ccdc136 chr6:29396427–29426995 12.24 23.01 33.47 13.60 NS *** NS NS NS 
 Cck chr9:121489823–121495689 122.7 365.6 258.2 134.5 *** ** NS NS ** NS 
 Cckbr chr7:105425819–105436338 6.02 13.39 4.45 7.88 ** NS NS *** NS NS 
 Col9a3 chr2:180598221–180642691 9.71 4.00 8.20 2.97 ** NS ** NS NS 
 Cox6a2 chr7:128173945–128206366 10.39 13.43 22.71 5.00 NS NS NS 
 Cplx3 chr9:57599991–57606281 1.77 4.45 1.59 2.40 ** NS NS ** NS NS 
 Cpne6 chr14:55510310–55517446 18.85 27.98 50.27 20.78 NS *** NS NS NS 
 Cpne7 chr8:123117373–123135185 3.94 5.12 17.78 3.75 NS *** NS *** NS *** 
 Crlf1 chr8:70493155–70504081 0.19 0.47 2.42 0.33 NS ** NS NS 
 Ctgf chr10:24595441–24598682 6.11 11.70 4.71 6.12 NS NS ** NS NS 
 Cyp2e1 chr7:140763831–140774977 0.18 0.39 8.39 0.60 NS ** NS ** NS ** 
 Dkkl1 chr7:45207524–45212152 3.96 15.75 2.61 4.80 *** NS NS *** ** NS 
 Emx1 chr6:85187930–85206015 2.63 9.24 2.39 3.56 ** NS NS ** NS 
 Epn3 chr11:94489598–94500270 1.05 0.25 1.71 0.29 ** NS ** NS 
 Fam183b chr11:58792801–58801960 3.60 1.03 6.29 1.60 NS NS NS NS 
 Fam81a chr9:70088500–70142424 14.41 30.47 14.91 21.89 ** NS NS *** NS NS 
 Fezf2 chr14:12341891–12345865 2.75 8.08 3.93 4.84 ** NS NS NS NS 
 Fxyd7 chr7:31032722–31051499 25.31 59.52 35.54 26.57 ** NS NS NS NS 
 Gbx2 chr1:89927961–89931176 0.58 0.42 4.60 0.36 NS *** NS ** NS 
 Gfra2 chr14:70890082–70979838 3.88 9.89 3.40 6.12 *** NS NS *** NS NS 
 Gh chr11:106300260– 301896 2.29 0.74 2.76 0.70 NS NS NS NS 
 Gm11549 chr3:36515056–36525258 4.74 15.38 1.98 7.21 *** NS *** 
 Gm14436 chr2:175470024–175483322 2.99 1.64 1.09 5.80 NS ** NS NS ** ** 
 Gm14446 chr19:34592887–34601968 0.42 1.19 0.32 0.66 NS NS NS NS 
 Gm3893 chr4:41889794–42462993 6.09 0.80 5.48 15.00 *** NS ** *** *** 
 Grm2 chr9:106644503–106656109 1.78 6.00 4.86 2.87 ** ** NS NS NS 
 Hkdc1 chr10:62383136–62422457 0.41 1.01 0.37 0.84 NS NS NS NS 
 Hs3st2 chr7:121392295–121502063 10.87 20.96 6.91 10.23 ** NS NS *** NS 
 Il12a chr3:68690643–68698547 0.35 1.27 0.16 0.43 NS NS NS NS 
 Kcnab3 chr11:69326257–69333041 7.91 21.47 10.65 10.64 *** NS NS ** NS 
 Kcng2 chr18:80294543–80364254 0.69 1.08 2.55 0.66 NS NS NS NS 
 Kcnq4 chr4:120696140–120747176 1.25 2.55 1.19 1.77 NS NS NS NS 
 Klk8 chr7:43797576–43803822 0.58 1.05 2.04 0.47 NS NS NS NS 
 Krt18 chr15:102028215–102032026 2.80 0.27 1.62 0.40 ** NS NS NS 
 Krt80 chr15:101348322–101370125 0.61 1.43 0.40 0.58 NS NS NS 
 Lars2 chr9:123366939–123462664 167.1 335.0 164.6 152.3 NS NS ** NS 
 Lbp chr2:158306492–158332852 4.04 0.92 2.17 1.28 *** NS ** NS NS 
 Lefty1 chr1:180935038–180938401 0.28 0.50 1.86 0.22 NS NS NS 
 Lrrc17 chr5:21483846–21645605 12.02 11.11 23.92 4.55 NS ** ** ** ** *** 
 Lrrc23 chr6:124769862–124779720 3.89 1.56 4.59 2.02 NS NS NS NS 
 Mapk11 chr15:89142483–89149629 4.87 10.69 4.54 5.97 ** NS NS *** NS NS 
 Mid1 chrX:169685246–169991234 20.20 61.93 24.41 11.22 *** NS NS *** NS 
 Mpped1 chr15:83780022–83858487 19.29 38.25 18.15 26.87 ** NS NS *** NS NS 
 Myl4 chr11:104502525–104595731 6.27 19.98 4.86 6.11 *** NS NS NS NS 
 Ndn chr7:62348276–62349927 49.07 112.8 61.25 39.45 *** NS NS NS *** NS 
 Necab3 chr2:154544387–154559006 14.49 38.38 17.71 17.89 *** NS NS *** NS 
 Neurod2 chr11:98325416–98329645 4.39 17.60 8.53 9.01 *** NS ** NS 
 Nov chr15:54745927–54753761 6.89 18.93 5.02 12.81 *** NS NS *** NS 
 Nrn1 chr13:36725624–36734738 27.79 78.38 56.59 33.71 *** ** NS NS NS 
 Ntn5 chr7:45677685–45694670 0.52 2.78 0.62 1.03 ** NS NS NS NS 
 Ntsr1 chr2:180499975–180544979 0.44 1.08 0.43 0.93 NS NS NS NS 
 Odf3b chr15:89377345–89379474 3.72 1.27 4.69 2.12 NS NS NS NS 
 Olfm1 chr2:28193092–28230736 193.0 439.6 270.6 205.6 ** NS NS NS NS 
 Olfm2 chr9:20667780–20746290 16.02 47.81 23.22 25.72 *** NS NS *** NS NS 
 Ovol2 chr2:144305175–144332080 0.79 2.84 0.44 1.08 ** NS NS ** NS NS 
 Pdzrn3 chr6:101149606–101377897 1.96 4.57 1.52 3.03 ** NS NS *** NS NS 
 Pnoc chr14:65400672–65425472 0.86 1.93 3.39 0.85 ** NS NS NS 
 Prkcd chr14:30595353–30626255 3.36 2.98 35.47 3.65 NS *** NS *** NS *** 
 Prss12 chr3:123446912–123506602 2.27 5.48 2.29 3.14 ** NS NS ** NS NS 
 Psrc1 chr3:108383803–108388231 0.71 2.49 0.92 0.98 ** NS NS NS 
 Rab37 chr11:115091430–115162522 1.28 1.36 7.63 1.13 NS *** NS *** NS ** 
 Ramp2 chr11:101246333–101248250 14.55 16.72 29.02 11.20 NS NS NS NS 
 Ramp3 chr11:6650147–6677475 5.74 5.10 22.35 4.11 NS *** NS *** NS ** 
 Resp18 chr1:75272201–75278593 23.71 45.55 39.27 22.10 ** NS NS NS NS 
 Rgs16 chr1:153740352–153745468 1.77 2.57 10.70 2.02 NS *** NS *** NS ** 
 Rpl27 chr11:101442244–101445596 94.92 157.3 193.6 85.81 NS ** NS NS NS 
 Rprm chr2:54084092–54085552 16.09 40.99 19.82 22.73 *** NS NS ** NS NS 
 Rsph1 chr17:31255019–31277356 5.46 2.01 5.92 2.25 NS NS NS 
 Rtn4r chr16:18127705–18152408 7.88 24.79 13.06 11.27 *** NS NS NS NS 
 Rtn4rl2 chr2:84871337–84887021 5.70 20.41 7.42 8.74 *** NS NS *** NS 
 Satb2 chr1:56793980–56975196 3.61 8.59 1.72 6.13 *** ** *** NS ** 
 Serpina3n chr12:104406707–104414329 2.59 4.08 7.68 2.94 NS ** NS NS NS 
 Serpini1 chr3:75557499–75642523 33.49 68.20 32.62 55.81 ** NS NS *** NS NS 
 Shox2 chr3:66971571–66981771 0.45 0.55 10.11 0.40 NS *** NS *** NS ** 
 Slc17a7 chr7:45163920–45176139 56.96 199.1 71.75 106.3 *** NS NS *** NS NS 
 Slc26a10 chr10:127172340–127180645 5.99 2.74 6.51 2.69 NS NS ** NS NS 
 Slc29a4 chr5:142702100–142722490 4.12 1.66 3.88 1.32 NS ** NS 
 Slc30a3 chr5:31086105–31093527 4.99 22.17 7.66 8.83 *** NS NS *** ** NS 
 Slc39a4 chr15:76612382–76617216 0.80 0.23 0.82 0.36 NS NS NS NS 
 Sostdc1 chr12:36314168–36318452 7.60 0.44 1.69 0.55 *** ** ** NS NS 
 Sstr1 chr12:58211803–58216036 1.45 4.38 1.11 3.74 ** NS ** NS 
 Sstr2 chr11:113618985–113626071 3.86 9.28 2.61 5.40 ** NS NS *** NS NS 
 Sstr3 chr15:78537014–78544345 1.11 3.11 1.53 1.81 ** NS NS NS NS 
 Stac2 chr11:98036623–98053462 8.09 19.49 8.16 11.00 *** NS NS *** NS NS 
 Stx1a chr5:135023571–135051099 19.32 54.54 18.87 28.70 *** NS NS *** NS NS 
 Tbr1 chr2:61804452–61814113 9.97 27.52 6.13 18.43 *** NS NS *** NS ** 
 Tgm3 chr2:130006385–130050399 0.06 0.49 0.08 0.33 NS NS NS NS 
 Tnnt1 chr7:4504662–4515975 3.09 2.53 16.78 1.28 NS *** NS *** NS ** 
 Tpm2 chr4:43514711–43524487 8.59 3.73 9.00 2.94 ** NS ** ** NS 
 Ttc9b chr7:27653923–27656207 46.52 97.31 47.70 36.33 ** NS NS *** ** NS 
 Ttr chr18:20665249–20674326 1520 63.89 304.8 49.22 *** *** *** *** NS *** 
 Tuba8 chr6:121210556–121226856 1.90 5.23 2.40 2.39 ** NS NS NS 
 Unnamed chr1:9548045–9631092 30.64 71.39 18.04 45.61 ** NS NS *** NS 
 Unnamed chr2:32379100–32381915 54.32 108.8 63.39 39.55 ** NS NS NS ** NS 
 Unnamed chr18:85071618–85083540 18.48 2.04 10.54 2.83 ** NS ** NS NS 
 Unnamed chr13:91368989–91388085 0.19 0.92 0.11 0.48 ** NS NS NS NS 
 Unnamed chr3:68869585–68872163 2.57 5.45 1.43 3.02 NS NS ** NS NS 
 Upb1 chr10:75406959–75440195 3.21 1.27 2.89 1.03 NS NS NS NS 
 Vip chr10:5639217–5647614 3.84 10.03 2.45 4.75 ** NS NS ** NS 
 Wnt10a chr1:74792018–74804175 1.28 3.28 1.22 1.33 NS NS NS 
 Wnt9a chr11:59306929–59333552 0.90 2.78 1.25 1.46 ** NS NS NS NS 
 Zic2 chr14:122475383–122480328 4.93 5.15 13.61 4.03 NS *** NS *** NS ** 
 Zic4 chr9:91368873–91389348 0.70 0.51 2.28 0.56 NS ** NS NS 
Gene dosage correlates Ccdc135 chr8:95055102–95078141 3.61 0.42 1.58 0.88 *** ** NS NS 
Slc29a4 chr5:142702100–142722490 4.12 1.66 3.88 1.32 NS ** NS 
 Sostdc1 chr12:36314168–36318452 7.60 0.44 1.69 0.55 *** ** ** NS NS 
 Ttr chr18:20665249–20674326 1520 63.89 304.8 49.22 *** *** *** *** NS *** 
Gene dosage discorrelates Abhd12b chr12:70154161–70183343 0.19 0.20 6.52 0.32 NS ** NS NS 
Adh1 chr3:138277644–138290691 0.79 0.28 3.75 0.22 NS ** NS ** NS 
 Agt chr8:124556586–124569707 5.59 3.96 12.65 4.09 NS ** NS *** NS 
 Alas2 chrX:150547416–150570622 3.86 2.73 1.16 0.89 NS ** ** NS 
 Aldob chr4:49535976–49549483 0.58 1.25 4.64 0.57 NS *** NS ** NS ** 
 Apoa1 chr9:46228629–46230466 0.39 0.24 25.59 0.08 NS ** NS NS NS 
 Apoa2 chr1:171225053–171226379 0.65 1.10 39.35 0.40 NS ** NS *** NS NS 
 Apoc1 chr7:19689480–19692658 2.80 1.73 29.69 1.09 NS *** NS ** NS 
 Apoc3 chr9:46233050–46235297 1.13 0.59 10.76 0.08 NS ** NS NS NS 
 Beta-s chr7:103826522–103827928 211.5 200.8 86.12 55.43 NS *** *** *** *** NS 
 Bex2 chrX:136066564–136068236 166.4 261.7 397.8 161.8 NS *** NS NS NS 
 Bok chr1:93685693–93695762 9.81 18.67 25.33 8.60 NS *** NS NS 
 Btg2 chr1:134074864–134079155 7.13 6.21 2.90 3.82 NS ** NS NS NS 
 C1ql2 chr1:120340581–120343174 0.84 0.88 6.15 0.36 NS *** NS *** NS ** 
 Calb2 chr8:110142537–110168206 2.64 3.38 20.26 2.36 NS *** NS *** NS *** 
 Camk2n2 chr16:20611600–20646462 25.84 69.20 50.19 33.05 ** NS NS NS 
