Selecting internal references is important for normalizing the loading quantity of samples in quantitative reverse-transcription PCR (qRT-PCR). In the present study, a systematic evaluation of reference genes among nine hepatocellular carcinoma (HCC) cell lines was conducted. After screening the microarray assay data of ten HCC cell lines, 19 candidate reference genes were preselected and then evaluated by qRT-PCR, together with ACTB, GAPDH, HPRT1 and TUBB. The expression evenness of these candidate genes was evaluated using RefFinder. The stabilities of the reference genes were further evaluated under different experimental perturbations in Huh-7 and MHCC-97L, and the applicability of the reference genes was assessed by measuring the mRNA expression of CCND1, CCND3, CDK4 and CDK6 under sorafenib treatment in Huh-7. Results showed that TFG and SFRS4 are among the most reliable reference genes, and ACTB ranks third and acts quite well as a classical choice, whereas GAPDH, HPRT1 and TUBB are not proper reference genes in qRT-PCR assays among the HCC cell lines. SFRS4, YWHAB, SFRS4 and CNPY3 are the most stable reference genes of the MHCC-97L under the perturbations of chemotherapy, oxidative stress, starvation and hypoxia respectively, whereas YWHAB is the most stable one of Huh-7 under all perturbations. GAPDH is recommended as a reference gene under chemotherapy perturbations. YWHAB and UBE2B, TMED2 and TSFM, and GAPDH and TSFM are the two best reference genes under oxidative stress, starvation and hypoxia perturbations respectively. TSFM is stable in both cell lines across all the perturbations.

Background

Hepatocellular carcinoma (HCC) is the most common type of primary liver cancer in adults. It is also the fourth most frequently diagnosed and the third leading cause of deaths among all cancers [1]. The molecular mechanisms underlying the initiation and progression of HCC remain elusive; however, they could most probably result from the changes in the expression levels of several susceptible genes. These cancer-related genes can definitely construct characteristic signal pathways and protein–protein interaction networks, which begin with the occurrence and development of HCC.

It would be helpful to maximize the impact of studies on gene expression. An important component of studying the HCC mechanisms is detecting the expression pattern in the transcriptome scale through high-throughput profiling assays such as microarray and RNA-seq. However, these high-throughput results require further validation in most of the circumstances. Real-time quantitative reverse-transcription PCR (qRT-PCR) has been proven to be a precise and flexible method for measuring a limited number of gene expressions [25]. The following two working principles have been used to determine the RNA expression in qRT-PCR: absolute and relative quantifications [68]. Absolute quantification primarily relies on the calibration curve for estimating the copy numbers of transcripts [9]. However, estimation bias may be inevitable because of the uncertainty of the initial value assignments. In the majority of cases, researchers determine the difference between concerned samples without considering the absolute abundance of mRNAs and then apply the relative quantification by assessing the fold change between the concerned samples; this process is also termed as the ΔCt or ΔCq method [10,11].

Several factors need to be considered in the relative quantification, which include the quality and amount of mRNA, the efficiency of the reverse transcriptase, the primer amplification efficiency and the systematic and random variations [1214]. Proper normalization is an important component of the precise measurement of mRNA and must deal with the cell count or the differences in tissue volume, the RNA concentration and purity variations, the efficiency of the reverse transcriptase and other amplification factors. Although a gene with absolutely stable expression never appears across all samples or treatment regimens, some relatively invariable ones are used as internal references [1519]. For example, ACTB, GAPDH, HPRT1 and TUBB are frequently used as reference genes in qRT-PCR and Northern blot assay [2027]. However, the mRNA levels of GAPDH are not always constant [2831] and may contribute to diverse cellular functions [32]. Thus, it is necessary to screen the most stably expressed reference gene(s) for a comparison of each individual expression.

In the present study, the most stably expressed 19 reference candidate genes were preselected from the microarray data of ten HCC cell lines and the stabilities of these putative reference genes together with ACTB, GAPDH, HPRT1 and TUBB were validated by qRT-PCR.

Methods

Cell lines and treatments

The following nine HCC cell lines were used in the present study: Huh-7, Hep3B, PLC/PRF/5, MHCC-97L, MHCC-97H, HCCLM3, SNU-398, SNU-449 and SNU-475. All the eight cell lines, except Huh-7, were from hepatitis B virus (HBV)-infected HCC patients. MHCC-97L, MHCC-97H and HCCLM3 were obtained from the Liver Cancer Institute, Fudan University (Shanghai, China) [33]. Huh-7 (catalogue number TCHu182) [34], Hep3B (catalogue number TCHu106) [35] and PLC/PRF/5 (catalogue number TCHu119) [36] were obtained from Shanghai Cellular Institute of Chinese Academy of Sciences (Shanghai, China). These six cell lines were grown in Dulbecco’s modified Eagle’s medium (DMEM; HyClone, U.S.A.) and supplemented with 10% FBS (Biochrom, Germany) and 1% penicillin/streptomycin (HyClone, U.S.A.). SNU-398 (ATCC® number: CRL-2233™), SNU-449 (ATCC® number: CRL-2234™) and SNU-475 (ATCC® number: CRL-2236™) were obtained from the American Type Culture Collection (ATCC) [37] and were cultured in Roswell Park Memorial Institute (medium) (RPMI)-1640 (HyClone, U.S.A.) supplemented with 10% FBS and 1% penicillin streptomycin. All the nine cell lines were maintained at 37°C in a 5% CO2 humidified incubator. The cells were grown to 80–90% confluence and harvested three times within ten passages. All the cells were periodically checked to ensure that there is no mycoplasma contamination.

Huh-7 and MHCC-97L cells were respectively treated with cisplatin and sorafenib (Selleck, U.S.A.) dissolved in DMSO for at least 24 h. The final concentrations of cisplatin and sorafenib were 7 and 5 µmol/l in Huh-7 cells respectively, while the final concentration of both cisplatin and sorafenib was 10 µmol/l in MHCC-97L cells. Huh-7 and MHCC-97L cells were treated with H2O2, with the respective final concentrations being 100 and 2 mmol/l. The starvation of Huh-7 and MHCC-97L cells corresponded with that of the cell lines cultured in 1.5 g/l glucose medium and compared with the control cells grown in 4.5 g/l high-glucose DMEM. Hypoxia was stimulated in the cell lines cultured in 2% O2 incubator for at least 24 h. The cell-counting kit-8 (CCK-8) cell proliferation assays (Dojindo, Japan) of Huh-7 and MHCC-97L cells were performed under the hypoxia stimulations (Supplementary Figure S3). Cell cycle analysis of the original and 5 µmol/l sorafenib-treated Huh-7 cells was conducted by flow cytometry using the Cell Cycle and Apoptosis Analysis Kit (Biyuntian, China) (Supplementary Figure S4).

