https://orcid.org/0000-0002-1078-8490
Figures
Abstract
Patient-derived cell models of dystrophic myogenesis and differentiation are valuable preclinical tools for early and mutation-based assessment of candidate therapeutic approaches. Quantitative measurement of gene expression within such models plays a key role in these studies, but normalisation of RT-qPCR data requires a panel of validated stably expressed reference genes. This study aims to identify stable reference genes for RT-qPCR assays in three human derived muscle immortalized cell lines: one healthy WT (from a 16-year-old donor), and two dystrophic lines, DMD1 (from a 11-year-old patient carrying a stop codon mutation on exon 59) and DMD2, from a 14-year-old patient carrying an exon 48–50 deletion. We screened a pool of 14 candidate genes (ACTB, HPRT1, RPL13A, RPS18, GAPDH, ALAS1, UBC, YWHAZ, IPO8, PSMC4, HSP90AB1, NONO, CSNK2A2, AP3D1), investigating stability of expression from proliferation through to 11 days of myogenic differentiation. Data were analysed using four complementary approaches (Bestkeeper, geNorm, Normfinder and DeltaCt) to determine the most appropriate references both within and between cell lines. Our study shows that RPS18, UBC, YWHAZ scored highly across all comparisons, and we therefore suggest that these three genes represent an appropriate reference panel for these human myogenic cell lines, regardless of genotype or differentiation stage.
Citation: Quarta R, Boccanegra B, Cristiano E, Ladisa A, Conte E, Ohana J, et al. (2026) Best reference genes for unbiased normalized transcript expression in normal and dystrophic human cell models of myogenesis. PLoS One 21(3): e0344973. https://doi.org/10.1371/journal.pone.0344973
Editor: Atsushi Asakura, University of Minnesota Medical School, UNITED STATES OF AMERICA
Received: June 5, 2025; Accepted: February 27, 2026; Published: March 13, 2026
Copyright: © 2026 Quarta et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the manuscript and its Supporting Information files.
Funding: This work was supported by Fondazione per il Sud (BTS grant 2018-PDR-00351 to OC) and by ERC SEEDS UNIBA (grant number 2023-UNBACLE-0245439 to BB). Part of the experimental work and the B. Boccanegra activities, were supported by funding of the Ministry of University and Research (MUR), National Recovery and Resilience Plan (NRRP), project MNESYS (PE0000006) - A Multiscale 652 integrated approach to the study of the nervous system in health and disease to ADL. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
1. Introduction
In vitro cellular models of skeletal muscle are widely used in preclinical studies of neuromuscular diseases (such as Duchenne Muscular Dystrophy, DMD) to better elucidate pathological mechanisms and as an early platform for drug testing. DMD is a X-linked inherited recessive, muscle-wasting disorder characterized by muscle degeneration, progressive disability and premature death [1]. The disease is caused by heterogeneous mutations in the gene that encodes for the dystrophin protein (DMD) which has a pivotal function in myofiber stability due to its central role in the dystrophin associated glycoprotein complex (DGC), a network of proteins that ensure a link between cytoskeleton and extracellular matrix. In the absence of dystrophin, the DGC cannot assemble, leaving myofibres vulnerable to contraction-induced injury. Repeated cycles of degeneration and regeneration lead to a complex pathogenetic cascade, characterized by chronic inflammation, oxidative stress and substitution of muscle tissue with fibrotic scarring [2]. DMD presently has no cure, but there are continued efforts within the scientific community to find therapies that correct the primary defect and restore dystrophin (such as exon-skipping), or that pharmacologically ameliorate aspects of the complex cascade initiated by dystrophin absence (such as corticosteroids to combat persistent inflammation). Animal models are widely used to obtain robust preclinical data, however the drive to refine, reduce and replace animal experiments (the 3Rs) has increasing favoured cell-based models (2D and 3D) for preclinical studies. Patient-derived cell culture models, in particular, offer a predictive and reliable tool to better support preclinical research, and isolated myoblasts can be immortalized to serve as an ongoing renewable resource. Evaluation of therapeutic efficacy (especially dystrophin restoration) typically requires myogenic differentiation of these proliferating myoblasts to form post-mitotic multinucleated myotubes, a process involving major changes in morphology and cellular activity, controlled by an elaborate network of regulatory mechanisms [3,4].
