Introduction

The prevalence of obesity in the United States has increased dramatically from approximately 12% in 1991 to approximately 42% three decades later1,2. Projections from the World Obesity Atlas suggest that by 2035, over 4 billion people worldwide, representing more than 51% of the global population, will have a BMI ≥ 25 kg/m² and be considered overweight or obese3. Obesity is a chronic, multifactorial, and complex disease resulting from excess accumulation of body fat. Environmental factors are mostly involved in the progression of this epidemic, e.g., excess food intake and a sedentary lifestyle, although genetic components cannot be excluded4,5. Obesity raises the risk of many comorbidities, e.g., diabetes, hypertension, and insulin resistance, thus directly or indirectly affecting the functioning of both the liver and kidney. The most common conditions affecting these organs in the context of an excessive BMI include non-alcoholic fatty liver disease (NAFLD) and chronic kidney disease (CKD)6. Obesity affects multiple metabolic functions of the liver and kidney. With prolonged obesity, inflammation develops as the influx of free fatty acids (FFAs) into the liver overwhelms physiological adaptive mechanisms, leading to endoplasmic reticulum (ER) stress, reactive oxygen species (ROS) formation, and liver cell dysfunction and damage by lipotoxicity7. Additionally, abnormal fat deposition and inflammation can influence the gene expression of different pathways and processes in the liver and kidney8,9. Increasing evidence suggests that the presence and severity of NAFLD are associated with an increased incidence of CKD10,11. Studies demonstrate that obesity may influence drug pharmacokinetics12; thus, it is necessary to address existing knowledge gaps to optimize therapeutic interventions and improve drug safety and efficacy.

Scientific community uses a variety of tools and techniques, including transcriptomics, proteomics, and metabolomics, to study the molecular mechanisms underlying obesity and the resulting pathophysiological changes that accompany this condition13,14. qPCR is a “gold standard” in mRNA quantification and an accurate technology for detecting changes in gene expression due to its high specificity, high sensitivity, and good repeatability15. To obtain accurate qPCR results, it is necessary to complete several technical steps and perform a range of quality controls. However, variations in the amount of starting material, enzymatic efficiency of reverse transcriptase, and the presence of reaction inhibitors can lead to quantification errors. Normalization of the expression data allows for a more accurate comparison between samples. It reduces non-specific variability associated with differences in RNA quantity and quality, usually by relative quantification, in which the expression level of a target gene is determined relative to the transcript of another gene, a so-called reference gene (RG) or housekeeping gene (HKG). The results are expressed as a ratio of target to reference (an endogenous control). The reliability of the normalized data largely depends on the robustness of the RG16. A stable reference gene is expected not to show inter-subject variation and not to be affected by the disease state or experimental conditions. The selection of appropriate reference genes (RGs) that remain relatively constant in specific tissue types in patients with particular conditions is one of the essential steps for accurate data analysis, and such data are in high demand17,18. Despite the growing need for increased reliability and accuracy of qPCR data, some common RGs are often used without proper validation or assessment of their stability in various tissues and pathophysiological conditions19,20,21,22,23.

The RefFinder24 is an online platform that integrates several algorithms, including geNorm25, NormFinder26, BestKeeper27, and comparative ΔCt method26, that have been developed to select suitable RGs. Based on the rankings from each program, RefFinder assigns an appropriate stability value to an individual gene and calculates the geometric mean of their values for the overall final ranking of the reference genes to determine their stability.

The current study aimed to identify stable RGs for gene expression analysis based on qPCR in human liver and kidney tissue for obesity research. We used seven candidates for reference genes: ACTB ( beta-actin), B2M (beta-2-microglobulin), GAPDH (glyceraldehyde-3-phosphate dehydrogenase), HPRT1 (hypoxanthine phosphoribosyltransferase 1), RPLP0 (ribosomal protein lateral stalk subunit P0), 18S rRNA, and PPIA (peptidylprolyl isomerase A; also known as cyclophilin A) and analyzed them in liver samples from 15 lean individuals and 17 donors with BMI ≥ 25 and in kidney samples from 13 lean individuals and 15 donors with BMI ≥ 25.

