Fourteen EBV transformed lymphoblast cell lines from healthy controls (FMR1 alleles – 6 to 40 repeats); grey zone carriers (41 to 55 repeats), premutation carriers (55 to 200 repeats) and full-mutation carriers (>200 repeats) 242526 were obtained from the MCRI tissue culture storage repository, or purchased from Coriell (Table 1). The cell-viability counts and cell composition were determined by trypan blue exclusion, using a hemocytometer. Cells were plated at a density of 10⁶ cell/ml in RPMI medium (Sigma StLuis, MO, USA)/10% Foetal bovine serum (CSL) at 37°C, 5% CO₂ overnight. Cell pellets were resuspended in RNA lysis buffer, (10 μl β-mercaptoethanol/1 ml of RLT buffer from Qiagen, Hilden, Germany), and stored at -80°C until RNA extraction.

Total RNA was extracted and purified using the RNeasy extraction kit (Qiagen Inc., Hilden, Germany). RNA concentrations were measured in triplicate using a NanoDrop ND-1000 Spectrophotometer, with purity being determined by the A260/A280 ratio using the expected values between 1.8 and 2. Each RNA sample was then diluted to 2 ng/ul. Sixty ul of each sample was left at room temperature for 0, 18, 24 and 96 hours. At each time point an aliquot of 15 ul was collected from each sample, 5 ul of which were used for capillary electrophoresis (Bio-rad) and 2 ul for cDNA synthesis.

The original study that investigated the relationship between FMR1 expression and CGG expansion was approved by the La Trobe University Human Ethics Committee, the Southern Tasmanian Health and Medical Human Research Ethics Committees, and by the Institutional Review Board of the University of California at Davis. The methods used for CGG repeat sizing and FMR1 mRNA assessments, as well as the results of correlation between these two measures in a larger sample, which included the present participants, were reported in the earlier study 4. An aliquot of total RNA was originally isolated from 3 ml of peripheral blood using Tempus Blood RNA tubes as previously described in: Loesch, et al. (2007) 4. After more than 1 year of storage at -80°C, the total RNA quality was assessed using capillary electrophoresis. All samples were diluted to 5 ng/ul, prior to reverse transcription. Here we utilized these earlier data on CGG repeat size, but validated and conducted new assays on the expression of FMR1ex3.4, FMR1ex13.14 and GUS.

Total RNA quality was assessed using the RNA HighSens Kit as per manufacturer's instructions (Bio-rad). Objective and subjective analysis of each RNA profile was performed. Objective assessment involved automatic collection of a number of (manually set) parameters which included percentage area under the lower marker (LM), small fragment region (SR), 5S, fast region (FR), 18S, Inter region (IR), 28S, post region (PR), the 28S:18S ratio and RQI (Figure 1A). Subjective assessment (Figure 1B and 1C) involved descriptive comparison of chromatographic features based on previous publications using this system 27.

Reverse transcription was performed one reaction per sample using the Multiscribe Reverse Transcription System, 50 units/ul (Applied Biosystems). The 7900HT Fast Real Time PCR (Applied Biosystems) was used to quantify FMR1ex3.4, FMR1ex13.14, DNMT1, GAPDH, B2M, and GUS, using the relative standard curve method. The target gene and the internal control gene dynamic linear ranges were performed on a series of doubling dilutions of RNA (160-0.5 ng/ul) of a selected peripheral bood mononuclear cell sample. Previously published sequences were used for RT-PCR primers and TaqMan probe for: FMR1exon3/4 and GUS 2; FMR1exon13/exon14 21, and DNMT1 22. FMR1exon3/4, FMR1exon13/14 and DNMT1 primers and probes were used at concentrations of 18 uM and 2 uM, respectively. GAPDH and B2M primer/probe mixes were obtained from PrimerDesign (PerfectProbe ge-PP-12-hu kit) and used at concentration of 2 uM. Each sample was assayed in duplicate 10 μl PCR reactions, consisting of 5.8 mM MgCl₂, 1 μl Buffer A (Applied Biosystems), 3.35 μl RNase-free water, 1.2 mM dNTPs, 0.01 units/μl of AmpliTaq Gold, 0.5 μl of TaqMan probe and 0.5 μl forward and 0.5 μl reverse primers, and 1 μl of the reverse transcription (cDNA) reaction. The annealing temperature for thermal cycling protocol was 60°C for 40 cycles. Samples were quantified in arbitrary units (au) in relation to the standard curves performed on each plate.