Preselection of reference candidate genes from microarray data

A total of 48 gene expression microarray (Affymetrix HG U133 Plus 2.0 Array) datasets of ten HCC cell lines (Supplementary Table S1) were collected. The datasets of MHCC-97L, MHCC-97H, HCCLM3 and HCCLM6 and those of two of Hep3B expression assays were obtained from our recent work [38]. The datasets of the other five cell lines and those of four expression assays of Hep3B were obtained from ArrayExpress [39] and Gene Expression Omnibus (GEO) databases [40,41]. Based on the pipeline of calculating the evenness of expression values across all samples, the candidate reference genes with low variation and high levels of microarray hybridization signal intensity (MAS5.0) were screened [42]. The following cutoffs were used: coefficient of variation (CV) <0.11, mean intensity and maximum fold-change (MFC) = Max (Ii)/Min (Ii) <1.4, where Ii denotes the intensity of the gene expression in the arrays of the i-th samples. Max and Min are the maximum and minimum values respectively. The candidate reference genes were analysed based on both probe-level (probe intensities obtained directly from microarray results) and gene-level (sum of all probe intensities) intensities.

RNA extraction and cDNA synthesis

A total of 5 × 106 cells were collected from each cell line. Total RNA was extracted using TRIzol reagent (Invitrogen) according to the manufacturer’s instructions. The quality and quantity of RNA were measured using NanoDrop ND-1000 (Thermo Scientific, Wilmington, DE, U.S.A.) through OD260/280 and OD260/230 ratios. A total of 500 ng of total RNA was reversed using a PrimeScript® RT reagent kit through poly-dT and random hexamer primers (TaKaRa DRR037) after treating with RNase-free DNase I (TaKaRa DRR2270). Each sample was replicated three times (biological replicates).

Design of primers and evaluation of amplification efficiencies

The primers of the 16 genes, namely, AIMP1, ANP32B, BCL2L13, CNPY3, CUGBP1, ENY2, HNRNPC, RPL22, SEC61B, SFRS4, TFG, TMED2, TROVE2, UBE2D3, UBE2V2 and YWHAB, were designed by PerlPrimer v1.1.16. The primers of TSFM and UBE2B were designed by NCBI primer BLAST. The primers of UBE2N and GAPDH were designed by Primer Premier 5.0. The published primer set was used for ACTB (NM_001101) [29]. The primers of HPRT1 (164518913c1), TUBB (34222261c1), CCND1 (77628152c1), CCND3 (209915551c1), CDK4 (345525417c1) and CDK6 (223718133c1) were selected from PrimerBank (https://pga.mgh.harvard.edu/primerbank/). To avoid the contamination of genomic DNA, the design of most of the primers was composed of cross exon–intron junctions and boundaries that were as far as possible from each other. Agarose gel electrophoresis and melting curve analysis were performed to assess the expected length of the PCR products. The sequencing of the amplicons confirmed the unique expected products. The amplification efficiencies and the specificity of these primer sets were evaluated using standard curve analysis of five-fold serial dilutions and dissociation curves according to previous descriptions [43,44]. Finally, the most efficient primers were selected (Supplementary Table S3).

Real-time qRT-PCR

Real-time qRT-PCR was performed using the SYBR® Premix Ex Taq™ Kit (TaKaRa DRR041) according to the manufacturer’s instructions, with minor modifications, in 96-well reaction plates using the Applied Biosystems 7500 Real-Time PCR System. Each sample was prepared in a total volume of 25 µl containing 1 µl of 5 µmol/l primer mix (200 nmol/l of each primer), 12.5 µl SYBR Green master mix, 0.5 µl rhodamine X (ROX) Reference Dye II and 9.5 µl RNase/DNase-free sterile water. The initial denaturation was carried out at 95°C for 10 s, followed by 40 cycles at 95°C for 5 s and 60°C for 34 s. The fluorescence data were collected in the 60°C extension phase, and each cell line was harvested in three biological replicates. Each sample was measured in three technical replicates.

The following five experimental perturbations were used in the present study: cisplatin treatment, sorafenib treatment, H2O2 treatment, starvation by low glucose and hypoxia. Under each condition, both Huh-7 and MHCC-97L cells were harvested from the three biological replicates, and each sample was measured in three technical replicates. The applicability of the reference genes was assessed by measuring the mRNA expression of CCND1, CCND3, CDK4 and CDK6 under sorafenib treatment in Huh-7 cells. The ΔΔCt algorithm was used to calculate the fold changes compared with those in the control samples.

Evaluation of qRT-PCR results

A web-based tool RefFinder was used to evaluate the stability of the candidate reference genes [45]. RefFinder integrates the geNorm [46], NormFinder [47,48], ΔCt [49] and BestKeeper [50] and evaluates the most stable gene. Each algorithm uses slightly different methods that are aimed at estimating both the intra- and the intergroup expression variations and allow the ranking of candidate genes based on the instability score [51]. geNorm operates on the assumption that the expression ratio of two ideal candidate genes is constant. NormFinder can indicate the optimal number of reference genes by calculating the accumulated S.D. ΔCt is used to compare the relative expressions of ‘pairs of genes’ by comparing their ΔCt values. Therefore, ΔCt algorithm can analyse large panels of genes based on the ‘process of elimination’. BestKeeper determines whether the candidate genes are differentially expressed under an applied treatment based on the crossing points. geNorm and NormFinder use the stability (actually instability) value, ΔCt uses the average of S.D., BestKeeper uses the S.D. of the crossing points and RefFinder uses the geometric mean of ranking values obtained from the above-mentioned four methods. These indexes are termed as instability scores (the smaller, the better) in the present study.

The result of each candidate gene is dependent on all the other ones because geNorm, NormFinder, ΔCt and RefFinder use all candidate genes to compute the instability score, indicating that the same candidate genes with different competitive partners will be scored differently. A candidate gene list with even expression levels is beneficial to assess the global stable centre. Therefore, the iterative method was used to compute the instability score using RefFinder and its submethods, namely, geNorm, NormFinder and ΔCt, by excluding the most unstable ones in each computing cycle.

Statistical analysis and visualization

The bubble charts were plotted using Excel. The matrix of the generalized pair plot was generated using the function ggscatmat of the R package GGally [52]. The violin plot was drawn using the web-tool BoxPlotR (http://shiny.chemgrid.org/boxplotr/), which uses the shiny package from RStudio. The reciprocals of the instability scores or the reciprocals of CV of microarray intensities, namely, the stability scores, were used to calculate the Pearson correlation coefficients between the various methods.

Ethics approval and consent to participate

Experiment materials used in this research were mostly the HCC cell lines. Among these cell lines, Huh-7 [30], Hep3B [31] and PLC/PRF/5 [32] were obtained from Shanghai Cellular Institute of Chinese Academy of Sciences (Shanghai, China). MHCC-97L, MHCC-97H and HCCLM3 were obtained from the Liver Cancer Institute, Fudan University (Shanghai, China) [29]. SNU-398, SNU-449 and SNU-475 were obtained from the ATCC.