 Car3 chr3:14863537–14872373 0.62 0.83 5.61 0.36 NS ** NS ** NS 
 Cbln1 chr8:87468852–87472592 1.78 2.79 6.02 1.47 NS ** NS NS 
 Ccdc136 chr6:29396427–29426995 12.24 23.01 33.47 13.60 NS *** NS NS NS 
 Cox6a2 chr7:128173945–128206366 10.39 13.43 22.71 5.00 NS NS NS 
 Cpne6 chr14:55510310–55517446 18.85 27.98 50.27 20.78 NS *** NS NS NS 
 Cpne7 chr8:123117373–123135185 3.94 5.12 17.78 3.75 NS *** NS *** NS *** 
 Crlf1 chr8:70493155–70504081 0.19 0.47 2.42 0.33 NS ** NS NS 
 Cyp2e1 chr7:140763831–140774977 0.18 0.39 8.39 0.60 NS ** NS ** NS ** 
 Cyp2f2 chr7:27119954–27133660 0.10 0.13 1.29 0.06 NS NS NS NS 
 Epn3 chr11:94489598–94500270 1.05 0.25 1.71 0.29 ** NS ** NS 
 Gbx2 chr1:89927961–89931176 0.58 0.42 4.60 0.36 NS *** NS ** NS 
 Gm11549 chr3:36515056–36525258 4.74 15.38 1.98 7.21 *** NS *** 
 Gm14436 chr2:175470024–175483322 2.99 1.64 1.09 5.80 NS ** NS NS ** ** 
 Gm3893 chr4:41889794–42462993 6.09 0.80 5.48 15.00 *** NS ** *** *** 
 Gpr4 chr7:19212537–19224176 1.24 1.62 3.41 1.68 NS NS NS NS 
 Hba-a1 chr11:32283671–32284493 99.48 82.94 35.18 25.72 NS *** *** *** *** NS 
 Hba-a2 chr11:32296488–32297310 0.04 12.53 0.00 0.00 NS ** ** ** *** NS 
 Hbb-b1 chr7:103812523–103813923 29.89 28.42 10.21 6.98 NS ** *** ** *** NS 
 Hp chr8:109575127–109579172 0.42 0.23 3.37 0.10 NS ** NS ** NS NS 
 Hpx chr7:105591610–105600116 0.29 0.34 5.30 0.10 NS ** NS ** NS NS 
 Kcng2 chr18:80294543–80364254 0.69 1.08 2.55 0.66 NS NS NS NS 
 Klk8 chr7:43797576–43803822 0.58 1.05 2.04 0.47 NS NS NS NS 
 Lef1 chr3:131110296–131224357 3.55 3.36 7.61 3.10 NS NS ** NS NS 
 Lefty1 chr1:180935038–180938401 0.28 0.50 1.86 0.22 NS NS NS 
 Lrrc17 chr5:21483846–21645605 12.02 11.11 23.92 4.55 NS ** ** *** ** *** 
 Nr4a1 chr15:101266845– 274794 58.45 40.20 19.74 32.31 NS *** NS *** NS NS 
 Ntf3 chr6:126101411–126166744 0.20 0.37 1.57 0.25 NS NS NS NS 
 Oprm1 chr10:6788600–7038209 0.09 0.14 0.16 0.80 ** NS NS *** NS ** 
 Pkp2 chr16:16213330–16272712 2.15 2.10 5.37 2.20 NS ** NS ** NS NS 
 Plekhd1 chr12:80692600–80724216 1.05 0.89 2.28 0.96 NS NS NS NS 
 Pnoc chr14:65400672–65425472 0.86 1.93 3.39 0.85 ** NS NS NS 
 Prkcd chr14:30595353–30626255 3.36 2.98 35.47 3.65 NS *** NS *** NS *** 
 Rab37 chr11:115091430– 162522 1.28 1.36 7.63 1.13 NS *** NS *** NS ** 
 Ramp2 chr11:101246333–248250 14.55 16.72 29.02 11.20 NS NS NS NS 
 Ramp3 chr11:6650147–6677475 5.74 5.10 22.35 4.11 NS *** NS *** NS ** 
 Rgs16 chr1:153740352–153745468 1.77 2.57 10.70 2.02 NS *** NS *** NS ** 
 Rpl27 chr11:101442244– 445596 94.92 157.3 193.6 85.81 NS ** NS NS NS 
 Satb2 chr1:56793980–56975196 3.61 8.59 1.72 6.13 *** ** *** NS ** 
 Serpina3n chr12:104406707–414329 2.59 4.08 7.68 2.94 NS ** NS NS NS 
 Shox2 chr3:66971571–66981771 0.45 0.55 10.11 0.40 NS *** NS *** NS ** 
 Slco2a1 chr9:103008488–103087959 0.16 0.15 0.83 0.18 NS NS NS NS 
 Snx31 chr15:36504061–36555572 0.18 0.11 0.94 0.26 NS NS NS NS 
 Sod3 chr5:52363803–52369738 8.14 7.31 3.73 3.27 NS ** NS NS 
 Spink8 chr9:109816625–109826754 0.61 0.67 4.34 0.29 NS NS NS NS 
 Tdo2 chr3:81958413–81975728 0.31 0.15 1.64 0.27 NS NS NS NS 
 Tnnt1 chr7:4504662–4515975 3.09 2.53 16.78 1.28 NS *** NS *** NS ** 
 Wnt3 chr11:103774174– 818021 1.10 1.13 2.67 1.18 NS NS NS NS 
 Zic2 chr14:122475383–480328 4.93 5.15 13.61 4.03 NS *** NS *** NS ** 
 Zic4 chr9:91368873–91389348 0.70 0.51 2.28 0.56 NS ** NS NS 
 Gene Locus WT FPKM 150 FPKM 315 FPKM 200 FPKM WT versus 150 WT versus 315 WT versus 200 150 versus 315 150 versus 200 315 versus 200 
Phenotypic correlates Alas2 chrX:150547416–150570622 3.86 2.73 1.16 0.89 NS ** ** NS 
Hba-a1 chr11:32283671–32284493 99.48 82.94 35.18 25.72 NS *** *** *** *** NS 
 Hbb-bs chr7:103826522–103827928 211.5 200.8 86.12 55.43 NS *** *** *** *** NS 
 Hbb-b1 chr7:103812523–103813923 29.89 28.42 10.21 6.98 NS ** *** ** *** NS 
 Igfbp2 chr1:72824479–72852471 26.53 13.43 12.87 6.57 NS *** NS NS 
 Oprm1 chr10:6788600–7038209 0.09 0.14 0.16 0.80 ** NS NS *** NS ** 
 Sod3 chr5:52363803–52369738 8.14 7.31 3.73 3.27 NS ** NS NS 
Phenotypic discorrelates Abhd12b chr12:70154161–70183343 0.19 0.20 6.52 0.32 NS ** NS NS 
Adam33 chr2:131050816–131063814 0.35 1.11 0.14 0.50 NS NS NS NS 
 Adcyap1 chr17:93199421–93205489 1.34 4.15 0.89 2.03 ** NS NS ** NS NS 
 Adh1 chr3:138277644–138290691 0.79 0.28 3.75 0.22 NS ** NS ** NS 
 Agt chr8:124556586–124569707 5.59 3.96 12.65 4.09 NS ** NS *** NS 
 Aldob chr4:49535976–49549483 0.58 1.25 4.64 0.57 NS *** NS ** NS ** 
 Apoc1 chr7:19689480–19692658 2.80 1.73 29.69 1.09 NS *** NS ** NS 
 Arhgap25 chr6:87459384–87533235 1.31 2.64 1.19 1.63 NS NS NS NS 
 Bean1 chr8:104170512–104219097 0.98 2.90 1.82 1.40 ** NS NS NS NS 
 Bex2 chrX:136066564–136068236 166.4 261.7 397.8 161.8 NS *** NS NS NS 
 Bok chr1:93685693–93695762 9.81 18.67 25.33 8.60 NS *** NS NS 
 C1ql2 chr1:120340581–120343174 0.84 0.88 6.15 0.36 NS *** NS *** NS ** 
 Cabyr chr18:12741354–12755142 1.44 3.54 1.81 1.73 NS NS NS NS 
 Calb2 chr8:110142537–110168206 2.64 3.38 20.26 2.36 NS *** NS *** NS *** 
 Camk2n2 chr16:20611600–20646462 25.84 69.20 50.19 33.05 ** NS NS NS 
 Car3 chr3:14863537–14872373 0.62 0.83 5.61 0.36 NS ** NS ** NS 
 Cbln1 chr8:87468852–87472592 1.78 2.79 6.02 1.47 NS ** NS NS 
 Ccdc135 chr8:95055102–95078141 3.61 0.42 1.58 0.88 *** ** NS NS 
 Ccdc136 chr6:29396427–29426995 12.24 23.01 33.47 13.60 NS *** NS NS NS 
 Cck chr9:121489823–121495689 122.7 365.6 258.2 134.5 *** ** NS NS ** NS 
 Cckbr chr7:105425819–105436338 6.02 13.39 4.45 7.88 ** NS NS *** NS NS 
 Col9a3 chr2:180598221–180642691 9.71 4.00 8.20 2.97 ** NS ** NS NS 
 Cox6a2 chr7:128173945–128206366 10.39 13.43 22.71 5.00 NS NS NS 
 Cplx3 chr9:57599991–57606281 1.77 4.45 1.59 2.40 ** NS NS ** NS NS 
 Cpne6 chr14:55510310–55517446 18.85 27.98 50.27 20.78 NS *** NS NS NS 
 Cpne7 chr8:123117373–123135185 3.94 5.12 17.78 3.75 NS *** NS *** NS *** 
 Crlf1 chr8:70493155–70504081 0.19 0.47 2.42 0.33 NS ** NS NS 
 Ctgf chr10:24595441–24598682 6.11 11.70 4.71 6.12 NS NS ** NS NS 
 Cyp2e1 chr7:140763831–140774977 0.18 0.39 8.39 0.60 NS ** NS ** NS ** 
 Dkkl1 chr7:45207524–45212152 3.96 15.75 2.61 4.80 *** NS NS *** ** NS 
 Emx1 chr6:85187930–85206015 2.63 9.24 2.39 3.56 ** NS NS ** NS 
 Epn3 chr11:94489598–94500270 1.05 0.25 1.71 0.29 ** NS ** NS 
 Fam183b chr11:58792801–58801960 3.60 1.03 6.29 1.60 NS NS NS NS 
 Fam81a chr9:70088500–70142424 14.41 30.47 14.91 21.89 ** NS NS *** NS NS 
 Fezf2 chr14:12341891–12345865 2.75 8.08 3.93 4.84 ** NS NS NS NS 
 Fxyd7 chr7:31032722–31051499 25.31 59.52 35.54 26.57 ** NS NS NS NS 
 Gbx2 chr1:89927961–89931176 0.58 0.42 4.60 0.36 NS *** NS ** NS 
 Gfra2 chr14:70890082–70979838 3.88 9.89 3.40 6.12 *** NS NS *** NS NS 
 Gh chr11:106300260– 301896 2.29 0.74 2.76 0.70 NS NS NS NS 
 Gm11549 chr3:36515056–36525258 4.74 15.38 1.98 7.21 *** NS *** 
 Gm14436 chr2:175470024–175483322 2.99 1.64 1.09 5.80 NS ** NS NS ** ** 
 Gm14446 chr19:34592887–34601968 0.42 1.19 0.32 0.66 NS NS NS NS 
 Gm3893 chr4:41889794–42462993 6.09 0.80 5.48 15.00 *** NS ** *** *** 
 Grm2 chr9:106644503–106656109 1.78 6.00 4.86 2.87 ** ** NS NS NS 
 Hkdc1 chr10:62383136–62422457 0.41 1.01 0.37 0.84 NS NS NS NS 
 Hs3st2 chr7:121392295–121502063 10.87 20.96 6.91 10.23 ** NS NS *** NS 
 Il12a chr3:68690643–68698547 0.35 1.27 0.16 0.43 NS NS NS NS 
 Kcnab3 chr11:69326257–69333041 7.91 21.47 10.65 10.64 *** NS NS ** NS 
 Kcng2 chr18:80294543–80364254 0.69 1.08 2.55 0.66 NS NS NS NS 
 Kcnq4 chr4:120696140–120747176 1.25 2.55 1.19 1.77 NS NS NS NS 
 Klk8 chr7:43797576–43803822 0.58 1.05 2.04 0.47 NS NS NS NS 
 Krt18 chr15:102028215–102032026 2.80 0.27 1.62 0.40 ** NS NS NS 
 Krt80 chr15:101348322–101370125 0.61 1.43 0.40 0.58 NS NS NS 
 Lars2 chr9:123366939–123462664 167.1 335.0 164.6 152.3 NS NS ** NS 
 Lbp chr2:158306492–158332852 4.04 0.92 2.17 1.28 *** NS ** NS NS 
 Lefty1 chr1:180935038–180938401 0.28 0.50 1.86 0.22 NS NS NS 
 Lrrc17 chr5:21483846–21645605 12.02 11.11 23.92 4.55 NS ** ** ** ** *** 
 Lrrc23 chr6:124769862–124779720 3.89 1.56 4.59 2.02 NS NS NS NS 
 Mapk11 chr15:89142483–89149629 4.87 10.69 4.54 5.97 ** NS NS *** NS NS 
 Mid1 chrX:169685246–169991234 20.20 61.93 24.41 11.22 *** NS NS *** NS 
 Mpped1 chr15:83780022–83858487 19.29 38.25 18.15 26.87 ** NS NS *** NS NS 
 Myl4 chr11:104502525–104595731 6.27 19.98 4.86 6.11 *** NS NS NS NS 
 Ndn chr7:62348276–62349927 49.07 112.8 61.25 39.45 *** NS NS NS *** NS 
 Necab3 chr2:154544387–154559006 14.49 38.38 17.71 17.89 *** NS NS *** NS 
 Neurod2 chr11:98325416–98329645 4.39 17.60 8.53 9.01 *** NS ** NS 
 Nov chr15:54745927–54753761 6.89 18.93 5.02 12.81 *** NS NS *** NS 
 Nrn1 chr13:36725624–36734738 27.79 78.38 56.59 33.71 *** ** NS NS NS 
