Novel ‘housekeeping’ genes and an unusually heterogeneous distribution of transporter expression profiles in human tissues and cell lines, assessed using the Gini coefficient

Steve O’Hagan, Marina Wright Muelas, Philip J. Day, Emma Lundberg, Douglas B. Kell

Abstract

We analyse two comprehensive transcriptome datasets from human tissues and human-derived cell lines in terms of the expression profiles of the SLC and ABC families of membrane transporters. The Gini index (coefficient) characterises inequalities of distributions, and is used in a novel way to describe the distribution of the expression of each transporter among the different tissues and cell lines. In many cases, transporters exhibit extremely high Gini coefficients, even when their supposed substrates might be expected to be available to all tissues, indicating a much higher degree of specialisation than is usually assumed. This is consistent with divergent evolution from a more restricted set of ancestors. Similar trends hold true for the expression profiles of transporters in different cell lines, suggesting that cell lines exhibit largely similar transport behaviour to that of tissues. By contrast, the Gini coefficients for ABC transporters tend to be larger in cell lines than in tissues, implying that some kind of a selection process has taken place. In particular, with some exceptions such as olfactory receptors and genes involved in keratin production, transporter genes are significantly more heterogeneously expressed than are most non-transporter genes. The Gini index also allows us to determine those transcripts with the most stable expression; these often differ significantly from the ‘housekeeping’ genes commonly used for normalisation in transcriptomics and qPCR studies. The lowest four in tissues are FAM32A, ABCB7, MRPL21 and PCBP1, while the lowest three in cell lines are SF3B2, NXF1 and RBM45. PCBP1 is both reasonably highly expressed and has a low Gini coefficient in both tissues and cell lines, and is an excellent novel housekeeping gene. Overall, our analyses provide novel opportunities for the normalisation of genome-wide expression profiling data.

Introduction

In human biology, genes are often broadly categorized as either "housekeeping" genes—those expressed ubiquitously to maintain fundamental cellular functions—or "specialized" genes, which are restricted to specific tissues. Membrane transporters, comprising the Solute Carrier (SLC) and ATP-binding cassette (ABC) families, are the primary gatekeepers of the cell, regulating the influx of nutrients and the efflux of metabolic waste and xenobiotics. Despite their importance, the degree of specialization across these families has remained difficult to quantify using traditional parametric methods.

This study utilizes the Gini coefficient, a statistical measure borrowed from economics, to provide a non-parametric assessment of expression inequality across the human transcriptome. By applying this metric to massive RNA-seq datasets, we reveal that membrane transporters are far more specialized than previously assumed and identify a new suite of highly stable "housekeeping" genes that offer superior performance for transcriptomic normalization.

Methodology

The Gini coefficient (or Gini index) typically measures income inequality within a population. In this biological context, a Gini of 0 represents "perfect equality" (uniform expression across all tissues), while a Gini of 1 represents "extreme inequality" (expression in only one tissue).

Using the KNIME analytics platform and the ineq R package, we calculated Gini indices for thousands of genes across two primary datasets from the Human Protein Atlas:

  1. The Tissue Atlas: Transcriptome data for 409 SLCs and 48 ABCs across 59 human tissue types.
  2. The Cell Atlas: Expression profiles for the same transporters across 56 human-derived cell lines.

Findings were further validated using Immunohistochemistry (IHC) to ensure that mRNA expression levels correlated with actual protein presence.

The Heterogeneity of Transporter Expression

Our analysis reveals that membrane transporters are among the most heterogeneously expressed genes in the human genome. The median Gini coefficient for SLC transporters was found to be 0.587, a value higher than the income inequality of any nation on Earth. This high degree of specialization is surprising; many transporters have Gini coefficients exceeding 0.9, indicating they are nearly tissue-specific.

This creates a "Specialization Paradox": many substrates for these transporters, such as glucose or basic amino acids, are required by every cell in the body. The extreme inequality in transporter distribution suggests a much higher level of evolutionary "tuning" than standard textbooks imply, likely reflecting divergent evolution from a restricted set of ancestors to meet specific physiological demands.