Availability of data and materials

All data generated or analysed during the present study are included in this published article (and its supplementary information files).

Results

Preselection

Previous studies have always used an arbitrary method of preselection of reference genes to be evaluated [2,12,14,19,47], thereby leading to missing of potential reference genes. Because of the availability of high-throughput transcriptome profiling technologies, such as gene expression microarray and RNA-seq, highly comprehensive candidate gene lists can be obtained from those profiling data [42,53]. In the present study, 19 candidate genes (Figure 1 and Supplementary Table S2) were preselected according to the evenness and the high expression level criteria across all the ten cell line 48 microarray datasets. Moreover, 10, 15 and 6 of the 19 candidate genes were preselected from gene, probe and both levels respectively (Figure 1C). ACTB, GAPDH, HPRT1 and TUBB were added to the candidate gene list for the next step of evaluation.

Results of preselecting the reference candidate genes from the expression microarray datasets

Figure 1
Results of preselecting the reference candidate genes from the expression microarray datasets

Items including gene (light-blue bubbles) and probe (light-green bubbles) are denoted by bubble charts of (A) all in Supplementary Table S2 and (B) pass the preselection cutoff (CV <0.11, , MFC <0.14). X- and Y-axes are denoted as CV and averaged intensity level respectively. Bubble areas are proportional to MCF values; (C) 10, 15 and 6 of the 19 (gene) candidates were preselected from gene, probe and both levels respectively.

Figure 1
Results of preselecting the reference candidate genes from the expression microarray datasets

Items including gene (light-blue bubbles) and probe (light-green bubbles) are denoted by bubble charts of (A) all in Supplementary Table S2 and (B) pass the preselection cutoff (CV <0.11, , MFC <0.14). X- and Y-axes are denoted as CV and averaged intensity level respectively. Bubble areas are proportional to MCF values; (C) 10, 15 and 6 of the 19 (gene) candidates were preselected from gene, probe and both levels respectively.

qRT-PCR evaluation

qRT-PCR was carried out to measure the Ct values of the candidate genes (Figure 2A), and the results showed that ACTB, TUBB and GAPDH have the highest transcript abundances (lowest Ct values) and UBE2N and GAPDH have the largest variations in transcript abundances among the measurements of the nine cell lines with multiple three biological replicates. The RefFinder, which integrates four algorithms, namely, geNorm, NormFinder, ΔCt and BestKeeper, was used to evaluate the expression stability from the Ct values. Through iterative assessment in RefFinder, it was observed that TFG and SFRS4 constantly maintained the top stable positions and finally reached the top two, followed by ACTB that finally achieved the third position (Figure 2B and Supplementary Table S4). The individual four submethods revealed slightly different results from those obtained in RefFinder (Figure 2C and Supplementary Table S4). All the three algorithms, except BestKeeper, selected TFG and SFRS4 as the most stably expressed genes in these HCC cell lines, whereas BestKeeper selected TMED2 and ACTB as the top two genes. The iterative ranking results of NormFinder (Supplementary Figure S1A and Table S4) and geNorm (Supplementary Figure S1B and Table S4) were steadier than those of RefFinder. ΔCt maintained the instability scores during iterative computations, because a candidate with a high instability score is only determined by all the candidates with low instability scores; thus, the exclusion of the most instable ones does not affect the scores of the others. BestKeeper produces steady results without the need for the iterative strategy. In conclusion, TFG and SFRS4 are the most stable reference genes among these nine HCC cell lines. Regarding the conservative choice, ACTB as the third most stable gene performs quite good. All the four algorithms denoted GAPDH, HPRT1 and TUBB as unstable genes in the HCC cell lines.

Results of assessing the internal reference genes using qRT-PCR

Figure 2
Results of assessing the internal reference genes using qRT-PCR

The sample size of each measurement was three biological replicates multiplied by three technical replicates (n=9). (A) Violin plots of Ct values of the 23 candidate genes arranged from left to right according to the ascending order of S.D. of Ct. White circles show the medians; box limits indicate the 25th and 75th percentiles; whiskers extend 1.5-times the interquartile range from the 25th and 75th percentiles; polygons represent density estimates of data and extend to extreme values. (B) Circular assessment of reference candidate genes by RefFinder. The gene with the highest instability score (geometric mean of ranking values) will be excluded in the next calculating cycle until the final two genes win. (C) Final assessments of the internal reference genes by RefFinder and its submethods, such as geNorm, NormFinder, ΔCt and BestKeeper, compared with the microarray-preselected results. Microarray-preselected results use the CV of intensities from gene (ArrayG) or probe level. The latter can select a probe with the minimum CV (ArrayPCV) or the maximum intensity (ArrayPInt).

Figure 2
Results of assessing the internal reference genes using qRT-PCR

The sample size of each measurement was three biological replicates multiplied by three technical replicates (n=9). (A) Violin plots of Ct values of the 23 candidate genes arranged from left to right according to the ascending order of S.D. of Ct. White circles show the medians; box limits indicate the 25th and 75th percentiles; whiskers extend 1.5-times the interquartile range from the 25th and 75th percentiles; polygons represent density estimates of data and extend to extreme values. (B) Circular assessment of reference candidate genes by RefFinder. The gene with the highest instability score (geometric mean of ranking values) will be excluded in the next calculating cycle until the final two genes win. (C) Final assessments of the internal reference genes by RefFinder and its submethods, such as geNorm, NormFinder, ΔCt and BestKeeper, compared with the microarray-preselected results. Microarray-preselected results use the CV of intensities from gene (ArrayG) or probe level. The latter can select a probe with the minimum CV (ArrayPCV) or the maximum intensity (ArrayPInt).

Stable reference genes under different experimental perturbations

The stabilities of the reference genes under experimental perturbations are important. The evaluation of the genes under such experimental perturbations shows complicated patterns. The candidate reference gene behaviour was slightly different in both MHCC-97L and Huh-7 (Figure 3 and Supplementary Table S5). In brief, SFRS4, YWHAB, SFRS4 and CNPY3 were the most stable reference genes of MHCC-97L under the perturbations of chemotherapy, oxidative stress, starvation and hypoxia respectively, whereas YWHAB was the most stable reference gene of Huh-7 under all the perturbations. From the viewpoint of the perturbations, GAPDH can be recommended as a reference gene under chemotherapy perturbations (Figure 3A), while YWHAB and UBE2B were the two best reference genes under oxidative stress (Figure 3B), TMED2 and TSFM were the two best reference genes under starvation (Figure 3C), GAPDH and TSFM were the two best reference genes under hypoxia (Figure 3D). TSFM showed stable expression in both cell lines across all the perturbations. Moreover, the second echelon should be TMED2 and TROVE2.