 Ntn5 chr7:45677685–45694670 0.52 2.78 0.62 1.03 ** NS NS NS NS 
 Ntsr1 chr2:180499975–180544979 0.44 1.08 0.43 0.93 NS NS NS NS 
 Odf3b chr15:89377345–89379474 3.72 1.27 4.69 2.12 NS NS NS NS 
 Olfm1 chr2:28193092–28230736 193.0 439.6 270.6 205.6 ** NS NS NS NS 
 Olfm2 chr9:20667780–20746290 16.02 47.81 23.22 25.72 *** NS NS *** NS NS 
 Ovol2 chr2:144305175–144332080 0.79 2.84 0.44 1.08 ** NS NS ** NS NS 
 Pdzrn3 chr6:101149606–101377897 1.96 4.57 1.52 3.03 ** NS NS *** NS NS 
 Pnoc chr14:65400672–65425472 0.86 1.93 3.39 0.85 ** NS NS NS 
 Prkcd chr14:30595353–30626255 3.36 2.98 35.47 3.65 NS *** NS *** NS *** 
 Prss12 chr3:123446912–123506602 2.27 5.48 2.29 3.14 ** NS NS ** NS NS 
 Psrc1 chr3:108383803–108388231 0.71 2.49 0.92 0.98 ** NS NS NS 
 Rab37 chr11:115091430–115162522 1.28 1.36 7.63 1.13 NS *** NS *** NS ** 
 Ramp2 chr11:101246333–101248250 14.55 16.72 29.02 11.20 NS NS NS NS 
 Ramp3 chr11:6650147–6677475 5.74 5.10 22.35 4.11 NS *** NS *** NS ** 
 Resp18 chr1:75272201–75278593 23.71 45.55 39.27 22.10 ** NS NS NS NS 
 Rgs16 chr1:153740352–153745468 1.77 2.57 10.70 2.02 NS *** NS *** NS ** 
 Rpl27 chr11:101442244–101445596 94.92 157.3 193.6 85.81 NS ** NS NS NS 
 Rprm chr2:54084092–54085552 16.09 40.99 19.82 22.73 *** NS NS ** NS NS 
 Rsph1 chr17:31255019–31277356 5.46 2.01 5.92 2.25 NS NS NS 
 Rtn4r chr16:18127705–18152408 7.88 24.79 13.06 11.27 *** NS NS NS NS 
 Rtn4rl2 chr2:84871337–84887021 5.70 20.41 7.42 8.74 *** NS NS *** NS 
 Satb2 chr1:56793980–56975196 3.61 8.59 1.72 6.13 *** ** *** NS ** 
 Serpina3n chr12:104406707–104414329 2.59 4.08 7.68 2.94 NS ** NS NS NS 
 Serpini1 chr3:75557499–75642523 33.49 68.20 32.62 55.81 ** NS NS *** NS NS 
 Shox2 chr3:66971571–66981771 0.45 0.55 10.11 0.40 NS *** NS *** NS ** 
 Slc17a7 chr7:45163920–45176139 56.96 199.1 71.75 106.3 *** NS NS *** NS NS 
 Slc26a10 chr10:127172340–127180645 5.99 2.74 6.51 2.69 NS NS ** NS NS 
 Slc29a4 chr5:142702100–142722490 4.12 1.66 3.88 1.32 NS ** NS 
 Slc30a3 chr5:31086105–31093527 4.99 22.17 7.66 8.83 *** NS NS *** ** NS 
 Slc39a4 chr15:76612382–76617216 0.80 0.23 0.82 0.36 NS NS NS NS 
 Sostdc1 chr12:36314168–36318452 7.60 0.44 1.69 0.55 *** ** ** NS NS 
 Sstr1 chr12:58211803–58216036 1.45 4.38 1.11 3.74 ** NS ** NS 
 Sstr2 chr11:113618985–113626071 3.86 9.28 2.61 5.40 ** NS NS *** NS NS 
 Sstr3 chr15:78537014–78544345 1.11 3.11 1.53 1.81 ** NS NS NS NS 
 Stac2 chr11:98036623–98053462 8.09 19.49 8.16 11.00 *** NS NS *** NS NS 
 Stx1a chr5:135023571–135051099 19.32 54.54 18.87 28.70 *** NS NS *** NS NS 
 Tbr1 chr2:61804452–61814113 9.97 27.52 6.13 18.43 *** NS NS *** NS ** 
 Tgm3 chr2:130006385–130050399 0.06 0.49 0.08 0.33 NS NS NS NS 
 Tnnt1 chr7:4504662–4515975 3.09 2.53 16.78 1.28 NS *** NS *** NS ** 
 Tpm2 chr4:43514711–43524487 8.59 3.73 9.00 2.94 ** NS ** ** NS 
 Ttc9b chr7:27653923–27656207 46.52 97.31 47.70 36.33 ** NS NS *** ** NS 
 Ttr chr18:20665249–20674326 1520 63.89 304.8 49.22 *** *** *** *** NS *** 
 Tuba8 chr6:121210556–121226856 1.90 5.23 2.40 2.39 ** NS NS NS 
 Unnamed chr1:9548045–9631092 30.64 71.39 18.04 45.61 ** NS NS *** NS 
 Unnamed chr2:32379100–32381915 54.32 108.8 63.39 39.55 ** NS NS NS ** NS 
 Unnamed chr18:85071618–85083540 18.48 2.04 10.54 2.83 ** NS ** NS NS 
 Unnamed chr13:91368989–91388085 0.19 0.92 0.11 0.48 ** NS NS NS NS 
 Unnamed chr3:68869585–68872163 2.57 5.45 1.43 3.02 NS NS ** NS NS 
 Upb1 chr10:75406959–75440195 3.21 1.27 2.89 1.03 NS NS NS NS 
 Vip chr10:5639217–5647614 3.84 10.03 2.45 4.75 ** NS NS ** NS 
 Wnt10a chr1:74792018–74804175 1.28 3.28 1.22 1.33 NS NS NS 
 Wnt9a chr11:59306929–59333552 0.90 2.78 1.25 1.46 ** NS NS NS NS 
 Zic2 chr14:122475383–122480328 4.93 5.15 13.61 4.03 NS *** NS *** NS ** 
 Zic4 chr9:91368873–91389348 0.70 0.51 2.28 0.56 NS ** NS NS 
Gene dosage correlates Ccdc135 chr8:95055102–95078141 3.61 0.42 1.58 0.88 *** ** NS NS 
Slc29a4 chr5:142702100–142722490 4.12 1.66 3.88 1.32 NS ** NS 
 Sostdc1 chr12:36314168–36318452 7.60 0.44 1.69 0.55 *** ** ** NS NS 
 Ttr chr18:20665249–20674326 1520 63.89 304.8 49.22 *** *** *** *** NS *** 
Gene dosage discorrelates Abhd12b chr12:70154161–70183343 0.19 0.20 6.52 0.32 NS ** NS NS 
Adh1 chr3:138277644–138290691 0.79 0.28 3.75 0.22 NS ** NS ** NS 
 Agt chr8:124556586–124569707 5.59 3.96 12.65 4.09 NS ** NS *** NS 
 Alas2 chrX:150547416–150570622 3.86 2.73 1.16 0.89 NS ** ** NS 
 Aldob chr4:49535976–49549483 0.58 1.25 4.64 0.57 NS *** NS ** NS ** 
 Apoa1 chr9:46228629–46230466 0.39 0.24 25.59 0.08 NS ** NS NS NS 
 Apoa2 chr1:171225053–171226379 0.65 1.10 39.35 0.40 NS ** NS *** NS NS 
 Apoc1 chr7:19689480–19692658 2.80 1.73 29.69 1.09 NS *** NS ** NS 
 Apoc3 chr9:46233050–46235297 1.13 0.59 10.76 0.08 NS ** NS NS NS 
 Beta-s chr7:103826522–103827928 211.5 200.8 86.12 55.43 NS *** *** *** *** NS 
 Bex2 chrX:136066564–136068236 166.4 261.7 397.8 161.8 NS *** NS NS NS 
 Bok chr1:93685693–93695762 9.81 18.67 25.33 8.60 NS *** NS NS 
 Btg2 chr1:134074864–134079155 7.13 6.21 2.90 3.82 NS ** NS NS NS 
 C1ql2 chr1:120340581–120343174 0.84 0.88 6.15 0.36 NS *** NS *** NS ** 
 Calb2 chr8:110142537–110168206 2.64 3.38 20.26 2.36 NS *** NS *** NS *** 
 Camk2n2 chr16:20611600–20646462 25.84 69.20 50.19 33.05 ** NS NS NS 
 Car3 chr3:14863537–14872373 0.62 0.83 5.61 0.36 NS ** NS ** NS 
 Cbln1 chr8:87468852–87472592 1.78 2.79 6.02 1.47 NS ** NS NS 
 Ccdc136 chr6:29396427–29426995 12.24 23.01 33.47 13.60 NS *** NS NS NS 
 Cox6a2 chr7:128173945–128206366 10.39 13.43 22.71 5.00 NS NS NS 
 Cpne6 chr14:55510310–55517446 18.85 27.98 50.27 20.78 NS *** NS NS NS 
 Cpne7 chr8:123117373–123135185 3.94 5.12 17.78 3.75 NS *** NS *** NS *** 
 Crlf1 chr8:70493155–70504081 0.19 0.47 2.42 0.33 NS ** NS NS 
 Cyp2e1 chr7:140763831–140774977 0.18 0.39 8.39 0.60 NS ** NS ** NS ** 
 Cyp2f2 chr7:27119954–27133660 0.10 0.13 1.29 0.06 NS NS NS NS 
 Epn3 chr11:94489598–94500270 1.05 0.25 1.71 0.29 ** NS ** NS 
 Gbx2 chr1:89927961–89931176 0.58 0.42 4.60 0.36 NS *** NS ** NS 
 Gm11549 chr3:36515056–36525258 4.74 15.38 1.98 7.21 *** NS *** 
 Gm14436 chr2:175470024–175483322 2.99 1.64 1.09 5.80 NS ** NS NS ** ** 
 Gm3893 chr4:41889794–42462993 6.09 0.80 5.48 15.00 *** NS ** *** *** 
 Gpr4 chr7:19212537–19224176 1.24 1.62 3.41 1.68 NS NS NS NS 
 Hba-a1 chr11:32283671–32284493 99.48 82.94 35.18 25.72 NS *** *** *** *** NS 
 Hba-a2 chr11:32296488–32297310 0.04 12.53 0.00 0.00 NS ** ** ** *** NS 
 Hbb-b1 chr7:103812523–103813923 29.89 28.42 10.21 6.98 NS ** *** ** *** NS 
 Hp chr8:109575127–109579172 0.42 0.23 3.37 0.10 NS ** NS ** NS NS 
 Hpx chr7:105591610–105600116 0.29 0.34 5.30 0.10 NS ** NS ** NS NS 
 Kcng2 chr18:80294543–80364254 0.69 1.08 2.55 0.66 NS NS NS NS 
 Klk8 chr7:43797576–43803822 0.58 1.05 2.04 0.47 NS NS NS NS 
 Lef1 chr3:131110296–131224357 3.55 3.36 7.61 3.10 NS NS ** NS NS 
 Lefty1 chr1:180935038–180938401 0.28 0.50 1.86 0.22 NS NS NS 
 Lrrc17 chr5:21483846–21645605 12.02 11.11 23.92 4.55 NS ** ** *** ** *** 
 Nr4a1 chr15:101266845– 274794 58.45 40.20 19.74 32.31 NS *** NS *** NS NS 
 Ntf3 chr6:126101411–126166744 0.20 0.37 1.57 0.25 NS NS NS NS 
 Oprm1 chr10:6788600–7038209 0.09 0.14 0.16 0.80 ** NS NS *** NS ** 
 Pkp2 chr16:16213330–16272712 2.15 2.10 5.37 2.20 NS ** NS ** NS NS 
 Plekhd1 chr12:80692600–80724216 1.05 0.89 2.28 0.96 NS NS NS NS 
 Pnoc chr14:65400672–65425472 0.86 1.93 3.39 0.85 ** NS NS NS 
 Prkcd chr14:30595353–30626255 3.36 2.98 35.47 3.65 NS *** NS *** NS *** 
 Rab37 chr11:115091430– 162522 1.28 1.36 7.63 1.13 NS *** NS *** NS ** 
 Ramp2 chr11:101246333–248250 14.55 16.72 29.02 11.20 NS NS NS NS 
 Ramp3 chr11:6650147–6677475 5.74 5.10 22.35 4.11 NS *** NS *** NS ** 
 Rgs16 chr1:153740352–153745468 1.77 2.57 10.70 2.02 NS *** NS *** NS ** 
 Rpl27 chr11:101442244– 445596 94.92 157.3 193.6 85.81 NS ** NS NS NS 
 Satb2 chr1:56793980–56975196 3.61 8.59 1.72 6.13 *** ** *** NS ** 
 Serpina3n chr12:104406707–414329 2.59 4.08 7.68 2.94 NS ** NS NS NS 
 Shox2 chr3:66971571–66981771 0.45 0.55 10.11 0.40 NS *** NS *** NS ** 
 Slco2a1 chr9:103008488–103087959 0.16 0.15 0.83 0.18 NS NS NS NS 
 Snx31 chr15:36504061–36555572 0.18 0.11 0.94 0.26 NS NS NS NS 
 Sod3 chr5:52363803–52369738 8.14 7.31 3.73 3.27 NS ** NS NS 
 Spink8 chr9:109816625–109826754 0.61 0.67 4.34 0.29 NS NS NS NS 
 Tdo2 chr3:81958413–81975728 0.31 0.15 1.64 0.27 NS NS NS NS 
 Tnnt1 chr7:4504662–4515975 3.09 2.53 16.78 1.28 NS *** NS *** NS ** 
 Wnt3 chr11:103774174– 818021 1.10 1.13 2.67 1.18 NS NS NS NS 
 Zic2 chr14:122475383–480328 4.93 5.15 13.61 4.03 NS *** NS *** NS ** 
 Zic4 chr9:91368873–91389348 0.70 0.51 2.28 0.56 NS ** NS NS 