RT-qPCR is an essential quantitative method for measuring pathological or therapeutic changes in gene expression in these cell culture models, however analysis of data first requires proper normalisation to internal “reference genes” or “housekeeping genes”: genes with expression patterns that are stable regardless of experimental conditions, and which are ideally present at levels comparable to any genes of interest [5,6]. Identifying appropriate reference genes within highly dynamic scenarios such as myogenic differentiation and concomitant pathological background is, however, challenging. Evidence is mounting that other popular reference genes, such as GAPDH and ACTB, are unsuitable candidates in several cellular models [7,8]. These studies have detected GAPDH involvement in many biological activities including membrane fusion [9], vesicular transport [10] and apoptosis pathways [11–13]. These multifunctional properties of GAPDH have been demonstrated to result in its changeable pattern of expression depending on different conditions [14]. Moreover, beta-actin (ACTB) has traditionally been regarded as an endogenous housekeeping gene and has been widely used as a reference gene/protein in quantifying expression levels. However, ACTB is closely associated with a variety of cancers, and accumulating evidence indicates that ACTB is de-regulated in liver, melanoma, renal, colorectal, gastric. In addition, studies on muscle cell cultures have demonstrated that ACTB expression progressively declines as differentiation proceeds throughout myogenesis. Such fluctuations may hinder its reliability as a housekeeping gene when performing RT-PCR studies on skeletal muscle derived cell cultures during differentiation. While some genes perform well in multiple scenarios, it is increasingly becoming clear that no truly universal reference genes exist, and that appropriate references should be determined empirically for every comparative situation. Furthermore, gene expression is inherently dynamic, and mRNA can be synthesised or degraded over rapid timescales: even with ostensibly stable genes, reliance on a single reference is unwise, and panels of two or three references are advisable to buffer against rare stochastic changes in expression (ideally these genes should be selected from different functional categories, i.e., not co-ordinately regulated) [15]. In this frame, we previously conducted studies on muscle cell cultures and on muscle samples derived from commonly used dystrophic animal models for preclinical DMD research (e.g., BL10-mdx and D2-mdx mouse models, DE50-MD dogs) in order to identify the most reliable housekeeping genes for RT-PCR studies [6].
In light of previous findings, this study aimed to establish suitable reference genes for three human-derived immortalized muscle cell lines: one healthy WT (from a 16-year-old donor), and two dystrophic lines, DMD1 (from a 11-year-old patient carrying a stop codon mutation on exon 59) and DMD2, from a 14-year-old patient carrying an exon 48–50 deletion. We screened a panel of 14 candidate reference genes, measuring expression in each cell line at different stages of differentiation and in the proliferative state. This set of genes (ACTB, HPRT1, RPL13A, RPS18, GAPDH, ALAS1, UBC, YWHAZ, IPO8, PSMC4, HSP90AB1, NONO, CSNK2A2, AP3D1) was selected from GeneCards database following the criteria of level expression (high) in human muscle tissue.
Expression data were analysed using four different algorithms: geNorm [16], Bestkeeper [17], DeltaCt [18] and Normfinder [19]. Each method is capable of determining optimal references a priori, but each uses different, but complementary criteria: reliably strong reference candidates should accordingly score highly in all four approaches.
2. Materials and methods
2.1. Cell culture: maintenance and sample collection
We used three paediatric patient- and muscle-derived satellite cell lines granted from the biobank MyoLine, previously immortalized and validated by the Mouly lab [20,21]: WT (AB1190), DMD1 (stop exon 59; AB1023) and DMD2 (deletion 48–50; AB1098). Cells were periodically checked for mycoplasma contaminations using MycoStrip®myocoplasma detection kit (InvivoGen cat# rep-mys-50). Briefly, cells were quickly thawed from separate vials and resuspended in fresh growth medium (passage number 8–9, Myoline biobank, France). Once cells reached approximately 70% of confluency, they were subcultured into separate flasks. Cells were cultured in Greiner CELLSTAR® T-75 flask (GN658195) and maintained in a proliferative state (myoblast) at 37°C by using skeletal muscle cell growth medium (S1 Table in S1 File). For ‘Myoblast stage’ samples, approx. 2.5 × 105 cells of each line were collected and used for RNA extraction.
To start myogenic differentiation, cells (2.5 × 105) were seeded in petri dishes with low serum differentiation medium (S1 Table in S1 File). Cells were incubated at 37°C throughout the differentiation process. Samples were harvested for RNA isolation at two time points, corresponding to different stages of the myogenic cascade: 6 and 11 days of differentiation. Differentiation medium was partially replaced (50:50 v/v) with fresh differentiation medium every 48 hours until each endpoint. Our entire dataset consists of three biological replicates from three separate culturing sessions of each sample (N = 3 per cell line).
2.2. Immunofluorescence staining
To characterise the state of differentiation of our cells lines we performed immunofluorescence assay to detect myosin heavy chain, as a myogenic marker. 2.5 × 105 cells/well were cultured in a 6-well plate (CytoOne® - Starlab) for 6 and 11 days in differentiation medium. At each timepoint, cells were fixed in 4% paraformaldehyde in PBS for 15 min at room temperature and then washed 3 times with PBS and blocked in Blocking Buffer (2% BSA + 1% Triton X-100) for 45 min. Anti-myosin primary antibody (MF-20-S; 1:5; Developmental Studies Hybridoma Bank) was diluted in Blocking Buffer (w/Triton) and incubated at RT for 1h. After three washes in PBS, cells were incubated with Alexa FluorTM 594-conjugated secondary antibodies (1:500; Thermo Fisher Scientific) for 1 h at RT. Nuclei were stained with Pure Blu Hoechst nuclear staining dye (1:100; Bio-Rad). Mounting Medium was added before imaging. Images were acquired using a Leica THUNDER Imager (11525687) with a 20x objective.