Stability evaluation of reference candidate genes in MHCC-97L and Huh-7 under different experimental perturbations

Figure 3
Stability evaluation of reference candidate genes in MHCC-97L and Huh-7 under different experimental perturbations

Reference candidate gene performances are denoted by bubble charts under (A) chemotherapy perturbations by cisplatin and sorafenib, (B) oxidative stress by H2O2, (C) starvation by low glucose and (D) hypoxia. The sample size of each measurement was three biological replicates multiplied by three technical replicates (n=9). Instability scores from circular assessment of reference candidate genes by RefFinder. The gene with the highest instability score (geometric mean of ranking values) will be excluded in the next calculating cycle until the final two genes win. X- and Y-axes are denoted as ReFinder instability scores from MHCC-97L and Huh-7 respectively. Bubble areas are proportional to ReFinder instability scores from the nine HCC cell lines (Figure 2C). TSFM, the best performing reference gene of both cell lines across all the perturbations, is denoted by red bubbles.

Figure 3
Stability evaluation of reference candidate genes in MHCC-97L and Huh-7 under different experimental perturbations

Reference candidate gene performances are denoted by bubble charts under (A) chemotherapy perturbations by cisplatin and sorafenib, (B) oxidative stress by H2O2, (C) starvation by low glucose and (D) hypoxia. The sample size of each measurement was three biological replicates multiplied by three technical replicates (n=9). Instability scores from circular assessment of reference candidate genes by RefFinder. The gene with the highest instability score (geometric mean of ranking values) will be excluded in the next calculating cycle until the final two genes win. X- and Y-axes are denoted as ReFinder instability scores from MHCC-97L and Huh-7 respectively. Bubble areas are proportional to ReFinder instability scores from the nine HCC cell lines (Figure 2C). TSFM, the best performing reference gene of both cell lines across all the perturbations, is denoted by red bubbles.

To check the applicability of the reference genes under sorafenib treatment in Huh-7 cells, the mRNA expression level changes of CCND1, CCND3, CDK4 and CDK6, which are believed to be down-regulated, were calculated based on the 23 references respectively [54,55]. Most of the 23 candidate genes were suitable for use under the sorafenib perturbation experiments in Huh-7 cells (Figure 4). The Huh-7 cell viability curve declaimed that the proliferation of cells treated with sorafenib was decreased compared with that in the control (Supplementary Figure S3). The cell cycle analysis declaimed that Huh-7 cells were arrested in G1 phase under sorafenib treatment compared with that in the control (P<0.01, Supplementary Figure S4).

Results of checking the applicability of reference genes under sorafenib treatment in the Huh-7

Figure 4
Results of checking the applicability of reference genes under sorafenib treatment in the Huh-7

Dot plot of fold-changes of the four reporter genes under sorafenib treatment that were calculated based on the 23 reference genes. Reference genes arranged from left to right according to the ascending order of performances under chemotherapy perturbations in Huh-7 cells. Circles show the mean values, and error bars show the S.E.M.

Figure 4
Results of checking the applicability of reference genes under sorafenib treatment in the Huh-7

Dot plot of fold-changes of the four reporter genes under sorafenib treatment that were calculated based on the 23 reference genes. Reference genes arranged from left to right according to the ascending order of performances under chemotherapy perturbations in Huh-7 cells. Circles show the mean values, and error bars show the S.E.M.

Discussion

A suitable reference gene should have following two characteristics: (i) it should not have a tissue specific to a gene expression, which is a valid reference gene, and should be expressed under almost all biological and experimental conditions, and (ii) it should have a low CV of expression levels [30]. Previous studies have recommended SFRS4 and TBP as the reference genes in HCV-induced HCC or breast carcinomas [29,43]. Among some putative reference genes selected experientially, HMBS and TBP were verified to be suitable for reference genes in HCC [4,56]. In the present study, SFRS4 and TFG were screened out by large-scale genomic dataset mining and qRT-PCR as the most stable reference genes in the HCC cell lines. TFG is a TRK-fused gene coded protein, which is a conserved regulator of protein secretion and oncogenesis and has been implicated in neuropathies [57,58]. SFRS4 encodes a member of the arginine-/serine-rich splicing factor family, which functions in mRNA processing. SFRS4 expressed in patients with alcoholic liver disease is a relatively stable reference gene used in qPCR technique and is not influenced by steatosis, alcoholic hepatitis, significant fibrosis and cirrhosis [59,60].

ACTB is a highly conserved protein and is one of the two non-muscular cytoskeletal actins. GAPDH is an important glycolytic enzyme and can catalyse the production of 1,3-bisphosphoglycerate from glyceraldehyde 3-phosphate. TUBB encodes the β-tubulin protein and acts as a structural component of microtubules [61]. HPRT1 is a transferase in the purine salvage pathway and catalyses the conversion of hypoxanthine into inosine monophosphate and guanine into guanosine monophosphate [62]. These four genes are generally used as reference genes in qRT-PCR [2023]. However, they are not always stable in several samples or conditions. A previous study has shown significantly different expressions of ACTB between malignant and non-malignant pairs, upon examination of 16 potential reference gene candidates in 17 untreated prostate carcinomas [15]. Moreover, the expression levels of ACTB and GAPDH were examined in 80 normal and tumour samples from colorectal, breast, prostate, skin and bladder tissues using qRT-PCR, which revealed that these genes were unsuitable as single reference genes [63]. HPRT1 has been evaluated in HBV-related HCC studies, but the results were found to be inconsistent [64,65]. TUBB has been evaluated as a reference gene in qRT-PCR assays among cell lines and pertubations but the results showed that it is not the most suitable reference gene [66,67]. In the present study, the expression of ACTB was found to be quite stable, whereas the other three genes exhibited dramatic variations among the HCC cell lines. To achieve a highly reliable measurement of gene expression, the combinatorial use of two or more reference genes (TFG/SFRS4/ACTB) is recommended.

No correlations were found between microarray-preselected and qRT-PCR-evaluated results (Supplementary Figure S2). The results indicated that microarray-based quantification is not sufficiently accurate to distinguish subtle differences in expression stability among genes with a similar performance and is only applicable for an approximate preselection. Reference candidate genes were preselected from the microarray data of ten HCC cell lines but were evaluated only in nine cell lines by qRT-PCR because of the unavailability of HCCLM6 in this step. This situation may introduce some variations between the preselected and the evaluated results. geNorm, NormFinder and ΔCt are highly correlated with each other and have higher consistencies than that of RefFinder. BestKeeper has a lower consistency than those of the other three algorithms and RefFinder, and it is the only algorithm that strongly recommended ACTB.