Generalized estimating equations (GEE) were used for the objective assessment of the chromatographs and mRNA integrity. Normalization of FMR1 and DNMT1 mRNA to 18S, 28S, GAPDH and GUS, expressed as a function of the degradation time, was assessed by fitting the values to a simple linear regression. The sign test was then used to examine the null hypothesis that the median slope was zero against the two-sided alternative of it not being equal to 0. The relationship between the qPCR results for FMR1ex3.4 and FMR1ex13.14 assays with the CGG expansion size was assessed using a significance test in a linear regression. The analysis was conducted using the publicly available R statistical computing package, version 1.191 (R development Core team, 2004).

FMR1 and DNMT1 mRNA quantities were examined using the relative standard curve method, in RNA samples 496, 497, 488, 584 and 585 artificially degraded at room temperature for 0, 18, 24 and 96 hours from 5 lymphoblast cell lines (Table 1). The samples were subjectively assessed as described in Figure 1 for the relationship between the capillary electrophoresis profiles and mRNA quantities of FMR1 (5' and 3') and DNMT1 (Figure 2).

Sample 496 (Figure 2; Panel I) displayed minor differences in the capillary electrophoresis profiles between 0 and 24 hours, with the most prominent change being observed between 24 and 96 hours (a large increase in the inter-peak area and the fast region, and a marked decrease in the 28S peak). This was also reflected in the 28S:18S ratios and RQI values, where there was no change between 0 and 24 hours, but a large decrease between 24 and 96 hours. In contrast, FMR1ex13.14 mRNA level decreased by more than half within the first 18 hours and remained at this level for the following time points. A similar trend was observed for FMR1ex3.4 mRNA, which decreased between 18 and 24 hours of incubation, and remained at the same level through to 96 hours. In contrast, there were no major changes in DNMT1 mRNA quantities for sample 496 throughout the time course.

Sample 497 was different to all the other RNA examined, as it displayed a second, smaller 28S peak at 0 hours (Figure 2; Pannel II E). The origin of this peak, which has been previously reported using the analogous system (Schroeder et al.; United states patent publication 2006/0246577 A1) is unknown; the absence of any other RNA degradation markers such as an increase in the inter-peak area and/or the fast region argue against a plausible explanation that it could represent 28S rRNA degradation. It is also unclear why a significant change in the capillary elecrophoresis profile for this sample was observed at 18 hours, when the larger 28S peak(s) almost completely disappeared (Figure 1, Panel II F), while at 24 hours there was an increase in the 28S:18S ratio and RQI (Figure 1, Panel II F and G). This may be linked to the structure of the anomolous 28S peak, the unusual integrity of which is beyond the scope of this study. Interestingly, the mRNA levels for FMR1 and DNMT1 did not change between 0 and 18 hours. The greatest increase in total RNA degradation of sample 497 was observed between 24 and 96 hours, as indicated by a significant increase in the inter-peak area and the fast region and a marked decrease in the 28S:18S ratio and the RQI value (Figure 2, Panel II G and H). However, again this trend was not mirrored by the degradation of FMR1 and DNMT1 mRNA, which showed no major differences between 24 and 96 hours.

For the sample 488, within the first 18 hours there was no difference in RQI, however the 28S:18S ratio decreased from 2.5 to 1.57, with a slight increase in the inter-peak area within the first 18 hours (Figure 2, Panel III, I and J). At 24 and 96 hours we observed a further decrease in the 28S:18S ratio which was mirrored by a moderate decrease in RQI, and a prominent increase in the inter peak area and the fast region (Figure 2, Panel III, K and L). In contrast, there was only a slight decrease in DNMT1 mRNA level within the first 18 hours, which remained at the same level for the following time points.

Samples 584 and 585 demonstrated similar kinetics of total RNA degradation. At 0 hours these samples had similar RQI values, but different 28S:18S ratios (Figure 2, Panels IV and V). A striking increase in total RNA degradation was observed between 24 and 96 hours. During this period for sample 584 the 28S:18S ratio dropped from 1.2 to 0, and RQI from 9.3 to 2.9 (Figure 2, Panels IV; O and P), while for sample 585 the 28S:18S ratio dropped from 1.64 to 0.77, and RQI from 9.7 to 7.7 (Figure 2, Panel V; S and T). This was related to a significant increase in the inter peak area and the fast region. In contrast, mRNA quantities of FMR1ex3.4 and FMR1ex13.14 were decreased by approximately half between 0 and 18 hours in both samples 584 and 585, and for DNMT1 in sample 584 about 3 fold between 0 and 24 hours.