Classes defined in Materials and Methods. Asterisks indicate statistical significance calculated by Cuffdiff (*P < 0.05, **P < 0.01, ***P = < 0.001, NS = not significant).

Table 4.

Transcript-level classification by ordering across allelic series

Allowed variation υ (%) 12 15 18 21 
Order: phenotypic severity 
 No effect 10 17 24 31 38 
 Partial correlate 17 23 28 30 30 28 
 Step-wise correlate 0.5 0.3 0.3 0.2 
 Discorrelate 95 88 77 66 55 46 39 34 
Order: mutant gene dosage 
 No effect 14 21 28 35 
 Partial correlate 14 25 34 40 44 45 44 41 
 Step-wise correlate 1.5 0.8 0.5 0.4 0.2 0.2 0.1 
 Discorrelate 83 73 62 51 41 34 28 24 
Allowed variation υ (%) 12 15 18 21 
Order: phenotypic severity 
 No effect 10 17 24 31 38 
 Partial correlate 17 23 28 30 30 28 
 Step-wise correlate 0.5 0.3 0.3 0.2 
 Discorrelate 95 88 77 66 55 46 39 34 
Order: mutant gene dosage 
 No effect 14 21 28 35 
 Partial correlate 14 25 34 40 44 45 44 41 
 Step-wise correlate 1.5 0.8 0.5 0.4 0.2 0.2 0.1 
 Discorrelate 83 73 62 51 41 34 28 24 

Shown are the percentages of 15 543 genes analyzed by RNA sequencing. All transcripts from a single gene combined and classified by relative FPKM between WT, HDQ150/150, HDQ315/+ and HDQ200/200 mice. Class definitions are described in Materials and Methods. For all υ > 21%, no effect category increase at the expense of correlates and discorrelates.