When the stable reference genes were screened under different stimulations, the results of the experimental perturbations showed complicated patterns. None of the candidate genes satisfied the requirements of both the evaluated cell lines under all the perturbations. Considering all the cell lines across all the perturbations, TSFM was the most balanced reference gene, followed by TMED2 and TROVE2. The TSFM gene encodes a mitochondrial translation elongation factor. The encoded protein is an enzyme that catalyses the exchange of guanine nucleotides on the translation elongation factor Tu during the elongation step of mitochondrial protein translation. A mutation in this gene results in severe infantile liver failure [68] and oxidative phosphorylation enzyme deficiency syndrome [69].

Huh-7 has more consistent reference behaviours than those of MHCC-97L. The best reference gene of Huh-7 cells constantly stuck to YWHAB across all the perturbations. YWHAB encodes a protein that belongs to the 14-3-3 family of proteins, which mediate signal transduction by binding to phosphoserine-containing proteins. YWHAB was initially reported as a reference gene in the present study, but another 14-3-3 family member YWHAZ has been selected as a suitable reference gene in several cell lines [66].

Most of the 23 candidates are all suitable under the sorafenib perturbation experiments in Huh-7. Although YWHAB and other top ranked candidates, such as UBE2B, UBE2D3 and GAPDH, lead to smaller fold changes of CCND1, CCND3, CDK4 and CDK6 than those of SFRS4, BCL2L13 and ACTB, we cannot consider the former group (YWHAB etc.) to perform worse than the latter group (SFRS4 etc.). In fact, the expressions of CCND1, CCND3, CDK4 and CDK6 decreased after sorafenib treatment, and the accurate fold-changes remained unknown. The exceedingly reduced value of a target gene can be generated from the actual increment of the reference genes.

It should be mentioned that the present study did not utilize a commonly used liver cell line, HepG2. Although HepG2 cell line and its derivate HepG2/C3A have been annotated as ‘hepatocellular carcinoma’ in the ATCC, the HepG2 cells were in fact isolated from liver biopsy specimens of primary hepatoblastoma (HB, originated from immature liver precursor cells) rather than HCC (originated from mature hepatocytes) [70], and are non-tumorigenic [71]. We believe that HepG2 is out of range of HCC cell lines [72] and have excluded it in our HCC studies designedly.

Although the present study was conducted only within the scope of HCC cell lines, the protocol can be easily applied to other cell lines or specimens. Furthermore, the evaluation of internal reference genes in the present study was conducted only among certain HCC cell lines and perturbations. If internal reference genes applicable to a broader range, such as those between HCC and normal liver or under special stimulations, were to be found, more sample types should be included.

Conclusion

The combinational use of two or more reference genes, such as TFG/SFRS4/ACTB, is recommended in qRT-PCR assays of HCC cell lines. GAPDH, YWHAB/UBE2B, TMED2/TSFM and GAPDH/TSFM are recommended as reference genes under the perturbations of chemotherapy, oxidative stress, starvation and hypoxia respectively.

Competing interests

The authors declare that there are no competing interests associated with the manuscript.

Funding

This work was supported by the National Key Research and Development Program of China [grant number 2016YFC0901903]; the National Natural Science Foundation of China [grant numbers 81602131, 91629301]; the National Basic Research Program of China [grant number 2013CB910802]; and the International S&T Cooperation Program of China [grant number 2014DFB30020]. None of the funding bodies had any part in the design of the study and collection, analysis and interpretation of data or in writing the manuscript.

Author contribution

Y.L. conducted the qRT-PCR assays, analysed and interpreted the stability scores of four algorithms regarding the candidate reference genes. Z.Q. analysed the qRT-PCR raw data. L.C. and J.Z. performed the biological replicated qRT-PCR assays. L.Z. participated in the revision of the article. F.Z. designed the overall project, collected the microarray data and selected the candidate genes, and also was a major contributor in writing the manuscript. All authors read and approved the final manuscript.