2.3. RNA extraction, cDNA synthesis and real time PCR
Total RNA was isolated using Total RNA Purification Plus Micro Kit (Norgen Biotek Corp. NR48500) following manufacturer’s instructions. RNA purification kit includes also a genomic DNA removal step and to confirm its purity, RNA was evaluated via nanodrop spectroscopy to confirm that 260/280 ratios were >2.0, and 260/230 ratios >1.8. Samples that did not met these parameters were considered not acceptable and therefore excluded. cDNA synthesis (800 ng of RNA) was performed using the iScript™ gDNA Clear cDNA Synthesis Kit (Bio-Rad) according to the manufacturer’s protocols. Following DNAse heat-inactivation, samples were reverse transcribed using oligodT and random priming. Quantitative Real Time PCR was performed via the CFX Duet Real-Time PCR system (Bio-Rad, Hercules, CA, USA) by using Prime PCR assays SYBR® green. Each reaction was performed in technical triplicate in Hard-shell® 96 well PCR plates using 8 ng of diluted cDNA (assuming 1:1 synthesis) per well for a total volume of 10 µl. Volumes and concentrations are listed in S2 Table in S1 File. The real-time PCR cycling protocol is described in S3 Table in S1 File. [22,23] All reference gene primers were purchased from Bio-Rad (Unique Assay IDs are shown in Table 1). Quantification cycle (Cq) values were determined via regression analysis of the amplification traces, and per-sample replicate Cq values were used to calculate sample Cq (arithmetic mean). Subsequent analysis used either Cq values (Bestkeeper, deltaCt) or linearised relative quantity (RQ) values (geNorm, Normfinder, expression analysis). Where multiple reference genes were used for normalisation, RQ values for each gene were combined (geometric mean) on a per-sample basis into normalisation factor (NF) values.
2.4. Reference gene analyses
Calculation of RQ values was performed as standard: efficiency(lowest Cq – sample Cq), where efficiency is the increase in total amplicon per cycle (thus here, 2.0, corresponding to 100% efficiency). All Cq values for a given gene are accordingly converted to a linear scale (0–1). Candidate reference gene data was assessed using four different approaches (geNorm, Bestkeeper, deltaCt and Normfinder). These identify suitable references independently, but each uses different methods to determine ranking. As we have shown previously [24–26], use of multiple complementary approaches strengthens analysis: genes consistently scored highly are likely to be strong candidates, while genes that score highly under some approaches but not others can provide insights of biological relevance. Furthermore, provided that a suitably large sample set is employed, data can be assessed both as a whole, and as subsets: here we determine reference genes appropriate for both healthy and dystrophic cell cultures (all samples), but also, we investigate whether different combinations of references emerge if data are assessed when segregated by genotype (healthy vs dystrophic cell line) or differentiation stage.
GeNorm [16] uses an iterative pairwise approach: pairwise variation values are determined for all combinations of candidate genes, and then ranked accordingly. The gene with the highest total pairwise variation (M) is then discarded, and the process repeated until only a single pair of genes remain: the “best pair”. Finally, all genes are ranked from ‘worst’ to ‘best pair’ by their M values, where low M values indicate high stability. This method assumes that unrelated genes exhibiting highly correlated expression values are likely to reflect cDNA content, however it should be noted that pairs of co-ordinately regulated genes (such as ribosomal protein genes) will also tend to score highly under geNorm. The geNorm algorithm uses data linearized to relative quantities (RQ).
Bestkeeper [17] uses raw Cq data to generate a hypothetical expression profile, the ‘bestkeeper’: representing the mean behaviour of all genes in the candidate panel. Individual candidates are then assessed for pairwise correlation with this and ranked accordingly, by Pearson correlation “r”, where high values indicate high correlation. The Bestkeeper approach assumes that the consensus expression profile of multiple genes increasingly reflects overall cDNA levels, and individual genes that closely match this profile are thus suitable references.
The DeltaCt [18] approach takes per-sample Cq differences between genes (dCt = Cq gene1 – Cq gene2) and then calculates the standard deviation (SD) of these values for the entire sample set: a pair of genes that vary in a consistent manner will thus produce low SD values, regardless of differences in relative abundance. Each gene is compared with all other genes in the panel, and then ranked by their mean SD value, with low overall values indicating high stability. The assumption is that genes that exhibit no strong differential expression (i.e., good candidate references) will exhibit the least overall divergence from other genes in the candidate pool.
Finally, Normfinder [19] empirically assesses expression variance across the sample set for each gene individually, and thus does not involve any pairwise comparative elements. Normfinder can assess a dataset as a whole, but can also be conducted with data assigned to specific groups (e.g., “healthy” or “dystrophic”). This grouped analysis allows intergroup variation to be identified and treated accordingly: genes exhibiting high overall stability but also a modest but consistent group-specific bias will consequently be scored less highly than genes with greater, but group-agnostic, variability. This latter approach also identifies a “best pair”: unlike the best pair identified by geNorm, this pair of genes can include two genes that vary in a consistently opposed fashion, such that their mean behaviour represents a more stable reference than any other gene alone.
