Integrity of RNA in long-term-stored cervical liquid-based cytology samples: implications for biomarker research

Abstract

Biobanks of cervical screening (LBC) samples annotated with disease status are an invaluable resource to support the development of tools for the risk stratification of disease. Although there is growing interest in the assessment of RNA-based biomarkers, little is known on the suitability and durability of stored clinical samples (commonly used in cervical screening) to support RNA-based research. RNA was extracted from 260 stored LBC samples. Storage at -80°C or -25°C allowed isolation of sufficient RNA for further analysis. RNA was found to be substantially degraded according to Agilent Bioanalyser data. Despite this, RT-qPCR was successful in 95% samples tested. These data suggest that biobanked LBC samples are suitable for RNA-based assessment even if stored for up to 14 years.

Multidisciplinary abstract

It is important to develop novel biomarkers that allow clinicians to accurately determine cervical disease stage. RNA biomarkers are easily and inexpensively assessed in hospital laboratories by reverse transcriptase quantitative PCR (RT-qPCR). This study examined the quality and quantity of RNA extracted from cervical screening samples stored long term at two different temperatures, -80 and -25°C. RNA was obtained from samples stored at either temperature. Although the RNAs were generally unsuitable for deep (next generation) sequencing approaches, they were suitable for RT-qPCR analysis. The results indicate that biobanked cervical screening samples can be used for RNA biomarker analyses.

Method summary

RNA was extracted from 260 cervical screening samples stored at either -80 or -25°C. An Agilent Bioanalyser was used to examine the level of degradation of a convenience sample of RNAs. Reverse transcriptase quantitative PCR (RT-qPCR) was used to quantify levels of two cellular mRNAs in all samples as a practical means of assessing suitability of the samples for mRNA biomarker analysis.

Executive summary

Background

• Low level knowledge of the suitability and durability of stored clinical samples, such as those used in cervical screening, for use in RNA-based research.

Experimental

• RNA extraction from cervical screening liquid based cytology (LBC) samples.

• RNA concentration and integrity analysis using NanoDrop and an Agilent Bioanalyser.

• RT-qPCR validation of detection and quantification of two cellular RNAs.

Results & discussion

• RNA can be prepared from >95% of stored LBC samples but those representing high grade disease (CIN2+) yielded statistically less RNA.

• Storage temperature -25°C allowed isolation of greater RNA quantities compared with -80°C but sufficient RNA was isolated from most samples stored at either temperature for subsequent molecular analysis.

• LBC RNAs were found to be degraded and unsuitable for next generation RNA sequencing.

• Over 95% of LBC RNA were suitable for RT-qPCR analysis.

• A test analysis using RT-qPCR demonstrated detection of β-actin and p16 RNAs.

Conclusion

• RNA prepared from long term stored LBC samples is suitable for RT-qPCR analysis of cellular RNA biomarkers of disease status.

Keywords: :

• biobanking
• cervical screening
• reverse transcriptase quantitative PCR
• RNA biomarkers
• RNA quality
• storage temperature

Biobanks are an extremely important resource for the HPV basic bioscience and clinical research communities. They offer the opportunity to test development of novel cervical screening technologies and to provide clinical samples on which to test basic research hypotheses. Biobanking is accepted as a key resource for progressing HPV research [1]. Well established biobanks such as the Kaiser Permanente collection (NIH) [2], the Swedish Cervical Cytology Biobank [3] and the Scottish HPV Archive [4] provide such a resource. Additionally, biobanks can ‘fast track’ evaluation of new technologies through enrichment for samples/conditions that would ordinarily take several years to collate prospectively.

This said, there is relatively little detailed information on the impact of long-term storage on the stability of nucleic acid within routinely taken cervical samples, including liquid based cytology (LBC) samples. This is particularly the case for RNA, an inherently unstable molecule, which nevertheless may be key to biomarker research [^5-8]. Certainly, the requirement for independent optimal biomarkers of transforming HPV infection is ever more pressing given the move to HPV-based screening and validation of candidates using banked samples, where longitudinal outcomes are known, will expedite their development.