Results

The expression stability of seven selected reference gene candidates was determined by measuring the relative expression of mRNA in kidney and liver tissue samples. Table 1 summarizes the anthropometric and clinical characteristics of the 42 study subjects. The kidney and liver samples were collected from 28 to 32 deceased donors, respectively (18 subjects provided both tissues). Table 2 provides descriptive statistics of the raw Ct values of all tested reference gene candidates.

Table 1 Clinical characteristics of the studied population.
Full size table
Table 2 The descriptive statistics of the raw Ct values of all tested reference gene candidates for the kidney and the liver.
Full size table

geNorm analysis

Vandesompele et al.25 assumed that the expression ratio of two ideal control genes is identical in all samples; in real-life conditions, the increasing variation in the expression ratio reflects the decreasing variation stability. Their study emphasized the errors of using a single control gene for normalization and indicated that the geometric mean of multiple control genes is required for accurate measures. geNorm analysis based on the gene-stability measure, M (elevated M values suggest higher pairwise variation), showed that the best reference genes in kidney tissue were RPLP0 and GAPDH. In contrast, the least stable gene was 18S rRNA (Fig. 1). In the liver tissue, ACTB & GAPDH were ranked best control genes, while – similarly to kidney tissue – 18S rRNA was found to be the gene with the least stable expression (Fig. 1).

Fig. 1
figure 1

The gene expression stability reflected by the gene-stability measure M was calculated using the geNorm method and the RefFinder online tool.

Full size image

NormFinder analysis

The mathematical method elaborated by Andersen et al.26 allows the identification of stably expressed genes among a set of reference gene candidates for each experiment. This strategy estimates intra- and intergroup expression variation between candidates and indicates the best control gene for normalization. The lower the derived stability value, the more stable the gene expression. In NormFinder analysis, the most stable reference genes were HPRT1, followed by RPLP0 in the kidney, and GAPDH, followed by RPLP0 in the liver (Fig. 2).

Fig. 2
figure 2

The stability value was calculated using the NormFinder method and the RefFinder tool to select the most stable reference genes in the kidney (blue graph) and liver (green graph).

Full size image

BestKeeper

BestKeeper is an Excel-based software developed by Pfaffl et al.27 that computes the descriptive statistics (Table 3) for each reference gene candidate. The algorithm uses the standard deviation (SD) of the mean Cq of each tested housekeeping gene and evaluates gene stability by pairwise bivariate correlations of Cq values of each gene; the x-fold expression change is calculated. Based on the calculated SD and coefficient of variation (CV) values, the gene candidates can be initially ranked from least to most stably expressed (highest and lowest variation, respectively). For genes considered stably expressed, the BestKeeper Index is calculated as the geometric mean of HKGs CP values where the degree of the root is the total number of HKGs included. The Pearson correlation coefficient between each HKG candidate and the index is then calculated. Pfaffl et al. assumed that any studied gene with SD [± Ct] > 1 can be considered inconsistent. In our set of studied control genes, BestKeeper selected 18S rRNA as the most stable one in the kidney and the liver tissue.

Table 3 Descriptive statistics of seven reference gene candidates were calculated using the BestKeeper technique and the RefFinder tool.
Full size table

ΔCt method

The approach proposed by Silver et al.28 is based on a technique similar to geNorm by Vandesompele et al.25, simplifying mathematical methodology. The idea is to calculate the ΔCt value for each pair of reference gene candidates. If the computed value remains constant, the two genes can be assumed to be stably expressed or co-regulated. Fluctuation in ΔCt values (higher SD) indicated that one of both genes is variably expressed. The genes with the least SD values are ranked as the best normalization control (most stably expressed), which in our study were RPLP0 followed by HPRT1 in the kidney and GAPDH followed by RPLP0 in the liver tissue (Fig. 3).

Fig. 3
figure 3

Average standard deviation (SD) values reflecting gene stability were calculated by the ΔCt method using the RefFinder tool.

Full size image

Comprehensive ranking

Considering the results of the four methods described above, it would be challenging to judge which approach to follow and ultimately choose the best reference gene(s) (Table 4).