Figure 6.

Expression correlates and discorrelates revealed by RNA-sequencing analysis. (A–F) Ordering of lines based on increasing phenotypic severity. Darker shading in arrows represents increased phenotypic severity. Bars represent striatal FPKM values of all isoforms of each gene from left to right of WT (open), HDQ150/150 (black), HDQ315/+ (gray) and HDQ200/200 (black). Definitions of correlate and discorrelate classes provided in Materials and Methods. (A–C) Examples of mRNAs with levels that correlate with phenotypic severity. (D–F) Examples of mRNAs with levels that discorrelate with phenotypic severity. (G–L) Ordering of lines based on increasing mutant gene dosage. Numbers in arrows indicate mutant gene dosage. Bars represent striatal FPKM values of all isoforms of each gene from left to right of WT (open), HDQ315/+ (gray), HDQ200/200 (black) and HDQ150/150 (black). Definition of correlate and discorrelate classes provided in Materials and Methods. (G–I) Examples of mRNAs with levels that correlate with mutant gene dosage. (J–L) Examples of mRNAs with levels that discorrelate with mutant gene dosage. Asterisks indicate p values calculated by Cuffdiff software based on dispersion model of variances (*P < 0.05, **P < 0.01, ***P < 0.001). Table 3 contains a complete list of all correlates and discorrelates that have statistical support.

Figure 6.

Expression correlates and discorrelates revealed by RNA-sequencing analysis. (A–F) Ordering of lines based on increasing phenotypic severity. Darker shading in arrows represents increased phenotypic severity. Bars represent striatal FPKM values of all isoforms of each gene from left to right of WT (open), HDQ150/150 (black), HDQ315/+ (gray) and HDQ200/200 (black). Definitions of correlate and discorrelate classes provided in Materials and Methods. (A–C) Examples of mRNAs with levels that correlate with phenotypic severity. (D–F) Examples of mRNAs with levels that discorrelate with phenotypic severity. (G–L) Ordering of lines based on increasing mutant gene dosage. Numbers in arrows indicate mutant gene dosage. Bars represent striatal FPKM values of all isoforms of each gene from left to right of WT (open), HDQ315/+ (gray), HDQ200/200 (black) and HDQ150/150 (black). Definition of correlate and discorrelate classes provided in Materials and Methods. (G–I) Examples of mRNAs with levels that correlate with mutant gene dosage. (J–L) Examples of mRNAs with levels that discorrelate with mutant gene dosage. Asterisks indicate p values calculated by Cuffdiff software based on dispersion model of variances (*P < 0.05, **P < 0.01, ***P < 0.001). Table 3 contains a complete list of all correlates and discorrelates that have statistical support.

Graphical examples of transcript levels with respect to mutant gene dosage are shown in Figure 6G–L. Gene dosage ordering provides similar distribution among classes as ordering by phenotypic severity (Table 4). The high numbers of partial correlates for a gene dosage ordering suggests that some mRNA-level alterations are more influenced by the concentration of mutant HD gene product than by the toxicity of longer repeat lengths.

We also used QRTPCR to examine three molecular markers for HD commonly used in mouse models. The gene products for cannabinoid receptor 1 (Cnr1), the dopamine receptor D2 (Drd-2), and the protein phosphatase 1 regulatory subunit 1B (Darpp32) were shown to be lowered in HD mice prior to the onset of motor and behavioral abnormalities (18–20) Consistent with prior studies, transcript-level reductions compared with wild type were shown for all mice with repeat lengths >150 CAGs (Fig. 7). Each marker has a similar mRNA-level profile across the allelic series with WT having high levels of marker mRNA and HDQ200/200 having the least. Each also has a statistically significant spike for HDQ315/+ indicating their mRNA levels discorrelated with phenotypic severity. Thus the degree of these marker mRNA reductions at 20 weeks is not a predictor of age of onset which has implications for therapeutic strategies designed to normalize their expression levels. This analysis shows the advantage of the use of an allelic series in assessing the contribution of specific molecular alterations to pathology.

Figure 7.

Levels of striatal mRNA from three commonly used markers of HD in mice. Bars represent mean mRNA levels relative to WT as determined by QRTPCR using is β-actin as an internal control. Darker shading in arrows represents increased phenotypic severity. Open bars indicate WT, grey heterozygous and black homozygous for expanded CAG repeat allele. The number of mice is shown at the base of each bar. RNA from each mouse analyzed individually to provide directly calculated statistics. Error bars show the standard error of the mean, asterisks above brackets indicate statistical significance by ANOVA with Tukey–Kramer multiple comparisons test. Asterisks directly above bars indicate significant difference from WT based on ANOVA with Dunnett multiple comparisons test (*P < 0.05, **P < 0.01 and ***P < 0.001).

Figure 7.

Levels of striatal mRNA from three commonly used markers of HD in mice. Bars represent mean mRNA levels relative to WT as determined by QRTPCR using is β-actin as an internal control. Darker shading in arrows represents increased phenotypic severity. Open bars indicate WT, grey heterozygous and black homozygous for expanded CAG repeat allele. The number of mice is shown at the base of each bar. RNA from each mouse analyzed individually to provide directly calculated statistics. Error bars show the standard error of the mean, asterisks above brackets indicate statistical significance by ANOVA with Tukey–Kramer multiple comparisons test. Asterisks directly above bars indicate significant difference from WT based on ANOVA with Dunnett multiple comparisons test (*P < 0.05, **P < 0.01 and ***P < 0.001).

Expanded repeat HD mutations effect processing and/or degradation rather than transcription alone

Three factors dictate mRNA levels: transcription, processing (e.g. splicing) and degradation rates. The ability of RNA sequencing to differentiate between isoforms from a single promoter provides a means of assessing the hypothesis that most transcript levels are altered by changing transcription rather than processing or degradation. This is achieved by considering all genes with more than one isoform transcribed from a single promoter. At steady state, these factors are related to each other by the following equation: KT × Pi = KD × FPKMi, where KT is the rate of production of full-length primary transcript in FPKM produced per sec, Pi is the proportion of the mRNA that is processed to that particular isoform, KD is the rate of degradation per second and FPKMi is the measured level of the specified particular isoform. For the major isoform (defined as the isoform with the highest FPKM from its promoter in the WT line) the equation becomes KT × Pmajor = KD,major × FPKMmajor. For all other isoforms from the same promoter, minor isoforms, the equation becomes KT × Pminor = KD,minor × FPKMminor. Dividing the two equations leads to the following relation FPKMminor/FPKMmajor = (KD,major × Pminor)/(KD,minor × Pmajor) a ratio we term the P/D ratio because it is dependent only on processing and degradation and independent of the rate of transcription from that promoter. Equivalent P/D ratios for any gene's transcripts between lines might be achieved if there is a change only in transcription rate or if expanded CAG repeats effect degradation and or processing rates equally for the compared isoforms. Both scenarios leading to equivalent P/D ratios are easily imagined. In contrast, a change in P/D ratio for the same transcript in two mouse lines indicates one or more of the factors KD,major, Pminor, KD,minor and Pmajor is altered between lines. Thus different P/D ratios for a transcript in different mouse lines indicate expanded CAG repeats influence processing and/or degradation rates. P/D ratios were calculated for 7434 minor transcripts from 5950 genes. Table 5 shows the proportion of transcripts changed in level where P/D ratios also change across a range of allowed variances. A majority of isoforms from genes changed in transcript level between lines show alteration in P/D ratio between WT and each mutant studied. Therefore, expanded CAG repeats usually alter transcript levels with involvement of processing and/or degradation rates rather than transcription rate alone.

Table 5.

Transcript-level changes with processing and/or degradation rate alteration

Allowed variation υ (%) 10 15 20 30 40 50 60 70 80 90 
WT versus HDQ150/150 
 Minor isoforms changed in level 6215 5161 4272 3475 2366 1692 1253 973 811 726 677 
P/DWTP/Dmut (%) 80 68 61 56 52 54 61 72 81 87 83 
WT versus HDQ315/+ 
 Minor isoforms changed in level 6470 5532 4620 3843 2661 1864 1338 1016 814 716 649 
P/DWTP/Dmut (%) 83 73 68 62 55 53 58 67 78 84 84 
WT versus HDQ200/00 
 Minor isoforms changed in level 6478 5535 4662 3882 2714 1930 1422 1100 894 768 676 
P/DWTP/Dmut (%) 82 71 64 59 53 52 56 63 73 81 84 
Allowed variation υ (%) 10 15 20 30 40 50 60 70 80 90 
WT versus HDQ150/150 
 Minor isoforms changed in level 6215 5161 4272 3475 2366 1692 1253 973 811 726 677 
P/DWTP/Dmut (%) 80 68 61 56 52 54 61 72 81 87 83 
WT versus HDQ315/+ 
 Minor isoforms changed in level 6470 5532 4620 3843 2661 1864 1338 1016 814 716 649 
P/DWTP/Dmut (%) 83 73 68 62 55 53 58 67 78 84 84 
WT versus HDQ200/00 
 Minor isoforms changed in level 6478 5535 4662 3882 2714 1930 1422 1100 894 768 676 
P/DWTP/Dmut (%) 82 71 64 59 53 52 56 63 73 81 84 

Table represents 7433 minor isoforms from 5950 genes. Major isoforms are not included in this table since they are used as a denominator in calculating P/D ratios, the processing–degradation ratio calculated as described in text. Minor isoforms changed in level do not have overlapping values when variation is added to the lesser of the two FPKM values (FPKMWT and FPKMmut) and subtracted from the greater. Variation was used to calculate a range for each P/D ratio with a maximum (FPKMminor+ υminor)/(FPKMmajorυmajor) and a minimum.

(FPKMminorυminor)/FPKMmajor+ υmajor). Overlapping ranges indicate that P/DWT = P/Dmut. Percentages indicate the number of isoforms changed in level with non-equivalent P/D ratios divided by the total number of isoforms changed in level. P/D ratios tend to be equivalent for transcripts that are not changed in levels between lines (χ2P < 0.0001 for comparisons between unchanged group and every changed group presented in this table).