In this study we aimed to determine the integrity and yield of RNA in a set of 260 cervical LBC samples representing the disease spectrum from the Scottish HPV Archive. We assessed different measures of quality and durability though molecular approaches. Implications of the results for future research on RNA using stored samples are discussed.

Materials & methods

Overview Scottish HPV archive

The Scottish HPV Archive is a biobank for HPV-associated research which comes under the auspice of NHS Lothian Bioresource-REC 20/ES/0061. It includes residual cervical samples from women attending routine cervical screening and colposcopy clinics and samples from HPV associated research consented for storage. Application to use the samples for this work was obtained through application to the Scottish HPV Archive and adjudication by the archive steering committee (application reference SA-066).

Sample set used for analysis

A total of 260 LBC samples were assessed. Storage times ranged from 3 to 11 years. 73 samples were stored at -25°C and 187 samples at -80°C. All samples were HPV-positive according to the RealTime High-Risk HPV test (Abbott Molecular, USA), or the Optiplex HPV Genotyping Assay (Diamex, Heidelberg, Germany). With respect to underlying disease status, 60 samples were negative for dysplasia; 69 were associated with cervical intraepithelial neoplasia (CIN); 27 were associated CIN2; and 104 were associated with CIN3.

RNA extraction

A 2.5 ml of original liquid-based cytology (LBC) samples were centrifuged at 350 × g for 5 min at 4°C. The pellet was then resuspended in 400 μl RNase-free PBS on ice. Using the RNAdvance Blood Kit (Beckman Coulter) RNA extraction was performed on the KingFisher Flex Purification System (Thermo Fisher Scientific, MA, USA) following the manufacturer's guidelines.

RNA quantification & cDNA synthesis

RNA samples were assessed by the NanoDrop (Thermo Fisher Scientific) to give an indication of quality and quantity. cDNA was synthesized using the Maxima First Strand cDNA synthesis kit (Thermo Fisher Scientific), with additional dsDNase, following the manufacturer's instructions.

RNA quality analysis

RNA quality was assessed by examining the RNA concentration and profiles and ratios of ribosomal RNA bands on an Agilent 2100 Bioanalyser. RNA integrity was reported in RIN values using the RIN software algorithm [9]. The DV200 tool (Agilent) was used to evaluate the percentage of fragments in an RNA population of >200 nucleotides and is used as a quality assessment standard [10].

Biomarker detection by RT-PCR

RT-PCR amplification was performed on all samples in triplicate on an ABI 7500 Fast RT-PCR system (Applied Biosystems) using 96-well plates. Amplification and quantification of the gene of interest (p16) and the β-actin housekeeping gene were performed in the same RT-PCR reaction, using probes labelled with FAM and HEX, respectively. The probes and primers used to amplify β-actin were: forward primer: 5′-AGCGCGGCTACAGCTTCA-3′. Reverse primer: 5′-CGTAGCACAGCTTCTCCTTAATGTC-3′. Probe (HEX): 5′-ATTTCCCGCTCGGCCGTGGT-3′. The probes and primers used to amplify p16 were: forward primer: 5′-GCACATTCATGTGGGCATTT-3′. Reverse primer: 5′-GACTCAAGAGAAGCCAGTAACC-3′. Probe (FAM): 5′-CCGGAAGCTGTCGACTTCATGACA-3′. The 20 μl reaction volume contained 1x Takyon ROX Probe MasterMix dTTP blue (Eurogentec, Seraing, Belgium) 0.9 μmol/l of each primer, 0.1 μmol/l of probe (Eurogentec) and 5 μl of sample cDNA. The thermal profile consisted of one cycle at 50°C for 2 min then 95°C for 3 min, followed by 40 cycles of 95°C for 10 s, 60°C for 1 min. A no template control and a positive control of cDNA synthesized from RNA prepared from W12E cells representing an HPV16 infection of cervical epithelial cells [11] were included on every plate. Samples were analyzed using β-actin as the internal reference control and relative quantification of the gene of interest was calculated using the Livak 2^-ΔΔCq method. Variation between experiments was controlled for by normalizing against the values for W12E cDNA to give final RNA relative expression values (REVs).