Abbreviations

     
  • ATCC

    American Type Culture Collection

  •  
  • ACTB

    actin beta

  •  
  • CCND1

    cyclin D1

  •  
  • CCND3

    cyclin D3

  •  
  • CDK4

    cyclin dependent kinase 4

  •  
  • CDK6

    cyclin dependent kinase 6

  •  
  • CNPY3

    canopy FGF signaling regulator 3

  •  
  • Cq

    quantification cycle

  •  
  • Ct

    threshold cycle

  •  
  • CV

    coefficient of variation

  •  
  • DMEM

    Dulbecco’s modified Eagle’s medium

  •  
  • GAPDH

    glyceraldehyde-3-phosphate dehydrogenase

  •  
  • HBV

    hepatitis B virus

  •  
  • HCV

    hepatitis C virus

  •  
  • HCC

    hepatocellular carcinoma

  •  
  • HMBS

    hydroxymethylbilane synthase

  •  
  • HPRT1

    hypoxanthine phosphoribosyltransferase 1

  •  
  • qRT-PCR

    quantitative reverse-transcription PCR

  •  
  • SRSF4

    serine and arginine rich splicing factor 4

  •  
  • TBP

    TATA-box binding protein

  •  
  • TFG

    TRK-fused gene

  •  
  • TMED2

    transmembrane P24 trafficking protein 2

  •  
  • TROVE2

    TROVE domain family member 2

  •  
  • TSFM

    Ts translation elongation factor, mitochondrial

  •  
  • TUBB

    tubulin beta class I

  •  
  • UBE2B

    ubiquitin conjugating enzyme E2 B

  •  
  • UBE2N

    ubiquitin conjugating enzyme E2 N

  •  
  • YWHAB

    tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein beta

References

1
Siegel
R.L.
,
Miller
K.D.
and
Jemal
A.
(
2016
)
Cancer statistics
.
CA Cancer J. Clin.
66
,
7
30
2
Rho
H.W.
,
Lee
B.C.
,
Choi
E.S.
,
Choi
I.J.
,
Lee
Y.S.
and
Goh
S.H.
(
2010
)
Identification of valid reference genes for gene expression studies of human stomach cancer by reverse transcription-qPCR
.
BMC Cancer
28
,
240
3
Fedrigo
O.
,
Warner
L.R.
,
Pfefferle
A.D.
,
Babbitt
C.C.
,
Cruz-Gordillo
P.
and
Wray
G.A.
(
2010
)
A pipeline to determine RT-QPCR control genes for evolutionary studies: application to primate gene expression across multiple tissues
.
PLoS ONE
5
,
e12545
4
Cicinnati
V.R.
,
Shen
Q.
,
Sotiropoulos
G.C.
,
Radtke
A.
,
Gerken
G.
and
Beckebaum
S.
(
2008
)
Validation of putative reference genes for gene expression studies in human hepatocellular carcinoma using real-time quantitative RT-PCR
.
BMC Cancer
8
,
350
5
Sanders
R.
,
Mason
D.J.
,
Foy
C.A.
and
Huggett
J.F.
(
2014
)
Considerations for accurate gene expression measurement by reverse transcription quantitative PCR when analysing clinical samples
.
Anal. Bioanal. Chem.
406
,
6471
6483
6
Wagner
E.M.
(
2013
)
Monitoring gene expression: quantitative real-time rt-PCR
.
Methods Mol. Biol.
1027
,
19
45
7
Vu
H.L.
,
Troubetzkoy
S.
,
Nguyen
H.H.
,
Russell
M.W.
and
Mestecky
J.
(
2000
)
A method for quantification of absolute amounts of nucleic acids by (RT)-PCR and a new mathematical model for data analysis
.
Nucleic Acids Res.
28
,
E18
8
Nakamura
H.
,
Dan
S.
,
Akashi
T.
,
Unno
M.
and
Yamori
T.
(
2007
)
Absolute quantification of four isoforms of the class I phosphoinositide-3-kinase catalytic subunit by real-time RT-PCR
.
Biol. Pharm. Bull.
30
,
1181
1184
9
Shahsiah
R.
,
Abdollahi
A.
,
Azmoudeh Ardalan
F.
,
Haghi-Ashtiani
M.T.
,
Jahanzad
I.
and
Nassiri Toosi
M.
(
2010
)
Result variation and efficiency kinetics in real-time PCR
.
Acta Med. Iran
48
,
279
282
10
Livak
K.J.
and
Schmittgen
T.D.
(
2001
)
Analysis of relative gene expression data using real-time quantitative PCR and the 2- ΔΔCT method
.
Methods
25
,
402
408
11
Pfaffl
M.W.
(
2001
)
A new mathematical model for relative quantification in real-time RT-PCR
.
Nucleic Acids Res.
29
,
e45
12
Mehta
R.
,
Birerdinc
A.
,
Hossain
N.
,
Afendy
A.
,
Chandhoke
V.
,
Younossi
Z.
et al
(
2010
)
Validation of endogenous reference genes for qRT-PCR analysis of human visceral adipose samples
.
BMC Mol. Biol.
11
,
39
13
Combes
J.D.
,
Grelier
G.
,
Laversanne
M.
,
Voirin
N.
,
Chabaud
S.
,
Ecochard
R.
et al
(
2009
)
Contribution of cell culture, RNA extraction, and reverse transcription to the measurement error in quantitative reverse transcription polymerase chain reaction-based gene expression quantification
.
Anal. Biochem.
393
,
29
35
14
Vazquez-Blomquist
D.
,
Fernández
J.R.
,
Miranda
J.
,
Bello
C.
,
Silva
J.A.
,
Estrada
R.C.
et al
(
2012
)
Selection of reference genes for use in quantitative reverse transcription PCR assays when using interferons in U87MG
.
Mol. Biol. Rep.
39
,
11167
11175
15
Ohl
F.
,
Jung
M.
,
Xu
C.
,
Stephan
C.
,
Rabien
A.
,
Burkhardt
M.
et al
(
2005
)
Gene expression studies in prostate cancer tissue: which reference gene should be selected for normalization?
J. Mol. Med. (Berl.)
83
,
1014
1024
16
Jung
M.
,
Ramankulov
A.
,
Roigas
J.
,
Johannsen
M.
,
Ringsdorf
M.
,
Kristiansen
G.
et al
(
2007
)
In search of suitable reference genes for gene expression studies of human renal cell carcinoma by real-time PCR
.
BMC Mol. Biol.
8
,
47
17
Ohl
F.
,
Jung
M.
,
Radonić
A.
,
Sachs
M.
,
Loening
S.A.
and
Jung
K.
(
2006
)
Identification and validation of suitable endogenous reference genes for gene expression studies of human bladder cancer
.
J. Urol.
175
,
1915
1920
18
Wang
Y.
,
Han
Z.
,
Yan
S.
,
Mao
A.
,
Wang
B.
,
Ren
H.
et al
(
2010
)
Evaluation of suitable reference gene for real-time PCR in human umbilical cord mesenchymal stem cells with long-term in vitro expansion
.
In Vitro Cell. Dev. Biol. Anim.
46
,
595
599
19
Wang
S.
,
Li
J.
,
Zhang
A.
,
Liu
M.
and
Zhang
H.
(
2011
)
Selection of reference genes for studies of porcine endometrial gene expression on gestational day 12
.
Biochem. Biophys. Res. Commun.
408
,
265
268
20
Zhang
X.
,
Ding
L.
and
Sandford
A.J.
(
2005
)
Selection of reference genes for gene expression studies in human neutrophils by real-time PCR
.
BMC Mol. Biol.
6
,
4
21
Huggett
J.
,
Dheda
K.
,
Bustin
S.
and
Zumla
A.
(
2005