Table 4 Ranking orders of gene candidates for the normalization control were obtained using geNorm, NormFinder, BestKeeper, and ΔCt methods (rank 1 – most stable gene; rank 7 – least stable gene). Genes marked in bold were selected as reference genes (RPLP0 & HPRT1 in the kidney and RPLP0 & GAPDH in the liver).
Full size table

RefFinder tool24 provides a comprehensive ranking, which supports choosing the most appropriate normalization control for the expression study. In our set of candidates, for the kidney tissue, the genes were ranked: RPLP0 (most stably expressed; ranking value = 1.565), HPRT1 (2.000), GAPDH (2.449), ACTB (4.120), 18S rRNA (4.304), B2M (5.439), and PPIA (most variable; ranking value = 5.733). For the liver tissue, the most stably expressed gene was selected GAPDH (ranking value = 1.189), followed by ACTB and RPLP0 (both with a ranking value of 2.783), HPRT1 (3.722), 18S rRNA (4.304), B2M (4.787), and PPIA (6.000). The final ranking of reference genes for both kidney and liver tissue is visualized on the recapitulative Fig. 4.

Fig. 4
figure 4

The final ranking of reference genes provided by the RefFinder tool and the pie charts corresponding to four tests (geNorm, NormFinder, BestKeeper, and ΔCt). In pie charts, the bigger the slice, the less stable the gene; exploded slices represent the two genes, which a particular test recognized as most stable. In the comprehensive ranking, the bigger the font, the more stable the gene.

Full size image

Discussion

The reproducibility of qPCR-based relative gene expression depends on reference genes with stability across individuals, tissue types, morbidity conditions, and other relevant experimental parameters. Failure to choose a suitable control for normalizing quantitative data will result in a biased gene expression profile and low precision, resulting in only large changes demonstrating statistical significance. Therefore, it is essential to determine whether potential RGs are suitable for specific experimental conditions. Unfortunately, there is still a misconception that most HKGs show stable expression in all sample types and experimental setups. However, the literature confirms that most genes do not exhibit characteristics of such stability. The most commonly used HKG expression varies across tissue types and morbidity conditions29,30,31. The novelty and the main advantage of our research is the expression stability examination of commonly used reference genes in the liver and kidney tissue samples from patients with obesity. We studied five housekeeping genes preselected to the 96-well kit of human RT2 Profiler PCR Arrays (Qiagen, CA, USA) (ACTB, B2M, RPLP0, HPRT1, GAPDH) and two genes selected based on literature related to kidney tissue (18S rRNA and PPIA)32. Before the qPCR analysis, we confirmed the absence of protein and adequate RNA quality by assessing the purity and quantity of total isolated RNA, as protein contamination can inhibit cDNA synthesis and qPCR reaction and lead to biased Ct values33. The post-qPCR melting curve analysis is a well-established and reliable method for determining non-specific amplification and low amplification efficiency. We used this method to confirm primer specificity, verifying a single peak in each sample.

Moreover, the RT2 Profiler PCR Arrays contain reverse-transcription control (RTC). This assay tests the efficiency of the reverse-transcription reaction performed with the RT2 First Strand Kit by detecting the template synthesized from the kit’s built-in external RNA control. There was no apparent inhibition of the reverse transcription reaction in any of the samples tested.

The candidate reference genes were selected for the analysis based on the difference in their physiological functions: cytoskeleton (ACTB)34, host immune surveillance (B2M)35, structural constituent of ribosome (RPLP0)36, metabolic salvaging of purines (HPRT1)37,38, glycolysis (GAPDH)39, protein synthesis (18S rRNA)40, protein folding (PPIA)41. As required by the MIQE (Minimum Information for Publication of Quantitative Real-Time PCR Experiments) guidelines, utilized reference genes must play a distinct role in cell metabolism, indicating that they are not co-regulated and may be used in conjunction to normalize gene expression16.