Statistical analysis

Statistical analysis was carried out using Graphpad Prism using a Student's t-test or a Mann-Whitney-U test.

Results & discussion

Clinical samples

All samples were taken routinely (cervical liquid based cytology (LBC) samples) as part of the cervical screening program in Scotland. Sample capture and fixation was consistent for all samples. Samples were not subjected to previous freeze-thaw prior to analysis for this study. There was a wide range of storage duration for the samples, with a minimum storage time of 3 years and maximum of 11 years. We selected 129 LBC samples with a histological grading ≤CIN1 and 131 samples with a histological grading of CIN2+ (260 LBC samples in total). Sample selection was performed initially to enrich for CIN2+ end points, but to have a component of controls (CIN1) to support a case-control study. The only inclusion criterion was therefore LBC samples that had information to allow classification into CIN2+ or ≤CIN1. Table 1 shows a summary of the sample cohort with histological grade of disease and HR-HPV positivity. 43% of LBCs were positive for HPV16 or HPV18.

Table 1. Number of samples representing each grade of cervical disease. All samples were confirmed HPV-positive, some samples were positive for more than one HPV type.

Disease grade	Total number of samples	HPV16	HPV18	Other HR-HPV
Normal	60	16	4	49
CIN1	69	22	2	66
CIN2	27	13	5	15
CIN3	104	44	6	73

RNA quantification according to underlying disease

RNA concentrations ranged between 0 and 3081 ng/μl (Figure 1A) with an average concentration of 76 ng/μl. 13/260 (4.6%) samples had a concentration of less than 1 ng/μl. Samples with a histological grade of CIN1 or less gave RNA concentrations ranging from 0.4 to 3081 ng/μl with an average concentration of 241.2 ng/μl. Only one sample yielded RNA at a concentration of <1 ng/μl. Samples with a histological grade of CIN2+ gave RNA concentrations ranging from 0 to 1257 ng/μl with an average concentration of 137.3 ng/μl. Twelve samples (9.2%) yielded RNA at a concentration of <1 ng/μl. Samples graded CIN2+ were statistically less likely to yield high RNA concentrations (p = 0.006). There is no clear reason why the average RNA concentration was lower for samples graded CIN2+. It is possible that upon sampling the clinician may notice signs of disease and may be more likely to sample gently, leading to fewer cells collected. Alternatively, cells isolated from high grade lesions contain aptypic nuclei which could be more fragile than those from low grade lesions leading to loss of RNA upon storage. Overall, the data show a wide range of RNA yields, but we were able to prepare sufficient RNA for subsequent analysis from 95.4% of samples. Therefore, the majority of stored LBC samples proved suitable for RNA preparation.

Figure 1. RNA concentration per grade of disease and temperature.

(A) Range of RNA concentrations (ng/μl) from all samples and from those grades normal or CIN1 (≤CIN1) and CIN2 or CIN3 (CIN2+). Red horizontal bars indicate the median value in each case. Comparing histological grades <=CIN1 and CIN2+ there was a statistically significant difference in RNA concentrations obtained **p = 0.006, Student's t-test. (B) Range of RNA concentrations (ng/μl) isolated per LBC storage temperature at either -80 or -25°C. (C & D) Normal distribution Q-Q plots showing deviation from normal (Gaussian) distribution (red dotted line) at -80 or -25°C.

RNA quantification according to temperature

A total of 188 samples had been stored at -80°C while 72 samples had been stored at -25°C (Table 2). Samples stored at -80°C had been stored for between 4 and 11 years. Due to a switch in biobanking protocol in 2017 from storage at -80°C to -25°C, those samples stored at the higher temperature had been stored for between 3 and 4 years. Storage temperature (-80 or -25°C) affected the general distribution of the range of RNA concentrations obtained (Figure 1B). Although there was a greater range of RNA concentrations (range = 0–3081 ng/μl) of samples stored at -80°C compared with samples stored at -25°C (range = 0.2–1257 ng/μl), probably due to the much greater number of samples stored at -80°C, only 39.9% of samples stored at -80°C gave a final RNA concentration >100 ng/μl compared with 54.8% of samples stored at -25°C. This indicates that storage at -80°C for extended periods of time may contain less cellular material, as reported previously, or compromise RNA isolation [12]. A separate study to answer this question would require a prospective collection of samples, which is outside the scope of the present study.