To generate an aggregate ranking, we calculated the geometric mean of the four algorithm scores on a per-gene basis. As bestkeeper ranks from low to high values, while all other methods rank from high to low values, all bestkeeper data was inverted (1-BK score) prior to use.
3. Results
3.1. Myotube characterization
First, we characterized the state of maturation of all three cell lines at day 6 and day 11 of differentiation via immunolabelling for myosin heavy chain (as a marker of terminal differentiation). Staining with the pan-myosin antibody MF20 revealed an intense signal in healthy myotubes, indicative of well-developed sarcomeric structures and efficient myogenic differentiation. In contrast, both dystrophic cell lines exhibited markedly lower MF20 labelling than healthy cells, with DMD1 showing even lower staining than DMD2 (at day 11), suggesting impaired expression or organization of contractile proteins. Morphological differences could also be detected in brightfield images (Fig 1).
Cell lines at different stages of differentiation: myoblast, day 6 and day 11 of differentiation. The bright-field images were taken with Eclipse Ti2 Nikon microscope at 10X magnification. Immunofluorescence staining showing heavy chain of myosin. Images show myosin at day 6 and day 11 of differentiation in all three cell lines. Scale bar: 100μm.
3.2. Raw Cq values
As shown in Fig 2, typical Cq values of the genes in our candidate panel spanned a range from ~15–27, with high abundance transcripts (GAPDH, ACTB, RPS18) exhibiting the lowest Cq values as expected. No marked differences associated with genotype/age were observed in raw Cq values in any candidate gene.
Individual (per sample) Cq values are indicated: solid blue circles: WT cells; solid red circles: DMD1; hollow red circles: DMD2.
3.3. GeNorm
The geNorm algorithm ranks genes by pairwise variation scores (M), identifying the pair of genes that show the most closely correlated patten of expression (the best pair). As shown in Fig 3A, assessment of the entire dataset revealed UBC and YWHAZ as the best pair, with GAPDH and ACTB also ranking highly, while HPRT1 and HSP90 were least stable. No genes exhibited M values below 0.5 (the commonly used threshold for suitability), suggesting that substantial variation existed across the sample set, either between cell lines, differentiation stages, or both. Supporting this, M values were lower when analysis was restricted by cell line (Fig 3B). Here, WT and DMD1 rankings were similar to the overall dataset, while DMD2 was markedly different, with AP3D1 and IPO8 forming the best pair (this latter gene ranked near last in WT and DMD1 cells). Following assessment of differentiation-stage-specific subsets (Fig 3C) several genes fell below the 0.5 threshold, indicating myogenesis represented the strongest contributor to variation in gene expression. Gene rankings in myoblasts and in well-differentiated cells (day 11) were broadly comparable to those in the overall dataset (with YWHAZ and GAPDH scoring highly, while HPRT1 and CSNK2 scored poorly), but rankings from cells at day 6 of differentiation (Fig 3C) showed less agreement, with AP3D1 and IPO8 being the best pair (similar to DMD2 alone). Our data thus suggested a pronounced divergence in gene stability profile at this specific timepoint.
Stability values (M) of individual genes ranked from least stable (left) to most stable (right) calculated via geNorm. (3A) Rankings are shown for the entire dataset. (3B) Rankings of individual cell lines: WT alone (upper left panel), all DMD (upper right panel), DMD1 and DMD2 only (bottom panel). (3C) Rankings of differentiation stage specific subsets (myoblast stage, day 6 and day 11).
GeNorm also determines whether additional reference genes (beyond the best pair) reduce variation (the common target being 0.2 or below). In agreement with the high M values noted above, the best pair alone was insufficient in most cases (myoblast- and day 11-specific subsets being exceptions). Use of 3 or 4 reference genes however markedly reduced variation.
3.4. Bestkeeper
The Bestkeeper algorithm compares individual genes with the consensus average of all candidate genes (the ‘bestkeeper’), ranking by Pearson correlation (thus high values indicate stronger reference candidates). As shown in Fig 4A, for our entire sample set YWHAZ, UBC and ACTB again scored highly, along with RPS18. At the other end of the scale, ALAS1, AP3D1, HPRT1 and IPO8 showed particularly poor correlations. Assessment of disease or cell-line specific datasets revealed prominent differences, however (Fig 4B), in dystrophic samples (either collectively or individually) RPS18, ACTB and UBC all ranked highly, while ALAS1, IPO8 and HPRT1 scored exceptionally poorly, but in WT samples ACTB had a markedly lower correlation coefficient, while HPRT1 was unexpectedly ranked highest. Assessment by differentiation stage (Fig 4C) suggested increasing stability with myogenic progression: several genes (such as ALAS1), exhibited negative correlation coefficients in myoblasts and at day 6, but at day 11 fully half the candidate panel pool correlated highly (r values >0.75).
Individual genes from the candidate panel ranked by Pearson correlation coefficient (r) against the “bestkeeper” (the mean expression behaviour of all genes in the candidate panel). High correlations indicate stronger reference candidates. (4A) Data is shown for the entire dataset. (4B) Genotype specific rankings. (4C) Differentiation stage specific rankings.