Discussion

The analysis of trends across an allelic series reveals several molecular features that are difficult or impossible to determine by multiple mutant versus wild-type comparisons. One feature is the lessening of HD mRNA levels with longer repeats. This effect is also seen for very long CAG repeats in the R6/2 transgene (3–5). Our analysis shown in Figure 2 revealed an unexpected feature of the R6 series. The R6/2Q50 brain transgene mRNA level is 3-fold less than for the R6/2Q110 line. There are several potential explanations for this result. One is that the extra 60 CAGs provide greater stability for the R6/2Q110 transgene mRNA. This does not seem to be the case for the full-length mRNAs from the endogenous locus, as shown in Figure 2B, where the brain knock-in HDQ50 mRNA is greater in level than that of the HDQ100. Alternatively, the location of the transgene might interact with repeat length allowing for an increase of transcription rate by the addition of 60 CAGs. One potential molecular explanation for this scenario is that site of transgene insertion (which is the same for both R6/2Q50 and R6/2Q110) might be in a relatively closed chromatin configuration at 50 CAGs compared with 110 CAGs. This feature is not shared by the endogenous locus suggesting a different potential for transcription at the R6/2 location versus endogenous HD.

The allelic series also provides a more powerful means of screening potential driver/responder/biomarker candidates than standard mutant versus wild-type analyses. Enhanced screening methods are needed given the thousands of transcript-level alterations associated with HD (21). There are practical considerations to be taken into account when using an allelic series for such a screen. First, given the present cost of RNA-sequencing, the addition of several additional lines to make up the series may limit the number of biological replicates that can be performed for an initial screen. This shortcoming is partially compensated by each mutant line in the series being a graded replicate. The strategy of using a low-cost screen with few biological replicates to identify candidate transcripts should be followed by confirmation with a large numbers of biological replicates. Figure 7 shows the use of 9–12 biological replicates for each line in the series using QRTPCR to determine striatal levels of Drd-2, Darpp32 and Cnr1 mRNAs. A second consideration is whether heterozygous and homozygous lines should be compared as members of the same series. Heterozygotes have a single gene dose of an expanded repeat allele and a dose of the wild-type allele. Homozygotes on the other hand have two gene doses of the expanded allele and no wild type. These genetic differences might influence transcript profiles in a series that mixes heterozygous and homozygous members. In this initial study, we chose members solely based on their phenotypic severity. The comparison HDQ315/+ heterozygotes (one mutant, one wild-type dose) to wild type (zero mutant, two wild-type gene doses) to the homozygous HDQ150/150 and HDQ200/200 lines (two mutant, zero wild-type gene doses) was used here to show that gene dosage correlates better than phenotypic severity with the levels of some transcripts (Table 4 and Fig. 6G–I). The presence of the wild-type allele or gene dosage of the mutant allele might also be responsible for the HDQ315/+ line breaking the downward trend for Drd-2, Darpp32 and Cnr1 mRNAs levels shown in Figure 7. The validity of comparison between heterozygous and homozygous series members is strengthened by a previous study showing 94% concordance between transcript-level alterations in HDQ150/150 mice (which lack WT HD) and R6/2-Q150 mice (which have wild-type endogenous HD) (9) and the qualitative similarity of the HD-like phenotypes seen in heterozygous and homozygous members of this series. A third practical consideration is the determination of which tissue should be used for RNA sequence analysis. In the case of HD, several brain regions are known to be affected by the expanded repeat mutation (1). In this initial study, we have chosen to determine the effects of the HD repeat mutation on striatal mRNAs, since it is considered to be the most affected brain region in HD. The analyses of other brain regions and integration of results with those of the striatum will be needed to provide a more complete picture of the effects of expanded repeat mutations in these HD mouse models.

There are numerous plausible theories as to the nature of differentially expressed genes associated with disease states. A change in any particular transcript's level might be a downstream effect coincident with but not causally related to pathology. Reproducible members of this class might serve as valuable biomarkers of pathology but not as candidates for therapeutic intervention. Alternately, a transcript's level change may be a protective response to the genetic insult of the expanded repeat in HD. Finally, a transcript's level change might be responsible for some or all of the pathology. Such drivers and protective responders might then be altered in level for therapeutic purposes. There are several potential pitfalls in the path towards a complete classification of transcript levels in HD. The allelic series provides a means of eliminating some of the major experimental pitfalls, most importantly the elimination of false positives caused by normal variation in experiments which assess large numbers of transcripts simultaneously. The allelic series also provides a means of differentiating those candidates where degree of transcript-level alteration follows degree of phenotypic severity (step-wise correlates) from those that might be subject to threshold effects (partial correlates).

Previous transcript profiling of HD mouse models and postmortem tissues from HD patients has provided lists of genes with consistent expression changes in HD. We took one such list (Table 3 of (21)) and found, of the 149 listed mRNAs, 100 genes met the criteria for inclusion in our RNA-sequencing analysis. Table 6 shows an overview of these HD-related gene expression changes. Surprisingly, there are several discorrelates amongst the HD-related transcripts. This discrepancy may be due to methodological differences given previous studies relied on HD versus normal RNAs assessed with chip hybridization methods. Overall, however, selecting these HD-related candidates enriches the correlate class at the expense of discorrelates showing a concordance with this and previous studies.

Table 6.

Classification for genes with transcript levels previously shown to be altered in HD

Allowed variation υ (%) 12 15 18 21 
Order: phenotypic severity 
 No effect 11 23 33 41 
 Partial correlate 10 20 32 34 41 41 44 
 Step-wise correlate 
 Discorrelate 96 87 76 63 55 36 26 15 
Order: mutant gene dosage 
 No effect 11 17 26 36 
 Partial correlate 32 49 54 67 65 67 63 57 
 Step-wise correlate 
 Discorrelate 62 48 41 28 23 16 11 
Allowed variation υ (%) 12 15 18 21 
Order: phenotypic severity 
 No effect 11 23 33 41 
 Partial correlate 10 20 32 34 41 41 44 
 Step-wise correlate 
 Discorrelate 96 87 76 63 55 36 26 15 
Order: mutant gene dosage 
 No effect 11 17 26 36 
 Partial correlate 32 49 54 67 65 67 63 57 
 Step-wise correlate 
 Discorrelate 62 48 41 28 23 16 11 

Shown are the percentages in each class of 100 genes with transcript levels previously shown to be altered in HD (from Table 3 of Seredenina and Luthi-Carter (21)). Class definitions are described in Materials and Methods. All transcripts from a single gene combined and classified by relative FPKM between WT, HDQ150/150, HDQ315/+ and HDQ200/200 mice. For all υ > 21%, no effect category increase at the expense of correlates and discorrelates.

Despite this concordance, literature searches with our best driver/responder/biomarker candidates fail to show any previously published work on these genes with regard to HD. As such the seven statistically supported phenotypic correlates presented in Table 3 represent new potential biomarkers or therapeutic targets. There is one neurotransmitter receptor, the μ-type opioid receptor (Oprm1), which is pharmacologically tractable (22). One has antioxidant function, the extracellular superoxide dismutase [Cu–Zn] (Sod3) whose activity might be replaced by chemical free-radical scavengers (23). One is involved in neuroprotection, the insulin-like growth factor binding protein 2 (Igfpb2) (24) and the other four are involved in hemoglobin production. Hbb-bs and Hbb-b1 code for hemoglobin β chains while Hbb-a1 codes the α chain. Alas2 codes for the first step in the heme biosynthetic pathway. Although classically considered erythroid specific, expression of hemoglobin has been shown to occur in rodent striatal neurons isolated by laser-capture microdissection (25). Blood contamination in our experiment was ruled out by assessing the levels of other blood mRNAs Rhag, Spta1 and Epb4.2 which were either absent or at least 6-fold below our lowest FPKM threshold for inclusion in this study (FPKM < 0.052). Although these seven candidates might be molecular drivers of HD pathology, they might also be responding to the presence of the HD mutation in either a beneficial or neutral manner.

Despite the advantages of the allelic series, its use to provide positive correlative results, as most studies do, would be subject to the logical problems of traditional inductivist methods of scientific inquiry (26). The new discorrelate class, however, provides a means of interpretation that does not rely on induction, rather falsification of the hypothesis that the expression level of a gene drives a pathological process. A discorrelate under any one set of conditions despite correlation under many other conditions argues against that change being a sole driver of pathology. More complex hypotheses involving threshold expression levels sufficient to cause pathology and combinations of expression alterations of groups of genes would also be falsifiable.

By seeking discorrelates, the potential for eliminating false positives becomes additive between different laboratory environments and conditions. The number of different conditions that might be applied to reveal discorelates is already quite large given the many attempted treatment regimens that have little effect on pathological severity. Such treatments may reveal further discorrelates by having disparate effects on gene expression levels.

The identification of pathological discorrelates is a particularly attractive approach to eliminating pathologically neutral alterations leaving a short list of potential drivers of pathology when one considers there are a finite number of processes, metabolites, proteins or in this case transcripts in human and mouse tissues. For example, the number of protein coding mRNAs in the entire mouse genome is estimated to be ∼30 000 (27). Since these can now be queried en-mass, analysis of the entire transcriptome to reveal a complete short list of therapeutically relevant changes is within reach. The addition of the simultaneous assessment of several members of an allelic series adds the power of enhanced discrimination to this approach. This work represents a first step of its type towards ultimately saturating the transcriptome to determine which, if any, are therapeutically relevant molecular alterations. This approach is applicable to any measurable molecular level or process for any disease where a graded pathology exists or can be experimentally produced.

Materials and Methods

Generation of allelic series

The original HD knock-in line (12) was backcrossed to C57BL/6J mice for 10 generations to establish congenic lines containing the expanded repeat mutation. Germline instability of the CAG repeat number was exploited to choose longer and shorter CAG repeat lengths for breeding of individual lines with different repeat lengths. To confirm that the lines used in this study were congenic to C57BL/6J we analyzed our RNA sequencing results by alignment to the C57BL/6J genome (mm10 from UCSC) using the aligner STAR (version 2.5.0b) in 2-pass mapping mode. Variants were called using GATK (version 3.5) following their Best Practices Guide. Sequences that differed from C57BL/6J were considered candidates for comparisons between each mutant and the wild type line. These candidates were well dispersed throughout the mouse genome representing every chromosome with an average distance between candidates of 1.2 million bases. The HDQ200/200 line had 2 of 403 candidates with sequences differing from the WT line. The HDQ150/150 line had 1 of 359 and the HDQ315/+ line had 0 sequence differences of 392 candidates. This analysis shows >99.5% sequence identity between the C57BL/6J wild-type line and each mutant line as well as only three sequence variations between mutants and C57BL/6J wild-type controls in over 25 million reads per line. These results are consistent with all mutant lines being congenic to C57BL/6J.