NormFinder, geNorm, and comparative ΔCt algorithms correlated to a high degree, as reflected in similar gene stability rankings. The BestKeeper algorithm, on the other hand, differed in its assessment from the previous procedures. It was initially designed to evaluate the overall suitability of a housekeeping gene for RT-qPCR in a subsequent two-step evaluation based on the average standard deviation of Ct and correlation analysis rather than to compare possible housekeeping genes. The other algorithms, like geNorm or the ΔCt method, implement pairwise comparisons of housekeeping genes with linear quantities or use linear quantity models, as seen in the NormFinder26. However, using single algorithms to distinguish between available RGs is a source of variability in RG stability and data analysis among different studies. Therefore, to determine the optimal RG in each cell type, tissue, model, and species, using a tool like RefFinder that integrates these four algorithms is optimal. The superiority of the RefFinder application lies in its ability to provide an overall final ranking based on individual orders from each algorithm, assigning the appropriate weight to an individual gene and calculating the geometric mean of their weights.

Many studies have been carried out concerning validating housekeeping genes in many different tissues and cell types. However, it has been challenging to find information on appropriate housekeeping genes for use in the human liver and kidney tissue samples of individuals with obesity. By analyzing the expression stability of seven potentially suitable housekeeping genes with four different mathematical algorithms, we identified RPLP0 and HPRT1 in human kidney tissue and RPLP0 and GAPDH in the liver to be reliable sets of housekeeping genes for the normalization of target gene expression in qPCR analysis in obese individuals. RPLP0, the acidic ribosomal phosphoprotein P0, is involved in protein synthesis; the 36B4 gene encodes a component of the 60S subunit of the ribosome. RPLP0 is part of a pentameric complex that forms a stem-like structure protruding from the 60S subunit into the cytoplasm42. RPLP0, PPIA, and YWHAZ were the top three most stable reference genes in qPCR experiments on adipose, hepatic tissues, and muscles of mice in diet-induced obesity43. Xu L. et al. presented that HPRT1 and GAPDH have stable expression during the development of hepatic steatosis and are ideal for gene expression normalization in the research of hepatic steatosis in obese mice44. Interestingly, Mehta et al.45 recommended HPRT1 and GAPDH as the best candidates for qPCR analysis of human visceral adipose samples. Considering hepatic steatosis is caused by excessive triglycerides (TG) accumulation and obesity involves abnormal TG accumulation in adipocytes, it is possible that the reference genes HPRT1 and GAPDH are not engaged in TG accumulation and lipid metabolism. Lu B. et al. demonstrated that the expression of 13 candidate reference genes commonly used in rat obesity research was differentially affected by dietary and feeding conditions and tissue. They identified that the most suitable endogenous controls in the hypothalamus are B2M and RPLP0, and in the jejunum, RPLP2 and RPLP046.

The parallel determination of two or more housekeeping genes in a given sample population or experimental condition is crucial to allow for the detection of any systematic bias. Normalization with multiple reference genes is becoming more prevalent, but studies that apply this normalization approach are limited. Geometric mean possible outlying values and abundance differences between the different genes. It is suggested that at least three stable control genes should be included in the normalization of expression samples16. However, due to cost and time constraints, analysis of multiple reference genes may not always be feasible. Thus, it should be noted that several studies have shown that in qPCR analyses, under certain conditions, the use of two carefully selected reference genes is highly sufficient44,45,46.

Interestingly, there are tissues or conditions in which the lack of stable expression of the most common reference genes has been proven, e.g., embryonic development models47,48 or neuronal injury models49,50. In these cases with dynamically transcribed endogenous genes, exogenous mRNA spike-in transcripts have proven to be reliable references for qPCR analysis. This spike-in control can be used for sample-to-sample normalization and reverse transcriptase efficiency analysis. Such transcripts simultaneously undergo reverse transcription and amplification with the target transcript. Johnston et al. have used exogenous luciferase mRNA to normalize qPCR results in examinations of damage to the nervous system51. It is worth mentioning that this method is also used to assess miRNA relative expression levels. Since there is no clear consensus in the research community about which endogenous RNA should be used as a reference gene for miRNA expression profiling in plasma or serum, many researchers choose to add a synthetic miRNA as a spike-in control, a common choice is Caenorhabditis elegans-miR-39 (Cel-miR-39)52.

The development of obesity is a complex event, and subtle changes in gene expression during the development of obesity and its comorbidities can impact a cascade of signaling pathways. Hopefully, our study will raise awareness among obesity researchers of the crucial need for reference gene validation and that the reference genes identified here can be a resource for future obesity research.