The similarities in FMR1 mRNA and total RNA degradation kinetics in both 584 and 585 samples may be related to these cell lines harbouring grey zone alleles, whereas samples 496 and 497 that showed different degradation kinetics were from premutation carriers (Table 1). Although, an in-depth investigation of this relationship is beyond the scope of this manuscript, the differences in FMR1ex3.4 mRNA degradation kinetics between the premutation and grey zone cell lines (Figure 2) were consistent with previous studies showing that the increased length of the CGG tract correlates with increased mRNA stability through hairpin formation within the 5'UTR repeat region 28.

Together these data suggested that the total RNA degradation rate moderately varies between the samples from different cell lines with the most prominent changes being observed between 24 and 96 hours. This was poorly correlated with the profile of FMR1 and DNMT1 mRNA degradation that predominantly occurred within the first 18 to 24 hours, indicating that there was no clear correlation between the rate of total RNA degradation from the subjective assessment of the chromatographs and mRNA degradation as determined by real-time PCR. Because FMR1 mRNA stability may be related to the size of the CGG repeat within its UTR and the pathology of FMR1 related disorders 28, in the following sections we have established a method to normalize for mRNA degradation independent of the CGG expansion size, so that the clinical relevance of the CGG related FMR1 mRNA toxicity can be identified in samples with variable rRNA quality.

Objective assessment of total RNA and mRNA degradation was performed in RNA samples from 14 cell lines (Table 1). Ten features were measured from each chromatograph and 6 variables measured using real-time PCR from the corresponding cDNA samples at 4 paired time points. The RNA from three fragile X cell lines were excluded from the FMR1 real-time PCR analysis for the RNA degradation study, as they had no FMR1 expression. The first aim of this approach was to objectively delineate which features of the chromatographs and gene expression profiles (GAPDH, B2M, GUS) could be used as predictors of the total RNA degradation as reflected by the degradation time. The second aim was to objectively delineate whether FMR1 and DNMT1 mRNA degradation correlated with the degradation time, and if not, which features could be used as predictors of the target gene mRNA integrity.

Based on the subjective assessment, we arbitrarily divided the degree of total RNA degradation into four categories: 0 hours – intact; 18 hours – early degradation; 24 hours – moderate to severe degradation; 96 hours – severe degradation, under the assumption that RNAs from different cell lines follow this progressive trend of degradation at the four time points. GEE were then utilised to provide an estimation of which parameters most closely reflected FMR1 and DNMT1 mRNA integrity in early versus moderate versus late degradation stages, and through the time course as a whole.

As expected, the most significant predictors of total RNA degradation from the combined and individual comparisons of chromatographic features were the percentage areas of 18S, 28S and the inter-peak region (Table 2). These features appeared to be suitable predictors of early, moderate and severe RNA degradation (p < 0.001). In contrast, 5S % area was a good predictor of only early to moderate RNA degradation (p < 0.05), and the small fragmentation region percentage area of moderate to severe RNA degradation (p < 0.05). However, the most prominent predictors of severe RNA degradation were the fast region % area (p < 0.001), the 28S:18S ratio (p < 0.001), the RQI (p < 0.001) and the lower marker % area (p = 0.072). Interestingly, in contrast to the 28S:18S ratio and the lower marker % area, the RQI could be also used as a predictor of moderate degradation (p < 0.05). The post region % area was the only parameter that was not associated with any stage of RNA degradation.

GUS and GAPDH mRNA degradation closely reflected moderate to severe degradation of total RNA (p < 0.001), best represented by the small fragmentation region % area. In contrast, B2M mRNA could only be used to predict early total RNA degradation (p = 0.093). Furthermore, it was not associated with any other chromatographic and gene expression parameters examined. For the target transcripts, DNMT1 and FMR1ex3/4, mRNA degradation was closely associated with moderate to severe degradation of total RNA (p < 0.001), and was best predicted by the small fragmentation region % area, and by GUS and GAPDH mRNA, all of which had similar degradation kinetics. In contrast, FMR1ex13.14 mRNA degradation appeared to be closely associated with early, moderate and severe total RNA degradation (p < 0.001), and was best predicted by % areas of 18S, 28S and the inter-peak region (p < 0.001). This analysis demonstrated that the degradation of mRNA was different between most internal controls and target genes examined, and that the degradation kinetics of specific mRNAs were not necessarily the same as those for total RNA and rRNA.