Motor and behavioral assessment

Groups of 32 HDQ315/+ (16 male, 16 female) and 26 WT (15 male, 11 female) were longitudinally assessed at 10, 20, 30, 40, 50 and 65 weeks of age. No major differences were seen between males and females for the tests shown so their data were combined. Rotarod analysis was performed once a day for three training days followed by 5 days of testing on a gradually accelerating Ugo Basile 7650 from (4–40 RPMs during first 5 min of a 10 min trial). Starting at 30 weeks of age mice were tested in a Bioseb Automatic foot misplacement apparatus (EB-Instruments, Tampa, FL). Mice walked on a 79 cm horizontal ladder (1 rung/cm) through a lit corridor (50 mm wide with opaque walls) to a shaded area and the number of missteps and time to traverse ladder was averaged over three trials. In cage activity was determined as we previously described (28) with 24 h of acclimation followed by 3 days and nights of data acquisition which was averaged for each mouse. Open field test was performed as previously described (28) in a 50 × 50 cm square arena with walls and floor made of plexiglass. Distance traveled and position during 4 min trials was determined by computerized video recording and the Ethovision 3.1 software (Noldus Information Technology, Wageningen, the Netherlands). The voluntary wheel cage (Lafayette Neuroscience, Lafayette, IN, USA) was used according to manufacturer's instructions with 1 day of acclimation was followed by 3 days and nights of data acquisition which was averaged for each individual mouse. Grip strength was measured as previously described by us (28) using a Chatillon grip strength meter (Wagner Instruments, Greenwich, CT, USA). All procedures were conducted in compliance with the Guide for the Care and Use of Laboratory Animals as adopted by the NIH and approved by the Institutional Animal Care and Use Committee of the University of Alabama at Birmingham.

Neuroanatomical analyses

Receptor binding autoradiography for D1 and D2 dopamine receptors was performed as we previously described (14). Immunocytochemistry to detect HD aggregates was performed on three 70-week-old HDQ315/+ mice and an age matched wild-type control. Brains fixed overnight in 4% paraformaldehyde were cryoprotected and stored at −80°C; 30 μm coronal sections were cut using a freezing, sliding microtome. Sections were stained using the HD N-18 primary antibody (8767, Santa Cruz Biotechnology, Santa Cruz, CA, USA) and a Cy3 conjugated anti-goat IgG antibody (705-165-003, Jackson ImmunoResearch, West Grove, PA, USA). The stained sections of the striatum were digitized using an Olympus DP73 camera, and images were taken using Olympus fluorescence microscope with 552–565 nm filter (WIG cube) for Cy3-antibody detection. An inverted grayscale image of antibody fluorescence was created using Adobe Photoshop software for Figure 4G and H. Stereology was performed as in Kumar et al. (30).

Analyses of DNA, mRNA and protein

Genotyping was performed by PCR across the repeats using reactions with final concentrations indicated: 10 ng/μl each primers CCCATTCATTGCCTTGCTG and GCGGCTGAGGGGGTTGA, 2 m betaine, 400 μm dNTPs, 15 mm Tris–HCl, pH 8.9, 8 mm ammonium sulfate, 75 μg/ml BSA, 1.25 mm MgCl and 0.5 mm β-mercaptoethanol and 2.0 ng/μl tail biopsy DNA and 0.04 U/μl LA Taq (TaKaRa). Reactions were carried by incubating 94°C 5 min, then 30 cycles of 94°C 30 s, 53°C 30 s, 72°C 3 min and following cycling 72°C 5 min. PCR products were analyzed by running on long 1% agarose gels, stained with ethidium bromide and visualized with UV light. All samples were compared with standards sized independently by Laragen, Inc. (Los Angeles, CA, USA) and mice >15 CAGs from target length were excluded. R6 mice were genotyped by Laragen. Allele-specific HD QRTPCR was previously described by us in 15.

RNA preparation was performed as previously described (29). cDNA for QRTPCR was made using 200 ng of sample RNA by the Applied Biosystems Reverse Transcription kit according to manufacturer instructions (Life Technologies, Grand Island, NY, USA). Assays for markers used pre-made Taqman primers and probes and mastermix from Appliedbiosystems Cnr1 (Mm01212171), Drd-2 (Mm00438545), DARPP32 (Mm00454892) and mouse β-actin control (4352341E). HD QRTPCR was as previously described (29). R6 transgene mRNA levels were determined by QRTPCR as we previously described (30) on RNA extracted from single hemispheres of brain at 6 weeks of age. All QRTPCR assays were performed in an ABI Prism 7900HT (Life Technologies, Grand Island, NY, USA). Groups included from 9 to 12 mice aged 24 ± 4 weeks. Western analysis was performed as previously described (30) but with 4–8% gradient polyacrylamide gels using primary antibodies against HD (MAB2166, Millipore, Darmstadt, Germany), polyglutamine (MAB1574, Millipore) and α-tubulin (T5168, Sigma, St Louis, MO, USA).

For RNA-sequencing analysis striatal mRNA samples of each genotype were pooled such that each mouse contributed equal amounts of total RNA to one of the four different groups: (i) WT n = 12 age 21.7 ± 0.1 weeks, (ii) HDQ150/150 n = 6 age 20.2 ± 0.1 weeks, (iii) HDQ315/+ n = 8 age 21.1 ± 0.6 weeks and (iv) HDQ200/200 n = 10 age 21.1 ± 0.6 weeks. Each group had an equal number of males and females. RNA quality was confirmed using the Agilent 2100 Bioanalyzer according to manufacturer instructions followed by two rounds of mRNA purification using oligo dT magnetic beads and cDNA synthesis using SureSelect Strand-specific RNA Library generation kit (Agilent, Santa Clara, CA, USA). cDNA production included fragmentation, repair of ends, A-tailing and the ligation of separate barcode identifiers for each of the four samples. cDNA libraries were quantitated using qPCR in a Roche Lightcycler 480 with a kit for library quantitation (Kapa Biosystems, Woburn, MA, USA). Paired end 2 × 50 bp sequencing runs were performed in the Illumina HiSeq2500. Sequence data were converted to FASTQ Sanger format using Illumina's bcl2fastq version 1.8.4 and aligned to the University of California, Santa Cruz mouse mm10 genome using TopHat version 2.0.11 and the short read aligner Bowtie version 2.1.0. Cufflinks version 2.1.1. was used to assemble transcripts and estimate abundances. Cuffdiff was used to find significant changes in transcript level for all six possible pairwise comparisons of the four samples (shown in Table 3).

Further refinement of the data was carried out in Excel spreadsheets. A lower cutoff for FPKM values of 0.3334 was established using Cuffdiff readouts by determining the minimum FPKM value that provided a status of OK (rather than NOTEST) when compared with a zero FPKM value for each of the six pairwise comparisons and averaging the six. To insure 5′ ends were represented fully, we defined 103 groups based on length of transcript and plotted the average counts of each group against median length of each group. In theory, longer transcripts should have more counts, however, we found a plateau starting after 4186 bp. Visual inspection of number of reads on longer transcripts confirm a large drop off in reads more than ∼4 kb from the polyA tail. Since half of the mice used were female, we also excluded Y-chromosome-encoded genes. Furthermore, since small changes in mitochondrial number between lines might have a dramatic effect on transcript levels of mitochondrial encoded genes, we limited our analyses to nuclear-encoded genes. Thus we only include transcripts based on the following inclusion criteria: (i) maximum transcript length for a gene is <4186 bp, (ii) at least one FPKM value for any comparison is >0.3334, (iii) gene is nuclear encoded and (iv) gene is not on the Y-chromosome.

Definitions of classes

For data in Tables 2, 4, 5 and 6, Excel spreadsheets were designed to determine relationships between a transcript in different lines allowing for variation υ which can be set arbitrarily as a percentage of FPKM value. FPKM values for a transcript in two lines were considered changed when the smaller FPKM + υ did not overlap with the larger FPKM−υ. Comparisons across the series were then made leading to the following classes for ordering by phenotypic severity (WT < HDQ150/150 < HDQ315/+ < HDQ 200/200). The ‘No effect’ class exhibits no change in levels across the series having FPKM values described as 

WT±υ=150±υ=315±υ=200±υ,
where WT represents wild-type FPKM, 150 represents HDQ150/150 FPKM, 315 represents HDQ315/+ FPKM and 200 represents HDQ200/200 FPKM.

The ‘step-wise correlate’ class exhibits mRNA levels that either decrease or increase with phenotypic severity having FPKM values described as 

WT±υ>150±υ>315±υ>200±υORWT±υ<150±υ<315±υ<200±υ.

The ‘partial correlate’ class is a less stringent form of the ‘step-wise’ class allowing for some equivalent values having FPKM values that are not all equal, are not ‘step wise’ and conform to the following 

WT±υ150±υ315±υ200±υORWT±υ150±υ315±υ200±υ.

The ‘discorrelate’ class exhibits a clear breakage of correlation between mRNA level and phenotypic severity with FPKM values described as any of the following 

WT±υ<150±υand150±υ>315±υor200±υOR
 
WT±υ>150±υand150±υ<315±υor200±υOR
 
WT±υor150±υ<315±υand315±υ>200±υOR
 
WT±υor150±υ>315±υand315±υ<200±υ.

For ordering based on mutant gene dosage, WT has 0 mutant gene doses, HDQ315/+ has one and both HDQ150/150 and HDQ200/200 have 2. Thus leading to the ordering (WT < HDQ315/+ < HDQ150/150 and HDQ200/200) and definitions below. The ‘No effect’ class exhibits no change in levels across the series having FPKM values described as 

WT±υ=315±υ=200±υ=150±υ.

The ‘step-wise correlate’ class exhibits mRNA levels that either decrease or increase with mutant gene dosage while 150 and 200 are equivalent allowing for reasonable experimental error having FPKM values described as 

WT±υ>315±υ>200±υand150±υand200±6%=150±6%OR
 
WT±υ<315±υ<200±υand150±υand200±6%=150±6%.

The ‘partial correlate’ class is a less stringent form of the ‘step-wise’ class allowing for some equivalent values having FPKM values that are not all equal, are not ‘step-wise’ and conform to the following: 

WT±υ315±υ200±υand150±υOR
 
WT±υ315±υ200±υand150±υ.

The ‘discorrelate’ class exhibits a clear breakage of correlation between mRNA level and gene dosage with FPKM values described as any of the following 

WT±υ<315±υand315±υ>200±υor150±υOR
 
WT±υ>315±υand315±υ<200±υor 150±υOR
 
WT±υ<315±υandWT±υ>200±υor150±υOR
 
WT±υ>315±υandWT±υ<200±υor150±υOR
 
WT±υ<150±υand200±υ<WT±υor315±υOR
 
WT±υ>150±υand200±υ>WT±υor315±υOR
 
315±υ<150±υand200±υ<WT±υor 315±υOR
 
315±υ>150±υand200±υ>WT±υor315±υ.

Data and mouse line sharing

RNAseq data are available from NCBI's Gene Expression Omnibus accession number GSE73743. Mouse lines are available from the Jackson Laboratory, Bar Harbor, ME, USA (stock numbers 016521-016525 and 004595).