)
Real-time RT-PCR normalisation; strategies and considerations
.
Genes Immun.
6
,
279
284
22
Edwards
D.R.
and
Denhardt
D.T.
(
1985
)
A study of mitochondrial and nuclear transcription with cloned cDNA probes. Changes in the relative abundance of mitochondrial transcripts after stimulation of quiescent mouse fibroblasts
.
Exp. Cell Res.
157
,
127
143
23
Lim
Y.C.
,
Desta
Z.
,
Flockhart
D.A.
and
Skaar
T.C.
(
2005
)
Endoxifen (4-hydroxy-N-desmethyl-tamoxifen) has anti-estrogenic effects in breast cancer cells with potency similar to 4-hydroxy-tamoxifen
.
Cancer Chemother. Pharmacol.
55
,
471
478
24
Kulkarni
H.
,
Göring
H.H.
,
Diego
V.
,
Cole
S.
,
Walder
K.R.
,
Collier
G.R.
et al
(
2012
)
Association of differential gene expression with imatinib mesylate and omacetaxine mepesuccinate toxicity in lymphoblastoid cell lines
.
BMC Med. Genomics
5
,
37
25
Xu
H.
,
Bionaz
M.
,
Sloboda
D.M.
,
Ehrlich
L.
,
Li
S.
,
Newnham
J.P.
et al
(
2015
)
The dilution effect and the importance of selecting the right internal control genes for RT-qPCR: a paradigmatic approach in fetal sheep
.
BMC Res. Notes
8
,
58
26
Zampieri
M.
,
Ciccarone
F.
,
Guastafierro
T.
,
Bacalini
M.G.
,
Calabrese
R.
,
Moreno-Villanueva
M.
et al
(
2010
)
Validation of suitable internal control genes for expression studies in aging
.
Mech. Ageing Dev.
131
,
89
95
27
Durrenberger
P.F.
,
Fernando
F.S.
,
Magliozzi
R.
,
Kashefi
S.N.
,
Bonnert
T.P.
,
Ferrer
I.
et al
(
2012
)
Selection of novel reference genes for use in the human central nervous system: a BrainNet Europe Study
.
Acta Neuropathol.
124
,
893
903
28
Valenti
M.T.
,
Bertoldo
F.
,
Dalle Carbonare
L.
,
Azzarello
G.
,
Zenari
S.
,
Zanatta
M.
et al
(
2006
)
The effect of bisphosphonates on gene expression: GAPDH as a housekeeping or a new target gene?
BMC Cancer
6
,
49
29
Waxman
S.
and
Wurmbach
E.
(
2007
)
De-regulation of common housekeeping genes in hepatocellular carcinoma
.
BMC Genomics
8
,
243
30
Zhou
L.
,
Lim
Q.E.
,
Wan
G.
and
Too
H.P.
(
2010
)
Normalization with genes encoding ribosomal proteins but not GAPDH provides an accurate quantification of gene expressions in neuronal differentiation of PC12 cells
.
BMC Genomics
11
,
75
31
Lin
J.
and
Redies
C.
(
2012
)
Histological evidence: housekeeping genes beta-actin and GAPDH are of limited value for normalization of gene expression
.
Dev. Genes Evol.
222
,
369
376
32
Sirover
M.A.
(
1999
)
New insights into an old protein: the functional diversity of mammalian glyceraldehyde-3-phosphate dehydrogenase
.
Biochim. Biophys. Acta
1432
,
159
184
33
Tang
Z.Y.
,
Ye
S.L.
,
Liu
Y.K.
,
Qin
L.X.
,
Sun
H.C.
,
Ye
Q.H.
et al
(
2004
)
A decade’s studies on metastasis of hepatocellular carcinoma
.
J. Cancer Res. Clin. Oncol.
130
,
187
196
34
Nakabayashi
H.
,
Taketa
K.
,
Miyano
K.
,
Yamane
T.
and
Sato
J.
(
1982
)
Growth of human hepatoma cells lines with differentiated functions in chemically defined medium
.
Cancer Res.
42
,
3858
3863
35
Knowles
B.B.
and
Aden
D.P.
(
1983
)
Human hepatoma derived cell line, process for preparation thereof, and uses therefor
.
U.S. Pat.
439,133
36
Alexander
J.J.
,
Bey
E.M.
,
Geddes
E.W.
and
Lecatsas
G.
(
1976
)
Establishment of a continuously growing cell line from primary carcinoma of the liver
.
S. Afr. Med. J.
50
,
2124
2128
37
Park
J.G.
,
Lee
J.H.
,
Kang
M.S.
,
Park
K.J.
,
Jeon
Y.M.
,
Lee
H.J.
et al
(
1995
)
Characterization of cell lines established from human hepatocellular carcinoma
.
Int. J. Cancer
62
,
276
282
38
Shen
H.
,
Zhong
F.
,
Zhang
Y.
,
Yu
H.
,
Liu
Y.
,
Qin
L.
et al
(
2015
)
Transcriptome and proteome of human hepatocellular carcinoma reveal shared metastatic pathways with significant genes
.
Proteomics
15
,
1793
1800
39
Kolesnikov
N.
,
Hastings
E.
,
Keays
M.
,
Melnichuk
O.
,
Tang
Y.A.
,
Williams
E.
et al
(
2015
)
ArrayExpress update--simplifying data submissions
.
Nucleic Acids Res.
43
,
D1113
D1116
40
Barrett
T.
,
Wilhite
S.E.
,
Ledoux
P.
,
Evangelista
C.
,
Kim
I.F.
,
Tomashevsky
M.
et al
(
2013
)
NCBI GEO: archive for functional genomics data sets–update
.
Nucleic Acids Res.
41
,
D991
D995
41
Edgar
R.
,
Domrachev
M.
and
Lash
A.E.
(
2002
)
Gene Expression Omnibus: NCBI gene expression and hybridization array data repository
.
Nucleic Acids Res.
30
,
207
210
42
Cheng
W.C.
,
Chang
C.W.
,
Chen
C.R.
,
Tsai
M.L.
,
Shu
W.Y.
,
Li
C.Y.
et al
(
2011
)
Identification of reference genes across physiological states for qRT-PCR through microarray meta-analysis
.
PLoS ONE
6
,
e17347
43
Lyng
M.B.
,
Laenkholm
A.V.
,
Pallisgaard
N.
and
Ditzel
H.J.
(
2008
)
Identification of genes for normalization of real-time RT-PCR data in breast carcinomas
.
BMC Cancer
8
,
20
44
Kosir
R.
,
Acimovic
J.
,
Golicnik
M.
,
Perse
M.
,
Majdic
G.
,
Fink
M.
et al
(
2010
)
Determination of reference genes for circadian studies in different tissues and mouse strains
.
BMC Mol. Biol.
11
,
60
45
Xie
F.
,
Xiao
P.
,
Chen
D.
,
Xu
L.
and
Zhang
B.
(
2012
)
miRDeepFinder: a miRNA analysis tool for deep sequencing of plant small RNAs
.
Plant Mol. Biol.
80
,
75
84
46
Hellemans
J.
,
Mortier
G.
,
De Paepe
A.
,
Speleman
F.
and
Vandesompele
J.
(
2007
)
qBase relative quantification framework and software for management and automated analysis of real-time quantitative PCR data
.
Genome Biol.
8
,
R19
47
Wang
Q.
,
Ishikawa
T.
,
Michiue
T.
,
Zhu
B.L.
,
Guan
D.W.
and
Maeda
H.
(
2012
)
Stability of endogenous reference genes in postmortem human brains for normalization of quantitative real-time PCR data: comprehensive evaluation using geNorm, NormFinder, and BestKeeper
.
Int. J. Legal Med.
126
,
943
952
48
Andersen
C.L.
,
Jensen
J.L.
and
Ørntoft
T.F.
(
2004
)
Normalization of real-time quantitative reverse transcription-PCR data: a model-based variance estimation approach to identify genes suited for normalization, applied to bladder and colon cancer data sets