Table 2. Number of samples stored at different temperatures and grade of disease.

Histology	Number of samples	Number stored at -80°C	Number stored at -25°C
Normal	60	34	26
CIN1	69	58	11
CIN2	27	26	1
CIN3	104	70	34
TOTAL	260	188	72

The normal distribution profile of RNA concentrations obtained from the samples stored at either temperature diverged from normality (Figure 1C & D) but both curves had a similar profile, therefore the divergence from normal range appeared similar in each case. Storage at either temperature allowed sufficient RNA from most LBC samples to be used for subsequent molecular analysis but storage at -25°C improved the probability of preparing greater amounts of RNA. This means that biobanked LBC samples, especially those stored at -25°C, provide a credible source of material for RNA biomarker work.

RNA integrity

Sixteen samples, all graded ‘normal’, were chosen for RNA integrity analysis. All had been stored at -80°C to represent a greater challenge since these samples had been stored for the longest period of time and it has been shown that samples stored long term at -80°C contain less cellular material [12]. According to RNA Integrity Number (RIN) values [9], RNA quality scored poorly, as previously reported [13] with RIN values ranging from 1.5 to 4.5, which is below the accepted value of >7.0 for RNA of suitable quality for next generation sequencing (Table 3). 13/16 samples had a DV200 value of ≥50 which indicates RNA of >200 nts in length (Table 3) [10]. Agilent tape station automated electrophoresis analysis showed a high degree of degradation in all samples: one or both of the ribosomal (18S and/or 28S) RNA bands were absent from the RNA preparation (Figure 2A). Figure 2B shows that the sample with the highest RIN value had a clear 28S ribosomal RNA peak and most RNA fragments were >200 nts in length. Figure 2C shows the RNA profile of the sample with the lowest RIN value. No ribosomal RNA peaks were present, and most RNA fragments were <200 nts in length. Figure 2D shows a profile representative of most samples showing a large proportion of RNA fragments smaller than 200 nts. These results indicate that, in most cases, RNA prepared from stored LBC samples is unsuitable for RNA sequencing. Future work should investigate integrity of RNA isolated from samples stored at -25°C and investigate samples representative of different grades of disease.

Figure 2. Quality analysis of RNA isolated from LBC samples stored at -80°C.

(A) Agilent tape station automated gel electrophoresis analysis of 16 RNAs (lanes 1–16) isolated from LBC samples. M, size markers in nucleotides (nts). Arrows to the right hand side indicate the electrophoretic migration positions of 28S and 16S ribosomal RNAs. Green lines indicate the end of each lane. (B) RNA fragment size profile of RNA sample 1. (C) RNA fragment size profile of RNA sample 9. (D) RNA fragment size profile of RNA sample 5. Size in nucleotides (nt) is indicated below the x-axes. The y-axes show sample integrity values. Migration positions of 28S and 16S ribosomal RNAs are indicated with blue lines.

Table 3. Concentration and RNA integrity score of the 16 randomly selected LBC samples.

Sample ID	Conc. (ng/µl)	RIN score	DV200
001	41.7	2.4	70.21%
002	62.0	4.5	59.78%
003	11.8	2	60.40%
004	31.8	2.5	88.45%
005	28.1	1.9	55.91%
006	21.8	2.3	65.93%
007	68.0	3	66.12%
008	53.0	2.1	65.42%
009	66.0	4.4	29.32%
010	55.0	2.8	37.47%
011	18.0	2.7	79.47%
012	41.1	4.1	72.88%
013	17.2	1.5	48.84%
014	36.4	3.8	60.17%
015	5.75	1.6	78.72%
016	43.2	2.3	73.11%

RNA integrity number is on a scale of 1 where 1 = RNA that is entirely degraded to 10, where 10 indicates RNA that is entirely intact. DV200 is the percentage of RNA fragments greater than 200 nts in length. A DV200 score of >70% indicates high quality RNA. A DV200 score of 50–70% indicates RNA of medium quality.