Funding

This work was supported by grants from the CHDI, Inc. (P.D., A.J.M.) and the National Institutes of Health (NS062216 to P.D., NS071168 to M.L., NS059537 to R.L.A., P30 NS47466 to T.vG., CA13148 to M.R.C. and D.K.C. and CFAR AI027767 to M.R.C and D.K.C.).

Acknowledgements

We thank Drs Rita Cowell and Gail V. W. Johnson for helpful gifts of anti-N-terminal HD antibodies and Dr Thomas Ryan for helpful discussions.

Conflict of Interest statement. None declared.

References

1
Reiner
A.
,
Dragatsis
I.
,
Dietrich
P.
(
2011
)
Genetics and neuropathology of Huntington's disease
.
Int. Rev. Neurobiol.
 ,
98
,
325
372
.
2
Huntington's Disease Collaborative Research Group
. (
1993
)
A novel gene containing a trinucleotide repeat that is expanded and unstable on Huntington's disease chromosomes
.
Cell
 ,
72
,
971
983
.
3
Cummings
D.M.
,
Alaghband
Y.
,
Hickey
M.A.
,
Joshi
P.R.
,
Hong
S.C.
,
Zhu
C.
,
Ando
T.K.
,
Andre
V.M.
,
Cepeda
C.
,
Watson
J.B.
et al
. (
2012
)
A critical window of CAG repeat-length correlates with phenotype severity in the R6/2 mouse model of Huntington's disease
.
J. Neurophysiol.
 ,
107
,
677
691
.
4
Dragatsis
I.
,
Goldowitz
D.
,
Del Mar
N.
,
Deng
Y.P.
,
Meade
C.A.
,
Liu
L.
,
Sun
Z.
,
Dietrich
P.
,
Yue
J.
,
Reiner
A.
(
2009
)
CAG repeat lengths > or =335 attenuate the phenotype in the R6/2 Huntington's disease transgenic mouse
.
Neurobiol. Dis.
 ,
33
,
315
330
.
5
Morton
A.J.
,
Glynn
D.
,
Leavens
W.
,
Zheng
Z.
,
Faull
R.L.
,
Skepper
J.N.
,
Wight
J.M.
(
2009
)
Paradoxical delay in the onset of disease caused by super-long CAG repeat expansions in R6/2 mice
.
Neurobiol. Dis.
 ,
33
,
331
341
.
6
Heng
M.Y.
,
Detloff
P.J.
,
Albin
R.L.
(
2008
)
Rodent genetic models of Huntington disease
.
Neurobiol. Dis.
 ,
32
,
1
9
.
7
Menalled
L.
,
Brunner
D.
(
2014
)
Animal models of Huntington's disease for translation to the clinic: best practices
.
Mov. Disord.
 ,
29
,
1375
1390
.
8
Ramaswamy
S.
,
McBride
J.L.
,
Kordower
J.H.
(
2007
)
Animal models of Huntington's disease
.
ILAR J.
 ,
48
,
356
373
.
9
Kuhn
A.
,
Goldstein
D.R.
,
Hodges
A.
,
Strand
A.D.
,
Sengstag
T.
,
Kooperberg
C.
,
Becanovic
K.
,
Pouladi
M.A.
,
Sathasivam
K.
,
Cha
J.H.
et al
. (
2007
)
Mutant Huntingtin's effects on striatal gene expression in mice recapitulate changes observed in human Huntington's disease brain and do not differ with mutant huntingtin length or wild-type huntingtin dosage
.
Hum. Mol. Genet.
 ,
16
,
1845
1861
.
10
Cha
J.H.
(
2007
)
Transcriptional signatures in Huntington's disease
.
Prog. Neurobiol.
 ,
83
,
228
248
.
11
Neueder
A.
,
Bates
G.P.
(
2014
)
A common gene expression signature in Huntington's disease patient brain regions
.
BMC Med. Genomics
 ,
7
,
60
.
12
Lin
C.H.
,
Tallaksen-Greene
S.
,
Chien
W.M.
,
Cearley
J.A.
,
Jackson
W.S.
,
Crouse
A.B.
,
Ren
S.
,
Li
X.J.
,
Albin
R.L.
,
Detloff
P.J.
(
2001
)
Neurological abnormalities in a knock-in mouse model of Huntington's disease
.
Hum. Mol. Genet.
 ,
10
,
137
144
.
13
Heng
M.Y.
,
Tallaksen-Greene
S.J.
,
Detloff
P.J.
,
Albin
R.L.
(
2007
)
Longitudinal evaluation of the Hdh(CAG)150 knock-in murine model of Huntington's disease
.
J. Neurosci.
 ,
27
,
8989
8998
.
14
Heng
M.Y.
,
Duong
D.K.
,
Albin
R.L.
,
Tallaksen-Greene
S.J.
,
Hunter
J.M.
,
Lesort
M.J.
,
Osmand
A.
,
Paulson
H.L.
,
Detloff
P.J.
(
2010
)
Early autophagic response in a novel knock-in model of Huntington disease
.
Hum. Mol. Genet.
 ,
19
,
3702
3720
.
15
Jin
J.
,
Peng
Q.
,
Hou
Z.
,
Jiang
M.
,
Wang
X.
,
Langseth
A.J.
,
Tao
M.
,
Barker
P.B.
,
Mori
S.
,
Bergles
D.E.
et al
. (
2015
)
Early white matter abnormalities, progressive brain pathology and motor deficits in a novel knock-in mouse model of Huntington's disease
.
Hum. Mol. Genet.
 ,
24
,
2508
2527
.
16
Mangiarini
L.
,
Sathasivam
K.
,
Seller
M.
,
Cozens
B.
,
Harper
A.
,
Hetherington
C.
,
Lawton
M.
,
Trottier
Y.
,
Lehrach
H.
,
Davies
S.W.
et al
. (
1996
)
Exon 1 of the HD gene with an expanded CAG repeat is sufficient to cause a progressive neurological phenotype in transgenic mice
.
Cell
 ,
87
,
493
506
.
17
Thomas
E.A.
,
Coppola
G.
,
Desplats
P.A.
,
Tang
B.
,
Soragni
E.
,
Burnett
R.
,
Gao
F.
,
Fitzgerald
K.M.
,
Borok
J.F.
,
Herman
D.
et al
. (
2008
)
The HDAC inhibitor 4b ameliorates the disease phenotype and transcriptional abnormalities in Huntington's disease transgenic mice
.
Proc. Natl. Acad. Sci. USA
 ,
105
,
15564
15569
.
18
Young
D.
,
Mayer
F.
,
Vidotto
N.
,
Schweizer
T.
,
Berth
R.
,
Abramowski
D.
,
Shimshek
D.R.
,
van der Putten
P.H.
,
Schmid
P.
(
2013
)
Mutant huntingtin gene-dose impacts on aggregate deposition, DARPP32 expression and neuroinflammation in HdhQ150 mice
.
PLoS One
 ,
8
,
e75108
.
19
Cha
J.H.
,
Frey
A.S.
,
Alsdorf
S.A.
,
Kerner
J.A.
,
Kosinski
C.M.
,
Mangiarini
L.
,
Penney
J.B.
Jr.
,
Davies
S.W.
,
Bates
G.P.
,
Young
A.B.
(
1999
)
Altered neurotransmitter receptor expression in transgenic mouse models of Huntington's disease
.
Philos. Trans. R. Soc. Lond. B. Biol. Sci.
 ,
354
,
981
989
.
20
Fernandez-Ruiz
J.
,
Moreno-Martet
M.
,
Rodriguez-Cueto
C.
,
Palomo-Garo
C.
,
Gomez-Canas
M.
,
Valdeolivas
S.
,
Guaza
C.
,
Romero
J.
,
Guzman
M.
,
Mechoulam
R.
et al
. (
2011
)
Prospects for cannabinoid therapies in basal ganglia disorders
.
Br. J. Pharmacol.
 ,
163
,
1365
1378
.
21
Seredenina
T.
,
Luthi-Carter
R.
(
2012
)
What have we learned from gene expression profiles in Huntington's disease?
Neurobiol. Dis.
 ,
45
,
83
98
.
22
Sadee
W.
,
Wang
D.
,
Bilsky
E.J.
(
2005
)
Basal opioid receptor activity, neutral antagonists, and therapeutic opportunities
.
Life Sci.
 ,
76
,
1427
1437
.
23
Salvemini
D.
,
Riley
D.P.
,
Cuzzocrea
S.
(
2002
)
SOD mimetics are coming of age
.
Nat. Rev. Drug Discov.
 ,
1
,
367
374
.
24
Fletcher
L.
,
Isgor
E.
,
Sprague
S.
,
Williams
L.H.
,
Alajajian
B.B.
,
Jimenez
D.F.
,
Digicaylioglu
M.
(
2013
)
Spatial distribution of insulin-like growth factor binding protein-2 following hypoxic-ischemic injury
.
BMC Neurosci.
 ,
14
,
158
.
25
Richter
F.
,
Meurers
B.H.
,
Zhu
C.
,
Medvedeva
V.P.
,
Chesselet
M.F.
(
2009
)
Neurons express hemoglobin alpha- and beta-chains in rat and human brains
.
J. Comp. Neurol.
 ,
515
,
538
547
.
26
Popper
K.R.
(
1959
)
The Logic of Scientific Discovery
 .
Basic Books
,
New York
.
27
Waterston
R.H.
,
Lindblad-Toh
K.
,
Birney
E.
,
Rogers
J.
,
Abril
J.F.
,
Agarwal
P.
,
Agarwala
R.
,
Ainscough
R.
,
Alexandersson
M.
,
An
P.
et al
. (
2002
)
Initial sequencing and comparative analysis of the mouse genome
.
Nature
 ,
420
,
520
562
.
28
Perry
G.M.
,
Tallaksen-Greene
S.
,
Kumar
A.
,
Heng
M.Y.
,
Kneynsberg
A.
,
van Groen
T.
,
Detloff
P.J.
,
Albin
R.L.
,
Lesort
M.
(
2010
)
Mitochondrial calcium uptake capacity as a therapeutic target in the R6/2 mouse model of Huntington's disease
.
Hum. Mol .Genet.
 ,
19
,
3354
3371
.
29
Dixon
K.T.
,
Cearley
J.A.
,
Hunter
J.M.
,
Detloff
P.J.
(
2004
)
Mouse Huntington's disease homolog mRNA levels: variation and allele effects
.
Gene Expr.
 ,
11
,
221
231
.
30
Kumar
A.
,
Kneynsberg
A.
,
Tucholski
J.
,
Perry
G.
,
van Groen
T.
,
Detloff
P.J.
,
Lesort
M.
(
2012
)
Tissue transglutaminase overexpression does not modify the disease phenotype of the R6/2 mouse model of Huntington's disease
.
Exp. Neurol.
 ,
237
,
78
89
.