**Background:** Housekeeping genes (HKGs) are assumed to have stable expression across cell types and are widely used as internal controls in gene expression studies. However, their expression can vary depending on experimental conditions, tissue type, and biological variables—including sex, which has historically been overlooked. In adipose tissue (AT) research, commonly used HKGs such as ACTB, GAPDH, and 18S have not been systematically evaluated for sex bias. This study aimed to assess the expression stability of six classical HKGs (GAPDH, HPRT1, PPIA, UBC, 18S, and RPL19) in human and mouse AT, and to identify novel sex-unbiased HKGs (suHKGs) through a comprehensive meta-analysis of transcriptomic data.
**Methods:** The authors conducted a systematic review of GEO (Gene Expression Omnibus) for transcriptomic studies of adipose tissue (expression profiling by array) in Homo sapiens and Mus musculus, covering 2000–2019. They selected the four most used microarray platforms for human (GPL570, GPL6244, GPL10558, GPL6947) and five for mouse (GPL1261, GPL6246, GPL6887, GPL6885, GPL16570). After applying inclusion criteria (≥10 AT samples, standardized data, no duplicates), 49 human studies (2724 samples) and 43 mouse studies (1072 samples) were retained. Only 51% of human studies and 51% of mouse studies reported sample sex; among mouse studies, 19 used only males, 2 only females, and 1 both sexes. Human samples with known sex included 681 male and 875 female samples; mouse samples included 559 known male and only 34 known female samples, precluding sex-stratified mouse analysis. Three statistical stability indicators were calculated per gene per study: coefficient of variation (CV), IQR/median, and MAD/median. Gene variability scores per platform were ranked, and the Rank Product (RP) method was used to integrate rankings across platforms, producing platform-independent stability rankings. For human suHKG selection, genes with low expression (TPM < 20 in GTEx) were filtered out, and the top 10% most stable genes in male and female metaRankings were intersected. Experimental validation was performed by qPCR on human visceral and subcutaneous AT from lean and obese male and female individuals, and on mouse AT from wild-type and Irs2−/− (insulin resistance model) C57BL/6 mice.
**Key Results:** The six classical HKGs showed marked differences in stability between sexes in human AT. In human females, HKGs displayed greater instability overall. Specifically, PPIA, UBC, and RPL19 showed high stability in both sexes; GAPDH was stable in males but unstable in females; HPRT1 and 18S exhibited low stability in both sexes. In mouse samples (predominantly male), 18S appeared stable, but experimental validation revealed significant sex bias. After filtering for expression (TPM ≥ 20) and selecting the top 10% most stable genes in each sex, 195 candidate suHKGs common to both sexes were identified. These included classical HKGs PPIA (rank: male 873, female 1589; TPM: male 233.6, female 236.1) and RPL19 (rank: male 1129.67, female 137.33; TPM: male 1708, female 1707), as well as novel candidates RPS8 (rank: male 194.67, female 178.33; TPM: male 957.9, female 944.5), RPS18 (rank: male 119.33, female 296.33; TPM: male 3168, female 3180), UBB (rank: male 228, female 252.29; TPM: male 254.1, female 249.8), and UBC (rank: male 267.33, female 706.33; TPM: male 447.7, female 396.9). Experimental validation confirmed that PPIA and RPL19 had low CV in Cp values across male and female samples, while 18S showed significant sex differences and high CV. Using 18S as a reference gene altered the apparent relative expression of target genes (IRS1, LEPR, PPARγ) in human AT, whereas PPIA and RPL19 produced consistent results. In mouse AT, Ppia and Rpl19 also performed well as suHKGs, while 18s introduced sex-dependent bias. Deconvolution analysis found no consistent differences in cell-type composition between male and female samples across datasets, suggesting the observed instability is not driven by cellular heterogeneity.
**Clinical Implications:** This study demonstrates that the choice of housekeeping gene can substantially affect gene expression results in adipose tissue research, particularly when sex is considered. Using an inappropriate HKG (e.g., 18S) can introduce systematic error and lead to incorrect biological or clinical conclusions. The authors recommend PPIA and RPL19 as validated sex-unbiased reference genes for human AT studies, and propose RPS8, RPS18, UBB, and UBC as additional candidates. For mouse AT, orthologs Ppia and Rpl19 are recommended. The findings highlight the critical need to include sex as a biological variable in study design and to validate HKG stability under specific experimental conditions. The open-access web tool (metafun-HKG) enables researchers to explore gene stability rankings and select appropriate suHKGs for their experimental designs. The authors note that nearly half of published studies fail to report sample sex, and mouse studies overwhelmingly use males, which may have biased previous HKG validation efforts.