Improved Methodology for Assessment of mRNA Levels in Blood of Patients with FMR1 Related Disorders

The mutations in an X linked FMR1 gene [1] have some unique qualities. While large CGG expansions (>200) 3' of its promoter give rise to fragile X syndrome – a neurodevelopmental disorder caused by silencing of the gene, smaller expansions (permutations – PM, 55–200 repeats) lead to late onset pathologies through entirely different mechanisms, involving a toxic gain of function of this gene's transcripts [2‐4]. Fragile X Tremour Ataxia Syndrome (FXTAS) is the most prevalent disorder associated with PM alleles occurring in 50% of all older carrier males [5]. It is progressive neurodegenerative disorder manifesting with ataxia and/or intension or other tremors, cognitive decline, psychiatric involvement and characteristic MRI and histopathological features [5‐8]. In females, these alleles are associated with a significant increase in premature ovarian failure [9]. A toxicity of an elevated FMR1 mRNA which is believed to be pathogenic in these conditions, has been supported in animal models where it leads to increased cell death in Drosophila [10] and mice [11]. Although, in humans the level of mRNA is usually assessed in whole blood, there is a significant correlation between FXTAS associated with these levels and the typical neurological [12], and psychiatric [13] manifestations. This implies that the assessments of FMR1 mRNA levels in carriers of small CGG expansions, apart from being essential in the mechanisms involved in RNA toxicity, may also have a diagnostic and prognostic significance.

FMR1 expression is primarily regulated by the state of methylation of its promoter. DNMT1 is an enzyme pivotal in this process [14‐16]. This implies that DNMT1 may be indirectly involved in the mechanisms leading to FMR1 related disorders. Thus, expression of both genes, FMR1 and DNMT1, is of interest in fresh and archival samples from individuals with diseases characterized by abnormal expansions within the CGG region. Unfortunately, archival samples and samples extracted from blood have variable total RNA and mRNA degradation, which we have found to be a confounding factor in the gene expression analysis, unless a method was established to normalize for FMR1 and DNMT1 mRNA degradation.

Previous real-time PCR studies normalized gene expression against total cell number, total RNA or mRNA of internal control gene/s, reviewed in [17]. The main constraints associated with using total cell number for normalization is that the RNA has to be of a very high quality and the cell counts have to be accurate, which is not possible for solid tissues. Total RNA has been particularly useful for normalization using absolute and relative standard curve methods for real-time PCR, and has been shown to produce biologically relevant and highly reproducible data [18]. However, the most frequently applied normalization method involves the use of internal control gene/s based on the rational that the mRNA levels of these control genes in the tested samples should reflect the effect of variables on the quantity and quality of the target gene mRNA, and at the same time be minimally effected by the tested variables [19, 20]. Previous real-time PCR expression assays for FMR1 and DNMT1 have utilized GUS and HPRT as internal control genes [2, 21, 22]. However, these assays did not take into consideration the appropriateness of these genes as controls in relation to the quantity and quality of the starting material. A number of studies have suggested that confirmation of total RNA quality prior to real-time PCR quantification via capillary electrophoresis is essential in obtaining meaningful gene expression data [23].

In this study the Experion system (Bio-rad) was used to assess the quality of rRNA by subjective/visual interpretation of the chromatographs for absence of peaks other than those of 18S, 28S and 5S, and/or by examining the 28S:18S ratio and the RNA Qiality Indicator (RQI) values. However, these approaches have their limitations which we addressed in the study. We also evaluated in depth the use of internal control genes and the capillary electrophoresis features for the normalization of mRNA degradation in partially degraded and intact RNA from patient lymphoblasts and blood. We then defined the best predictor of target gene mRNA degradation, utilizing it for normalization in the whole blood RNA with a high variability of rRNA quality. In this sample set we examined the relationship between the FMR1 mRNA and the CGG expansion in a non-coding portion of this gene, and found that by incorporating the 3' and 5' mRNA analysis data from patients with small to intermediate CGG expansions, we have significantly improved upon the current approach used to examine the FMR1 mRNA toxicity, demonstrating the clinical relevance of our strategy.

The specific aim of this study was to establish an optimal method to normalize for the degradation of target gene FMR1 mRNA in a sample set showing large variability in rRNA quality, and demonstrate it's clinical/biological relevance. Previous studies have normalized FMR1 expression to GUS [2, 21]. However, we questioned the appropriateness of GUS as an internal control for our sample set, where we have observed a large variation in rRNA degradation, particularly as there are no previous studies that examined the rates of mRNA degradation for both FMR1 and GUS. Since different mRNA species degrade depending on their length and secondary structure [18, 23, 29], we assessed if GUS was a suitable control for FMR1 mRNA degradation, and if not, which of the capillary electrophoresis and real-time PCR parameters would provide a better normalization method. Because of the potential relevance of DNMT1 to FMR1 gene regulation [30‐33], we have included a parallel analysis of DNMT1 as as target gene in some of the studies.

We initially demonstrated using the Experion system that for the artificially degraded RNA samples, both the 28S:18S ratio and the RQI were most useful as predictors of severe RNA degradation, whereas the greatest changes in stability of different transcripts examined occurred during early to moderate stages of RNA degradation. Thus, the 28S:18S ratio and the RQI were not suitable predictors of mRNA stability, at least in our settings. Since, the RQI is closely related to a more widely used RNA Integrity Number (RIN) from an analogous Agilent system (Bio-Rad electrophoresis technical note 5761), our findings suggest that RIN may be also inappropriate as a normalization tool in our settings. Furthermore, the subjective assessment of general chromatographic features did not provide a useful estimation of mRNA degradation, as real-time PCR could still be used to obtain biologically relevant mRNA data in samples with chromatographs indicating severe rRNA degradation. In another approach, we determined that of the 10 selected chromatographic features, 18S, 28S and the inter-peak region % areas were the most reliable predictors of total RNA degradation when examined as a function of the degradation time. However, the normalization of the target genes to 18S and 28S chromatographic features was found to be inferior to the use of the internal control genes.