RIN: RNA integrity.

Detection & quantification of β-actin by RT-qPCR

Next, we investigated whether the RNAs isolated from the LBC samples were suitable for RT-qPCR analysis. Upon detection of β-actin RNA, 95.9% of samples gave a Ct-value that was below 31 which was the cut-off point indicated by reverse-transcriptase-negative controls (Figure 3A). Almost all samples with higher NanoDrop concentrations gave lower Ct-values as expected (Figure 3B). However, a few samples that had RNA concentrations above 1 ng/μl also had very high Ct values (Figure 3B, red dots) indicating poor quality RNA. 4.1% of samples were outside the suitable range for RT-qPCR. Thus, most stored LBC samples proved suitable for RNA preparation and subsequent RT-qPCR amplification of cellular RNAs suggesting that nucleic acid was sufficiently well preserved.

Figure 3. Range of and correlation of β-actin levels with RNA concentration.

(A) Range of Ct values obtained in RT-qPCR analysis of β-actin expression in all RNA samples. The red horizontal line indicates the median value. 95.9% of samples gave a Ct value lower than the cut off Ct = 31.0 selected by evaluating negative controls. (B) Correlation of RNA concentration with β-actin levels detected by RT-qPCR. Red dots indicate those samples (4.1%) with high Ct values which were considered invalid because they were outside the suitable range for RT-qPCR quantification.

Biomarker assessment using p16

We tested detection of known biomarker p16 in RNA from this sample cohort to determine if the RNAs were suitable for molecular biomarker analysis by RT-qPCR. p16 levels rise during cervical cancer progression due to overexpression of the HPV E7 oncoprotein which upregulates p16. p16 protein levels are an accepted sensitive and specific biomarker for clinically significant cervical disease [^14-17]. In fact, a p16 antibody-based in-clinic test shows good performance for detecting high-grade cervical disease [14,17,18]. In this study, p16 RNA was detected in 247/260 LBC samples (95%). 57/60 graded ‘normal’ by histology, 62/69 CIN1, 27/27 CIN2, 101/104 CIN3 (Figure 4). There was considerable variation in p16 expression within the different groups, especially in the ≤CIN1 population (Table 4). Relative p16 expression values of samples graded ‘normal’ ranged from 0.0001 to 0.198. Samples graded CIN1 had a range of relative expression values from 0.0001 to 0.536. Variability was somewhat reduced in samples graded CIN2+ and to a lesser extent in CIN3, although it should be noted that there were far fewer samples with a histological grade of CIN2. CIN2 values ranged from 0.0012 to 0.26. CIN3 values ranged from 0.0008 to 0.457. There was a significant statistical difference in detection levels of p16 RNA between those samples graded as normal and CIN3 (p > 0.001) and between those graded CIN1 and CIN3 (p > 0.001). Comparing CIN2 to CIN3 there was also a significance difference (p = 0.036). These observations agree with data from previous studies which used nucleic acid sequence-based amplification (NASBA) to quantify p16 transcripts during cervical disease progression [19,20]. We did not analyze HPV-positive cancer samples but we would expect, from the results of other studies, that levels of p16 mRNA would be greatest in LBC samples representing cervical cancer [19,20]. In future, co-assessment of RNA with p16 immunohistochemistry protein levels in the sample set would be of value to determine potential correlations.

Figure 4. Relative expression values of p16 RNA in each grade of disease.

p16 relative expression values were calculated by the Livak method and expressed as log 10 relative values in samples with histological grades normal, CIN1, CIN2 and CIN3. Horizontal bars indicate median values (Table 4). The statistical difference between the different groups is shown in p-values on the graph. Note that the y-axis is in log scale.

Table 4. Range and median of relative expression values for p16 detection in the four histologically defined disease groups.

Histology	Number of samples	p16 RNA quantity range (REVs)	p16 RNA quantity median (REVs)
Normal	57/60	0.0001–2.761	0.002186
CIN1	62/69	0.0001–3.049	0.004484
CIN2	27/27	0.0012–0.26	0.008441
CIN3	101/104	0.0008–0.457	0.03173

Relative expression values were calculated relative to the endogenous β-actin control and normalized to p16 expression in the W12E control RNA sample.