3.5. DeltaCt
DeltaCt assesses the extent to which each individual reference candidate varies from all other genes (on a per-sample basis), and thus genes with low mean dCt standard deviations are considered more appropriate references. As shown in Fig 5A, the highest scoring candidates for our complete sample set were RPS18, UBC, YWHAZ and GAPDH, and these genes also tended to score highly in disease- or cell-line specific datasets (Fig 5B). Rankings in WT cells identified PSMC4 and YWHAZ as the best references, however these genes were of only intermediate ranking in both DMD cell lines, where UBC was instead ranked most stable. As with Bestkeeper (see above), expression stability increased with myogenic maturity (Fig 5C): dCt values ranged from 1.3–2.0 at day 6, yet almost all were <1.0 at day 11, again indicating differentiation was the greatest contributor to expression variation. Overall, UBC, RPS18 and YWHAZ continued to rank highly.
Individual genes ranked by summed dCt standard deviations: high values (left) indicate poor candidate references, while low values (right) indicate stronger candidates. (5A) Data is shown for the entire dataset. (5B) Genotype-specific rankings. (5C) Differentiation stage specific rankings.
3.6. Normfinder
Normfinder determines empirical stability across the dataset on a gene-by-gene basis (rather than via pairwise comparisons) but can also assess data using user-assigned groups (genotype, age, etc.) to identify genes exhibiting subtle group-specific expression patterns (‘grouped analysis’). Here, low stability values correspond to highly stable genes. We first used ungrouped analysis: for our complete dataset (Fig 6A) RPS18, UBC and GAPDH were the strongest candidates, while HSP90, HPRT1 and IPO8 were weakest. Cell-line and disease-specific subsets suggested slightly different rankings (Fig 6B) but UBC and RPS18 still consistently ranked highly, and the former gene was also a strong candidate during differentiation (Fig 6C), where in agreement with other assessment methods, overall expression stability increased (lower scores) with myogenic maturation. We then assessed our complete dataset using grouped analysis (Fig 6D), taking the entire dataset but assigning data to groups either by “disease” (two groups; healthy, dystrophic), “cell-line” (three groups; WT, DMD1 and DMD2) or by “differentiation” (three groups; myoblast, day 6, day 11). Stability values were markedly higher (more variable) when grouped by differentiation stage, supporting myogenesis as the major source of dataset expression variability. Nevertheless, RPS18 and particularly UBC performed well in all cases (the latter being consistently the highest ranking, and forming one half of the ‘best pair’ in all comparisons). Similarly, HPRT1, HSP90 and IPO8 scored poorly in all comparisons.
(6A) Data is shown for the entire dataset. (6B) Genotype specific rankings. (6C) Differentiation stage subsets rankings. (6D) Rankings of complete dataset using grouped analysis, grouped by “disease”, “cell line” and “differentiation stage”.
3.7. Combined analysis
To aggregate the outputs of these four algorithms into an overall ranking, we generated the geometric mean of the scores (see methods) for each analysis (Fig 7A). RPS18, UBC YWHAZ, ACTB and GAPDH all scored highly within the full dataset, while HPRT1, HSP90 and IPO8 scored poorly. Final rankings across subset analyses were less consistent, with ACTB in particular being highly variable during myogenesis (ranked highly in dystrophic samples, myoblasts and day 11 differentiated myotubes, but second-last in day 6 myotubes) and across genotypes (high ranking in DMD cells -particularly DMD1- but not in WT). These data suggest that despite ostensibly strong performance in our full dataset, ACTB is not a suitable reference for myogenic scenarios.
Aggregate geNorm, deltaCt, Bestkeeper and Normfinder scores of all candidate genes in the entire dataset (7A), ranked from least to most stable (left to right). (7B) Combined scores for the entire dataset showing genotype specific rankings. (7C) Differentiation stage specific rankings.
Overall, UBC and (to a lesser extent) RPS18, YWHAZ and GAPDH typically emerged as strong candidates under all comparative scenarios, suggesting that these four might be a suitable reference panel.
3.8. Validation
GeNorm analysis suggested that 3–4 reference candidates were necessary for effective normalisation. Given their high rankings (both overall and within specific conditional subsets), we generated normalisation factors (NF) from UBC, RPS18 and YWHAZ, with or without further addition of GAPDH. Comparison of the resultant 4-gene and 3-gene NFs revealed minimal differences (i.e., inclusion of GAPDH was unnecessary) (S1 Fig in S1 File).
To validate this 3-gene NF, we then used a within-dataset approach as described previously: essentially, using high scoring genes to normalise consistently poor-scoring genes. As HPRT1, IPO8 and HSP90 were near-uniformly ranked last in our analysis, these were accordingly selected for normalisation to our NF. Normalisation of HPRT1 revealed distinct expression patterns across cell lines (Fig 8). In WT cells, expression was high in myoblasts but decreased during differentiation. DMD2 cells showed a similar but less pronounced pattern. In contrast, DMD1 cells exhibited lower expression in myoblasts that increased progressively throughout differentiation. IPO8 expression decreased between the myoblast stage and day 11 of differentiation across all cell lines, with more variable levels observed at day 6. HSP90 expression increased throughout differentiation in all cell lines. Notably, basal expression at the myoblast stage was approximately 10-fold lower in DMD2 compared to WT cells.