Primarily these observations indicate that the degradation kinetics of rRNA may be heavily size dependent, as the small 5S rRNA subunit, 160 nucleotides, was found to be a good predictor of only early to moderate RNA degradation, whereas the 18S and 28S % areas, 1770 and 3770 nucleotides respectively, could be used to predict RNA degradation at early, moderate and late stages of degradation. For mRNA quantitation, by qPCR, however, it has been previously suggested that the length of the amplicon, rather than that of the whole mRNA molecule, may be a more important indicator of degradation kinetics, particularly because fragmentation of a long mRNA may only result in a loss of the molecule for qPCR detection if the RNA break occurs within the product sequence [23].

We've found that the location of the amplicon, may be just as important as the size in determining the effect of mRNA degradation on qPCR performance. This was initially observed by examining the differences in FMR1ex3.4 (5') and FMR1ex13.14 (3') qPCR data throughout the RNA degradation time course. Since most FMR1 transcripts contain the 195-bp exon 14 [34], and all contain exons3/4, both assays target FMR1 transcripts of a similar size and abundance, which is experimentally reflected by the highly significant correlation between the two assays for high quality total RNA samples from the peripheral blood of patients with small to intermediate size expansions (Additional file 1). Since the amplicons for both assays are also of a similar size, if the location of the amplicon is unimportant in determining RNA degradation kinetics, its effects should be similar on the qPCR results for both assays. However there were clear differences between the 5' and 3' FMR1 results for the sample set with the high variability in total RNA quality, which may be explained by the 3' alternative splice site being more susceptible to RNA breaks than the 5' common region. Thus, thermal stress at the 3' site may result in the loss of more molecules for qPCR detection, than that at the 5' site, which is consistent with our findings in the RNA samples extracted from whole blood where RNA degradation had far less impact on the biological significance of the FMR1ex3.4 data than on the FMR1ex13.14 results (Figure 4B).

We have also found that due to the differences between degradation kinetics of mRNA and rRNA, normalization of target mRNA levels using rRNA profiles, as indicated by capillary electrophoresis, was inferior to the use of the internal control genes. Therefore, we proposed an alternative approach, independent of capillary electrophoresis, where by targeting each transcript at multiple sites for qPCR analysis, the confounding impact of mRNA breaks at any specific location on qPCR performance can be minimized. Specifically, in a previous study we've found that the FMR1 mRNA levels were proportional to the size of CGG expansions within the grey zone and lower PM range [4]. In 30 of the original samples we've found the rRNA quality to vary significantly. Half of the samples would have had to be omitted from further analysis unless an alternative method to capillary electrophoresis was found to normalize for mRNA degradation.

We examined FMR1 expression in these samples without omission by targeting FMR1 transcript at 5' and 3' ends standardized to GUS, as we have shown that this strategy is superior to the use of rRNA profiles from the capillary electrophoresis analyses of the artificially degraded samples. Of the 30 samples we found four outliers that had a significant impact on qPCR performance particularly for the 3' FMR1ex13.14 assay. In 3 out of the 4 outliers this was not directly related to rRNA quality. By combining FMR1 mRNA 3' and 5' end analyses, the confounding impact of these outliers on the correlation between FMR1 expression and CGG size was minimized, and the statistical significance of the data doubled.

In summary, we have demonstrated that in artificially degraded RNA samples a number of chromatographic features including 18S, 28S, the inter-peak region, the 28S:18S ratio and RQI can be used as predictors of different stages of total RNA degradation. However, their use for normalization of target gene mRNA degradation was inferior to the use of internal control genes, of which GUS was the most appropriate as it closely reflected the target gene mRNA degradation kinetics. For the target FMR1 mRNA we've shown that the location of the amplicon, may be just as important as the size of its transcript in determining the effect of mRNA degradation on qPCR performance. Furthermore, we've found that by targeting the FMR1 transcript at multiple sites, the confounding impact of mRNA breaks due to fragmentation within any specific qPCR product sequence, and the subsequent loss of mRNA molecules for qPCR detection, was minimized. In clinical practice, this approach may be extremely useful, as there is extensive variation in the time lapse between the blood collection and RNA extraction for different samples, which contributes variation in rRNA and mRNA quality. Our strategy allows for a more accurate comparison between samples without sample loss.

Whilst we have shown that this approach is applicable for the FMR1 mRNA in RNA extracted from whole blood in our patient subset, there are also potential implications on the development of diagnostic tests for the levels of FMR1 mRNA toxicity associated with a number of clinical conditions [2‐9, 12, 13]. These findings are also likely to have a broader application to expression studies of other genes using precious archival or fresh blood samples with a large variation in rRNA quality where sample omission is not an option. More importantly, the approach presented in this manuscript may be useful to any diagnostic application where mRNA integrity may be compromised [35], including future developments of methods for detection of placental mRNA in maternal plasma, foetal RNA markers for non-invasive prenatal diagnosis of pregnancy associated diseases and foetal chromosomal aneuploidies, as well as clinical uses of plasma or whole blood RNA for detection of tumour RNA in cancer diagnostics.