REV: Relative expression value.

Limitations of the study

This was a retrospective study using a random selection of samples from a limited, pre-selected cohort where we were unable to choose the properties of the LBC samples which we analyzed other than histological grade of disease. Although we examined roughly equal numbers of samples graded ≤CIN1 and CIN2+, which was our primary criterion, there was an uneven distribution of samples stored at each temperature and the length of storage was quite variable, especially for samples stored at -80°C. This is a limitation of the study, which means that we cannot make firm conclusions on how temperature and time of storage affects RNA quality and quantity. The data suggest that samples stored at -25°C for up to 4 years are more likely to yield useful RNA, but a custom-designed prospective study is required to properly address this question. The small number of samples examined for quality using a Bioanalyser is another limitation of the study. In future, it would be important to analyse samples stored at -25°C, and a much larger number of samples representing an equal number of samples graded ≤CIN1 and CIN2+.

Conclusion

Certainly, the development and evaluation of any RNA biomarker test of cervical disease will be facilitated through the assessment of well-annotated LBC samples for use in RT-qPCR assays. In this study, by assessing the range of RNA concentrations that could be extracted from 260 LBC samples, stored frozen for between 5 and 14 years, we demonstrated that stored samples provide a credible source of material for RNA biomarker work.

A triage test, as an adjunct to primary HPV testing, based on molecular testing rather than cytology suits the molecular age since it is empirical rather than subjective. The biggest attraction of an RT-qPCR-based test is its suitability for use with existing clinical laboratory protocols and automated machinery. Use of biobanked samples can expedite the assessment of RNA biomarkers detectable by RT-qPCR; the fact that routinely collected and long-stored samples can be used for this application is very encouraging.

Future perspective

Development of a set of robust RNA biomarkers to be used in RT-qPCR, a standard test in clinical laboratories, to accurately risk stratify HPV infection could potentially reduce patient care costs and reduce patient anxiety regarding disease outcome. RNA isolation and analysis is low cost, suitable for automation, and because it is eminently quantifiable, may be applicable to artificial intelligence analysis. The current study analysed clinician taken samples, however in future with the increasing move to self-collection, biobanks that incorporate such samples and stability exercises that evaluate their longevity for biomarker research will be of significant value.

Author contributions

MPJ White: investigation, writing original draft; A Stevenson: investigation; H Elasifer: resources, project administration; C Davies: investigation; K Nomikou: investigation; K Cuschieri: conceptualization, writing, review and editing; SV Graham: conceptualization, supervision writing original draft, review and editing and funding acquisition.

Financial disclosure

This work was funded by a Medical Research Council Confidence in Concept award reference: MC_PC_19039. K Cuschieri's institution has received research funding or gratis consumables to support research from the following commercial entities in the last 3 years: Cepheid, Euroimmun, GeneFirst, SelfScreen, Hiantis, Seegene, Roche, Hologic, Barinthus Biotherapeutics PLC & Daye. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

Writing disclosure

No writing assistance was utilized in the production of this manuscript.

Ethical conduct of research

Application to use the samples for this work was obtained through application to the Scottish HPV Archive and adjudication by the archive steering committee (application reference SA-066).

Acknowledgments

The authors would like to thank the Scottish HPV Archive Steering Group for considering the request for and granting access to the samples used in this study.

Competing interests disclosure

K Cuschieri has attended advisory board meetings for Hologic and Barinthus Biotherapeutics PLC, for the former, UK travel has been paid. The authors have no other competing interests or relevant affiliations with any organization or entity with the subject matter or materials discussed in the manuscript apart from those disclosed.

Additional information

Funding

This work was funded by a Medical Research Council Confidence in Concept award reference: MC_PC_19039. K Cuschieri's institution has received research funding or gratis consumables to support research from the following commercial entities in the last 3 years: Cepheid, Euroimmun, GeneFirst, SelfScreen, Hiantis, Seegene, Roche, Hologic, Barinthus Biotherapeutics PLC & Daye.