Comparative Analysis: Tissues vs. Cell Lines

Generally, human-derived cell lines maintain transport profiles that mirror their tissues of origin. However, a notable divergence was observed in ABC transporters (efflux pumps), which exhibited higher Gini indices in cell lines than in tissues. This suggests that laboratory culture conditions may exert a selection pressure, favoring specific transport behaviors or resulting in the loss of certain efflux capabilities over time.

In terms of organ-specific expression, the kidney demonstrated the highest overall expression of transporters, consistent with its role in filtration and reabsorption. Conversely, the pancreas showed the lowest transporter expression, a finding that may explain the clinical difficulty of targeting drugs to pancreatic tissues.

Redefining Housekeeping Genes

A critical application of the Gini index is the identification of genes with the most stable expression across all conditions. Traditional housekeeping genes used for qPCR and transcriptomic normalization, such as GAPDH or ACTB, often possess Gini indices between 0.2 and 0.3, indicating significant variability.

Our analysis identified several genes with much lower Gini indices (0.11–0.14). The most promising candidate is PCBP1. It maintains a very low Gini index in both tissues (0.139) and cell lines (0.135) and is expressed at high enough levels (median >100 TPM) to be easily detected. Other stable candidates include FAM32A, ABCB7, and MRPL21 in tissues, and SF3B2, NXF1, and RBM45 in cell lines.

Conclusion

The application of the Gini coefficient to transcriptomic data provides a robust framework for quantifying gene specialization. The discovery that transporters are unusually heterogeneous highlights a gap in our understanding of substrate availability and tissue-specific biochemistry. Furthermore, the identification of PCBP1 as a "super-housekeeping" gene provides a novel opportunity for more accurate normalization in genome-wide expression profiling. Future research utilizing single-cell RNA-seq will be essential to determine if these patterns of inequality persist at the sub-organ level.

Summary updated

Abstract

Abstract We analyse two comprehensive transcriptome datasets from human tissues and human-derived cell lines in terms of the expression profiles of the SLC and ABC families of membrane transporters. The Gini index (coefficient) characterises inequalities of distributions, and is used in a novel way to describe the distribution of the expression of each transporter among the different tissues and cell lines. In many cases, transporters exhibit extremely high Gini coefficients, even when their supposed substrates might be expected to be available to all tissues, indicating a much higher degree of specialisation than is usually assumed. This is consistent with divergent evolution from a more restricted set of ancestors. Similar trends hold true for the expression profiles of transporters in different cell lines, suggesting that cell lines exhibit largely similar transport behaviour to that of tissues. By contrast, the Gini coefficients for ABC transporters tend to be larger in cell lines than in tissues, implying that some kind of a selection process has taken place. In particular, with some exceptions such as olfactory receptors and genes involved in keratin production, transporter genes are significantly more heterogeneously expressed than are most non-transporter genes. The Gini index also allows us to determine those transcripts with the most stable expression; these often differ significantly from the ‘housekeeping’ genes commonly used for normalisation in transcriptomics and qPCR studies. The lowest four in tissues are FAM32A, ABCB7, MRPL21 and PCBP1, while the lowest three in cell lines are SF3B2, NXF1 and RBM45. PCBP1 is both reasonably highly expressed and has a low Gini coefficient in both tissues and cell lines, and is an excellent novel housekeeping gene. Overall, our analyses provide novel opportunities for the normalisation of genome-wide expression profiling data.
BibTeX
@misc{ohagan2018novel,
  author = {Steve O’Hagan and Marina Wright Muelas and Philip J. Day and Emma Lundberg and Douglas B. Kell},
  title = {{Novel ‘housekeeping’ genes and an unusually heterogeneous distribution of transporter expression profiles in human tissues and cell lines, assessed using the Gini coefficient}},
  year = {2018},
  doi = {10.2139/ssrn.3155744},
  howpublished = {SSRN},
}
Figure