Expression of HPRT1, IPO8 and HSP90 (poor ranking candidates) across cell lines throughout differentiation using a normalisation factor derived from RPS18, UBC and YWHAZ. Two-way ANOVA with post-hoc multiple comparisons (Tukey’s). For HSP90, significant effect of differentiation (P < 0.0001) and significant effect of cell-line (P = 0.0024), for IPO8, significant effect of differentiation (P < 0.0001). For HPRT1, no significant overall effects of either differentiation or cell-line, but a significant interaction between the two (P = 0.0025). For MYOG, significant effect of differentiation (P < 0.0001) and significant effect of cell-line (P < 0.0001).
We also normalized MYOG and DMD expression using our NF. MYOG expression increased markedly as differentiation proceeded across all cell lines. DMD expression showed steady increases in the WT cell line, while both DMD lines exhibited more variable patterns.
Collectively, these data indicate that use of UBC, RPS18 and YWHAZ allows normalisation of gene expression in our healthy and dystrophic cell lines, during proliferation and throughout differentiation.
4. Discussion
Well-ordered and coordinated myogenic differentiation is essential, both for muscle development during embryogenesis and for regeneration/repair of adult skeletal muscle [27]. Muscle is a complex and heterogeneous tissue with a plethora of genetic pathways regulating the myogenic program during differentiation [28]. Study of gene expression during differentiation requires normalization, but selecting suitable reference genes within such transcriptionally dynamic scenarios is challenging. In this study, we used four complementary reference gene assessment methods (GeNorm, Bestkeeper, DeltaCt and Normfinder) to assess a large candidate gene panel in proliferating and differentiating, patient-derived, cell-culture models of healthy and dystrophic muscle. By using the consensus of all four methods, across both the entire dataset, and cell-line-, disease- or differentiation-state-specific subsets, we identified strong reference candidates suited to comparative scenarios within these models. We show that RPS18, UBC and YWHAZ consistently ranked highly within the full dataset, while HPRT1, HSP90 and IPO8 similarly scored poorly. These genes have pedigree across the broader literature: RPS18 was also found to be stable in over 20 developmental stages in skeletal muscle of Landrace pigs [29]. UBC has been identified as a reliable reference gene across various tissue types [30,31]. However, this gene has also been reported to be overexpressed in muscle tissues under certain conditions, such as glucocorticoid treatment [32]. UBC was highly stable under the conditions studied here, but exposure to glucocorticosteroids was not assessed: given these drugs represent the gold standard therapy for DMD, further investigation is warranted. YWHAZ was also found by others to be a suitable reference gene in different murine muscle tissue types as well as in a C2C12 myoblast cell line model [33].
A notable finding of our study is that expression was most variable when samples were assessed by differentiation status, rather than disease or cell-line. Variability at day 11 was comparatively low regardless of algorithm, indicating that mature myotubes are largely transcriptionally stable, however rankings (and overall stability values) were markedly less consistent at day 6 (with Bestkeeper even reporting negative Pearson correlations for several genes at this stage). Our data thus suggests that progression through myogenesis is the primary driver of variability, contributing substantially more to expression variation than patient of origin or dystrophic phenotype. This is perhaps understandable: myogenic differentiation is a highly complex process, orchestrated by distinct, hierarchical gene regulatory networks acting at specific temporal and spatial stages of development [4,34], including myogenic regulatory factors (MRFs), upstream regulators such as the Pax family and Six homeodomain transcription factors, as well as microRNAs [34–36]. While this complex scenario has well-defined start- and end-points (myoblasts and myofibres, respectively), it is not surprising that greater variability is observed between samples at intermediate myogenic stages, both because extensive transcriptional changes occur throughout the myogenic program, and because individual cell lines might not mature at equivalent rates, despite equivalent differentiation times. This was demonstrated clearly by ACTB, which was near-uniformly poorly stable at day 6, while typically ranking highly in proliferating myoblasts or differentiated myotubes. Previous studies have shown that ACTB and GAPDH exhibit the highest variability across prenatal and postnatal stages of porcine skeletal muscle development [29], and our results here extend this to human-derived samples, suggesting that these genes are not suitable references for RT-qPCR experiments assessing expression in differentiating skeletal muscle.
Expression of HPRT1 (another commonly used reference gene) was also revealed to be variable following normalization to our 3-gene NF, both between cell lines and throughout myogenesis. This was especially clear for DMD1, where the expression of this gene differed markedly from the other lines (indicating patient-related variation). This finding contrasts with other studies in which this gene was considered a suitable reference for RT-PCR experiments in murine models [33] and rat tissues [37], and again supports the notion that no gene is universally appropriate across all genotypes and tissues.
In conclusion, we show that RPS18, UBC and YWHAZ are appropriate references for normalizing gene expression across the immortalised healthy and dystrophic humans cell lines studied here (both proliferating and during differentiation). We note, however, that this does not imply they are universally applicable to human myogenic cultures, and we stress that studies using other cell lines, dystrophic genotypes or differentiation contexts would require additional validation. Conducting comprehensive housekeeping gene studies, as here and by others [6,26,38] prior to RT-qPCR analysis, is essential to identify the most suitable references for the specific scenarios under consideration (which can vary with sample type and pharmacological treatment).