The target gene expression in the RNA samples from the cell lines were normalized to the best predictors of FMR1 and DNMT1 mRNA degradation as determined by the objective assessment of chromatographs and mRNA integrity. 18S % area, 28S % area, GUS mRNA or GAPDH mRNA quantity were used for normalization, and expressed as a function of the degradation time. The normalization method that provided the most constant (least significant) values throughout the time course was considered as the most optimal of the predictors tested for the target gene mRNA degradation (Figure 3; the sign test was used to examine the null hypothesis that the median slope was zero against the two-sided alternative of it not being equal to 0). We found that for DNMT1 both 18S (p = 0.066) and 28S (p = 0.066) % area were poor normalization features compared to GAPDH (p = 0.388) and GUS (p = 0.774). Similarly, for FMR1ex3.4, both 18S (p = 0.11) and 28S (p = 0.11) % area normalization provided less constant values than GAPDH (p = 0.51) and GUS (p = 0.254). For FMR1ex13.14, 18S (p = 0.11) and 28S (p = 0.11) % area as well as GAPDH (p = 0.11) normalization provided less constant values than GUS (p = 0.34). Thus, it appeared that normalization of the target genes to 18S and 28S chromatographic features was overall inferior to the use of internal control genes, GAPDH and GUS. Although 18S and 28S % areas could be still be used as normalization markers for FMR1ex3.4 and FMR1ex13.14, the most optimal of the predictors tested for both DNMT1 and FMR1 was GUS.

We have previously demonstrated using freshly extracted RNA, that FMR1 expression was significantly elevated in carriers of CGG expansion, compared with normal controls of a similar age, and that the expression was proportional to the size of CGG expansions within the grey zone and lower premutation range 4. Subsequent analysis of these stored samples revealed a high variability in rRNA quality, which posed major confounder concerns. We have therefore determined whether these samples could provide clinically relevant data using the normalization criteria tested in the study.

The FMR1ex3.4/GUS mRNA levels assessed here closely corresponded to the levels of the earlier study in freshly extracted RNA samples4, but using the relative standard curve as opposed to the delta-Ct method. As expected we have also found a significant correlation (p = 0.028) between the FMR1ex3.4/GUS and FMR1ex13.14/GUS assays indicating that most samples had intact FMR1 mRNA (Figure 4A) despite the observed variability in rRNA quality as exemplified by the chromatographs of samples 350 and 351, with the 28S:18S ratio between 2.1 and 0.4, and RQI between 9.1 and 3.8 (Figure 4C and 4D). We have also found a significant linear correlation between FMR1ex3.4/GUS mRNA levels and the CGG expansion size, (p = 0.046) (Figure 4B) indicating the biological relevance of the data. However, there was no significant correlation between FMR1ex13.14/GUS mRNA levels and the CGG expansion size (p = 0.1) (Figure 4B).

A number of samples (samples 339, 354, 358 and 360 – colour coded), did not follow the common pattern of correlation between the FMR1ex3.4/GUS and FMR1ex13.14/GUS assays (Figure 4A), and the FMR1ex3.4/GUS and FMR1ex13.14/GUS assays with the CGG expansion size (Figure 4B). Two of these samples, 339 and 360 were also outliers for rRNA quality assessment with the 28S:18S ratios of 3.3 and 0.24, and RQI values of 10 and 3, respectively (Figure 4D). Furthermore, there was no uniform correlation between rRNA integrity and FMR1 mRNA quality for both the FMR1ex3.4/GUS and FMR1ex13.14/GUS assays in these samples. Each of these samples appeared to be significantly affected at either the 5' or 3' sites (Figure 4A, B and 4C). Only one sample (360), with poor rRNA profile, appeared to have FMR1 mRNA integrity compromised at both the 5' and 3' sites (Figure 4B and 4D). The confounding impact of these outliers was minimized when the data for the FMR1 mRNA 3' and 5' end analyses were combined, as the significance of the FMR1 correlation with CGG expansion size increased to p = 0.018.