.
Cancer Res.
64
,
5245
5250
49
Silver
N.
,
Best
S.
,
Jiang
J.
and
Thein
S.L.
(
2006
)
Selection of housekeeping genes for gene expression studies in human reticulocytes using real-time PCR
.
BMC Mol. Biol.
7
,
33
50
Pfaffl
M.W.
,
Tichopad
A.
,
Prgomet
C.
and
Neuvians
T.P.
(
2004
)
Determination of stable housekeeping genes, differentially regulated target genes and sample integrity: BestKeeper–Excel-based tool using pair-wise correlations
.
Biotechnol. Lett.
26
,
509
515
51
Demidenko
N.V.
,
Logacheva
M.D.
and
Penin
A.A.
(
2011
)
Selection and validation of reference genes for quantitative real-time PCR in buckwheat (Fagopyrum esculentum) based on transcriptome sequence data
.
PLoS ONE
6
,
e19434
52
Emerson
J.W.
,
Green
W.A.
,
Schloerke
B.
,
Crowley
J.
,
Cook
D.
,
Hofmann
H.
et al
(
2013
)
The Generalized Pairs Plot
.
J. Comput. Graph. Stat.
22
,
79
91
53
Kolquist
K.A.
,
Schultz
R.A.
,
Furrow
A.
,
Brown
T.C.
,
Han
J.Y.
,
Campbell
L.J.
et al
(
2011
)
Microarray-based comparative genomic hybridization of cancer targets reveals novel, recurrent genetic aberrations in the myelodysplastic syndromes
.
Cancer Genet.
204
,
603
628
54
Liu
L.
,
Cao
Y.
,
Chen
C.
,
Zhang
X.
,
McNabola
A.
,
Wilkie
D.
et al
(
2006
)
Sorafenib blocks the RAF/MEK/ERK pathway, inhibits tumor angiogenesis, and induces tumor cellapoptosis in hepatocellular carcinoma model PLC/PRF/5
.
Cancer Res.
66
,
11851
11858
55
Broecker-Preuss
M.
,
Müller
S.
,
Britten
M.
,
Worm
K.
,
Schmid
K.W.
,
Mann
K.
et al
(
2015
)
Sorafenib inhibits intracellular signaling pathways and induces cell cycle arrest and cell death in thyroid carcinoma cells irrespective of histological origin or BRAF mutational status
.
BMC Cancer
15
,
184
56
Ceelen
L.
,
De Spiegelaere
W.
,
David
M.
,
De Craene
J.
,
Vinken
M.
,
Vanhaecke
T.
et al
(
2011
)
Critical selection of reliable reference genes for gene expression study in the HepaRG cell line
.
Biochem. Pharmacol.
81
,
1255
1261
57
Ishiura
H.
,
Sako
W.
,
Yoshida
M.
,
Kawarai
T.
,
Tanabe
O.
,
Goto
J.
et al
(
2012
)
The TRK-fused gene is mutated in hereditary motor and sensory neuropathy with proximal dominant involvement
.
Am. J. Hum. Genet.
91
,
320
329
58
Witte
K.
,
Schuh
A.L.
,
Hegermann
J.
,
Sarkeshik
A.
,
Mayers
J.R.
,
Schwarze
K.
et al
(
2011
)
TFG-1 function in protein secretion and oncogenesis
.
Nat. Cell Biol.
13
,
550
558
59
Boujedidi
H.
,
Bouchet-Delbos
L.
,
Cassard-Doulcier
A.M.
,
Njiké-Nakseu
M.
,
Maitre
S.
,
Prévot
S.
et al
(
2012
)
Housekeeping gene variability in the liver of alcoholic patients
.
Alcohol. Clin. Exp. Res.
36
,
258
266
60
Gabriel
M.
,
Delforge
Y.
,
Deward
A.
,
Habraken
Y.
,
Hennuy
B.
,
Piette
J.
et al
(
2015
)
Role of the splicing factor SRSF4 in cisplatin-induced modifications of pre-mRNA splicing and apoptosis
.
BMC Cancer
15
,
227
61
Liu
N.
,
Xiong
Y.
,
Ren
Y.
,
Zhang
L.
,
He
X.
,
Wang
X.
et al
(
2015
)
Proteomic profiling and functional characterization of multiple post-translational modifications of tubulin
.
J. Proteome Res.
14
,
3292
3304
62
Nguyen
K.V.
,
Naviaux
R.K.
and
Nyhan
W.L.
(
2017
)
Human HPRT1 gene and the Lesch-Nyhan disease: substitution of alanine for glycine and inversely in the HGprt enzyme protein
.
Nucleosides Nucleotides Nucleic Acids
36
,
151
157
63
de Kok
J.B.
,
Roelofs
R.W.
,
Giesendorf
B.A.
,
Pennings
J.L.
,
Waas
E.T.
,
Feuth
T.
et al
(
2005
)
Normalization of gene expression measurements in tumor tissues: comparison of 13 endogenous control genes
.
Lab. Invest.
85
,
154
159
64
Fu
L.Y.
,
Jia
H.L.
,
Dong
Q.Z.
,
Wu
J.C.
,
Zhao
Y.
,
Zhou
H.J.
et al
(
2009
)
Suitable reference genes for real-time PCR in human HBV-related hepatocellular carcinoma with different clinical prognoses
.
BMC Cancer
9
,
49
65
Liu
S.
,
Zhu
P.
,
Zhang
L.
,
Ding
S.
,
Zheng
S.
,
Wang
Y.
et al
(
2013
)
Selection of reference genes for RT-qPCR analysis in tumor tissues from male hepatocellular carcinoma patients with hepatitis B infection and cirrhosis
.
Cancer Biomark.
13
,
345
349
66
Chua
S.L.
,
See Too
W.C.
,
Khoo
B.Y.
and
Few
L.L.
(
2011
)
UBC and YWHAZ as suitable reference genes for accurate normalisation of gene expression using MCF7, HCT116 and HepG2 cell lines
.
Cytotechnology
63
,
645
654
67
Diesel
L.F.
,
dos Santos
B.P.
,
Bellagamba
B.C.
,
Pretto Neto
A.S.
,
Ely
P.B.
,
Meirelles Lda
S.
et al
(
2015
)
Stability of reference genes during tri-lineage differentiation of human adipose-derived stromal cells
.
J. Stem Cells
10
,
225
242
68
Vedrenne
V.
,
Galmiche
L.
,
Chretien
D.
,
de Lonlay
P.
,
Munnich
A.
and
Rötig
A.
(
2012
)
Mutation in the mitochondrial translation elongation factor EFTs results in severe infantile liver failure
.
J. Hepatol.
56
,
294
297
69
Smeitink
J.A.
,
Elpeleg
O.
,
Antonicka
H.
,
Diepstra
H.
,
Saada
A.
,
Smits
P.
et al
(
2006
)
Distinct clinical phenotypes associated with a mutation in the mitochondrial translation elongation factor EFTs
.
Am. J. Hum. Genet.
79
,
869
877
70
Aden
D.P.
,
Fogel
A.
,
Plotkin
S.
,
Damjanov
I.
and
Knowles
B.B.
(
1979
)
Controlled synthesis of HBsAg in a differentiated human liver carcinoma-derived cell line
.
Nature
282
,
615
616
71
Knowles
B.B.
,
Howe
C.C.
and
Aden
D.P.
(
1980
)
Human hepatocellular carcinoma cell lines secrete the major plasma proteins and hepatitis B surface antigen
.
Science
209
,
497
499
72
Qiu
G.H.
,
Xie
X.
,
Xu
F.
,
Shi
X.
,
Wang
Y.
and
Deng
L.
(
2015
)
Distinctive pharmacological differences between liver cancer cell lines HepG2 and Hep3B
.
Cytotechnology
67
,
1
12

Author notes

*

These authors contributed equally to this work.

This is an open access article published by Portland Press Limited on behalf of the Biochemical Society and distributed under the Creative Commons Attribution License 4.0 (CC BY).

Supplementary data