References
- 1. Mercuri E, Bönnemann CG, Muntoni F. Muscular dystrophies. Lancet Lond Engl. 2019;394(10213):2025–38.
- 2. Aartsma-Rus A, Van Deutekom JCT, Fokkema IF, Van Ommen G-JB, Den Dunnen JT. Entries in the Leiden Duchenne muscular dystrophy mutation database: an overview of mutation types and paradoxical cases that confirm the reading-frame rule. Muscle Nerve. 2006;34(2):135–44. pmid:16770791
- 3. Kang J-S, Krauss RS. Muscle stem cells in developmental and regenerative myogenesis. Curr Opin Clin Nutr Metab Care. 2010;13(3):243–8. pmid:20098319
- 4. Hernández-Hernández JM, García-González EG, Brun CE, Rudnicki MA. The myogenic regulatory factors, determinants of muscle development, cell identity and regeneration. Semin Cell Dev Biol. 2017;72:10–8. pmid:29127045
- 5. Rho H-W, Lee B-C, Choi E-S, Choi I-J, Lee Y-S, Goh S-H. Identification of valid reference genes for gene expression studies of human stomach cancer by reverse transcription-qPCR. BMC Cancer. 2010;10:240. pmid:20507635
- 6. Boccanegra B, Lenti R, Mantuano P, Conte E, Tulimiero L, Piercy RJ, et al. Determination of qPCR reference genes suitable for normalizing gene expression in a novel model of Duchenne muscular dystrophy, the D2-mdx mouse. PLoS One. 2024;19(11):e0310714. pmid:39535998
- 7.
Glare EM, Divjak M, Bailey MJ, Walters EH. β-Actin and GAPDH housekeeping gene expression in asthmatic airways is variable and not suitable for normalising mRNA levels. 2023
- 8. Zhao L-M, Zheng Z-X, Zhao X, Shi J, Bi J-J, Pei W, et al. Optimization of reference genes for normalization of the quantitative polymerase chain reaction in tissue samples of gastric cancer. Asian Pac J Cancer Prev. 2014;15(14):5815–8. pmid:25081706
- 9. Glaser PE, Gross RW. Rapid plasmenylethanolamine-selective fusion of membrane bilayers catalyzed by an isoform of glyceraldehyde-3-phosphate dehydrogenase: discrimination between glycolytic and fusogenic roles of individual isoforms. Biochemistry. 1995;34(38):12193–203. pmid:7547960
- 10. Tisdale EJ. Glyceraldehyde-3-phosphate dehydrogenase is required for vesicular transport in the early secretory pathway. J Biol Chem. 2001;276(4):2480–6. pmid:11035021
- 11. Sawa A, Khan AA, Hester LD, Snyder SH. Glyceraldehyde-3-phosphate dehydrogenase: nuclear translocation participates in neuronal and nonneuronal cell death. Proc Natl Acad Sci U S A. 1997;94(21):11669–74. pmid:9326668
- 12. Sunaga K, Takahashi H, Chuang DM, Ishitani R. Glyceraldehyde-3-phosphate dehydrogenase is over-expressed during apoptotic death of neuronal cultures and is recognized by a monoclonal antibody against amyloid plaques from Alzheimer’s brain. Neurosci Lett. 1995;200(2):133–6. pmid:8614562
- 13. Hara MR, Agrawal N, Kim SF, Cascio MB, Fujimuro M, Ozeki Y, et al. S-nitrosylated GAPDH initiates apoptotic cell death by nuclear translocation following Siah1 binding. Nat Cell Biol. 2005;7(7):665–74. pmid:15951807
- 14. Moe TK, Ziliang J, Barathi A, Beuerman RW. Differential expression of glyceraldehyde-3-phosphate dehydrogenase (GAPDH), beta actin and hypoxanthine phosphoribosyltransferase (HPRT) in postnatal rabbit sclera. Curr Eye Res. 2001;23(1):44–50. pmid:11821985
- 15. Bustin SA, Benes V, Garson JA, Hellemans J, Huggett J, Kubista M, et al. The MIQE guidelines: minimum information for publication of quantitative real-time PCR experiments. Clin Chem. 2009;55(4):611–22. pmid:19246619
- 16. Vandesompele J, Preter KD, Roy NV, Paepe AD. Accurate normalization of real-time quantitative RT-PCR data by geometric averaging of multiple internal control genes. Genome Biol. 2002;3(7):RESEARCH0034.