It should be noted that the expression analysis was performed on post-mortem tissue, so natural degradation of RNA is possible, which could be a limitation of this study. However, due diligence was applied during tissue collection, storage, and RNA isolation.

Materials and methods

Human tissue specimens

All study aspects were conducted in accordance with the ethical principles of the Good Clinical Practice Guidelines and the Declaration of Helsinki. Forty-two necropsy specimens of liver and kidney were obtained from the National Disease Research Interchange (NDRI) (Philadelphia, PA). The protocol was granted expedited approval by the Rutgers Biomedical and Health Sciences (Pro2019001020), and it was determined that subject consent was not required since specimens were acquired from the NDRI tissue bank and were unidentifiable. NDRI requires all tissue source sites in the NDRI network to obtain informed consent in writing from any tissue donor (or the next-of-kin) for using that tissue for research. Any tissue with signs of disease (i.e., malignancy, cirrhosis) would be rejected. Tissue specimens were snap-frozen immediately after collection, shipped on dry ice, and stored at − 80 °C until analysis. Each sample was characterized in terms of the donor’s age, race, sex, body weight (kg), BMI (kg/m2), medical history, concomitant diseases, medication list, surgical history, date and time of death, cause of death; date and time of autopsy (including post-mortem interval [PMI]). Donors of tissue specimens were divided into categories based on their BMI value. A BMI of 18.5–25 kg/m2 was interpreted as healthy weight, while overweight individuals were characterized by BMI ≥ 25 kg/m2. Before the RNA extraction assay, a single tissue block (from a single patient) was removed from − 80 °C, placed on a tray with liquid nitrogen, and crushed into pieces. Several small pieces were immediately collected into a weigh boat on dry ice. The tissue specimens were on dry ice during this procedure and never allowed to thaw entirely.

RNA extraction and cDNA synthesis

Total RNA was extracted from the liver and kidney samples using the AllPrep DNA/RNA/Protein Mini Kit (Qiagen, CA, USA) and purified using the RNeasy Mini Kit (Qiagen, CA, USA). Both procedures were performed according to the manufacturer’s protocols. Both tissues were homogenized (approximately 30–35 mg per sample) in a dedicated buffer using TissueLyser (Qiagen, CA, USA). RNA quantity and quality were assessed using a NanoDrop 2000 (Thermo Scientific, MA, USA). The yield was determined in duplicate at an absorbance of 260 nm. All samples showed 260/280 ratios exceeding 1.8, excluding protein presence. The cDNA (0.5 µg) was synthesized (in a reverse transcribed polymerase chain reaction (RT-PCR)) using a commercial RT2 First Strand Kit (Qiagen, CA, USA) according to the manufacturer’s recommendations. The cDNA samples were stored at − 20 °C until further analysis.

Quantitative PCR

Real-time quantitative PCR was performed for the analysis of 5 housekeeping genes using 96-well plates RT2 Profiler PCR Array Systems, a SYBR-Green-based method (Qiagen, CA, USA) in QuantStudio 7 Pro Real-Time PCR instrument (Thermo Scientific, MA, USA). Primers for two additional housekeeping genes (18S rRNA and PPIA) were selected from ‘off the shelf’ (Qiagen, CA, USA). qPCR reactions were initiated by activating Hot Start DNA Taq Polymerase at 95 °C for 10 min, followed by 40 amplification cycles. The thermal cycling profile consisted of denaturation at 95 °C for 15 s and annealing and extension at 60 °C for 1 min with subsequent fluorescence data acquisition. A melting curve was generated to discriminate between specific and non-specific amplification products (in all cases, the ramp time was 1 °C/s). The cut-off Ct was set at 35 cycles for all analyses.

Stability of reference gene

The gene expression stabilities of all candidate RGs (ACTB, B2M, RPLP0, HPRT1, GAPDH, 18S rRNA, and PPIA) in liver and kidney samples were analyzed with comparative ΔCt method26, geNorm25, NormFinder26, and BestKeeper27 programs based on untransformed mean Ct values from three replicates. For the consensus ranking of RG candidates, the geometric mean of ranks from these analyses was calculated using an online tool, RefFinder24.