- 17. Pfaffl MW, Tichopad A, Prgomet C, Neuvians TP. Determination of stable housekeeping genes, differentially regulated target genes and sample integrity: BestKeeper--Excel-based tool using pair-wise correlations. Biotechnol Lett. 2004;26(6):509–15. pmid:15127793
- 18. Silver N, Best S, Jiang J, Thein SL. Selection of housekeeping genes for gene expression studies in human reticulocytes using real-time PCR. BMC Mol Biol. 2006;7:33. pmid:17026756
- 19. Andersen CL, Jensen JL, Ørntoft TF. 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. 2004;64(15):5245–50. pmid:15289330
- 20. Thorley M, Duguez S, Mazza EMC, Valsoni S, Bigot A, Mamchaoui K, et al. Skeletal muscle characteristics are preserved in hTERT/cdk4 human myogenic cell lines. Skelet Muscle. 2016;6(1):43. pmid:27931240
- 21. Mamchaoui K, Trollet C, Bigot A, Negroni E, Chaouch S, Wolff A, et al. Immortalized pathological human myoblasts: towards a universal tool for the study of neuromuscular disorders. Skelet Muscle. 2011;1:34. pmid:22040608
- 22. Boccanegra B, Mantuano P, Conte E, Cerchiara AG, Tulimiero L, Quarta R, et al. LKB1 signaling is altered in skeletal muscle of a Duchenne muscular dystrophy mouse model. Dis Model Mech. 2023;16(7):dmm049930. pmid:37427454
- 23. Boccanegra B, Cappellari O, Mantuano P, Trisciuzzi D, Mele A, Tulimiero L, et al. Growth hormone secretagogues modulate inflammation and fibrosis in mdx mouse model of Duchenne muscular dystrophy. Front Immunol. 2023;14:1119888. pmid:37122711
- 24. Hildyard JCW, Wells DJ. Identification and validation of quantitative PCR reference genes suitable for normalizing expression in normal and dystrophic cell culture models of myogenesis. PLoS Curr. 2014;6:ecurrents.md.faafdde4bea8df4aa7d06cd5553119a6. pmid:24634799
- 25. Hildyard JCW, Finch AM, Wells DJ. Identification of qPCR reference genes suitable for normalizing gene expression in the mdx mouse model of Duchenne muscular dystrophy. PLoS One. 2019;14(1):e0211384. pmid:30699165
- 26. Hildyard JCW, Wells DJ, Piercy RJ. Identification of qPCR reference genes suitable for normalising gene expression in the developing mouse embryo. Wellcome Open Res. 2022;6:197. pmid:35509373
- 27. Cossu G, Biressi S. Satellite cells, myoblasts and other occasional myogenic progenitors: possible origin, phenotypic features and role in muscle regeneration. Semin Cell Dev Biol. 2005;16(4–5):623–31. pmid:16118057
- 28. Bentzinger CF, Wang YX, Rudnicki MA. Building muscle: molecular regulation of myogenesis. Cold Spring Harb Perspect Biol. 2012;4(2):a008342. pmid:22300977
- 29. Niu G, Yang Y, Zhang Y, Hua C, Wang Z, Tang Z, et al. Identifying suitable reference genes for gene expression analysis in developing skeletal muscle in pigs. PeerJ. 2016;4:e2428. pmid:27994956
- 30. Chua SL, See Too WC, Khoo BY, Few LL. UBC and YWHAZ as suitable reference genes for accurate normalisation of gene expression using MCF7, HCT116 and HepG2 cell lines. Cytotechnology. 2011;63(6):645–54. pmid:21850463
- 31. Mosley Y-YC, HogenEsch H. Selection of a suitable reference gene for quantitative gene expression in mouse lymph nodes after vaccination. BMC Res Notes. 2017;10(1):689. pmid:29208024
- 32. Marinovic AC, Zheng B, Mitch WE, Price SR. Ubiquitin (UbC) expression in muscle cells is increased by glucocorticoids through a mechanism involving Sp1 and MEK1. J Biol Chem. 2002;277(19):16673–81. pmid:11872750
- 33. Ma J, Chen J, Gan M, Chen L, Zhao Y, Niu L, et al. Comparison of reference gene expression stability in mouse skeletal muscle via five algorithms. PeerJ. 2022;10:e14221. pmid:36275473
- 34. Buckingham M, Rigby PWJ. Gene regulatory networks and transcriptional mechanisms that control myogenesis. Dev Cell. 2014;28(3):225–38. pmid:24525185
- 35. Gagan J, Dey BK, Dutta A. MicroRNAs regulate and provide robustness to the myogenic transcriptional network. Curr Opin Pharmacol. 2012;12(3):383–8.
- 36. Liu N, Williams AH, Maxeiner JM, Bezprozvannaya S, Shelton JM, Richardson JA, et al. microRNA-206 promotes skeletal muscle regeneration and delays progression of Duchenne muscular dystrophy in mice. J Clin Invest. 2012;122(6):2054–65. pmid:22546853
- 37. Wang X, Zhao H, Ni J, Pan J, Hua H, Wang Y. Identification of suitable reference genes for gene expression studies in rat skeletal muscle following sciatic nerve crush injury. Mol Med Rep. 2019;19(5):4377–87. pmid:30942461
- 38. Hildyard JCW, Taylor-Brown F, Massey C, Wells DJ, Piercy RJ. Determination of qPCR Reference Genes Suitable for Normalizing Gene Expression in a Canine Model of Duchenne Muscular Dystrophy. J Neuromuscul Dis. 2018;5(2):177–91. pmid:29614692