Human housekeeping genes are compact

Eli Eisenberg and Erez Y. Levanon

Compugen Ltd., 72 Pinchas Rosen Street, Tel Aviv 69512, Israel

Abstract
arXiv:q-bio/0309020v1 [q-bio.GN] 30 Sep 2003

We identify a set of 575 human genes that are expressed in all conditions tested in a publicly
available database of microarray results. Based on this common occurrence, the set is expected
to be rich in “housekeeping” genes, showing constitutive expression in all tissues. We compare
selected aspects of their genomic structure with a set of background genes. We find that the
introns, untranslated regions and coding sequences of the housekeeping genes are shorter, indicating
a selection for compactness in these genes.

1
 The amazing diversity of the human body stems from the different expression patterns of
genes in different tissues. Although most genes show constitutive expression in only a subset
of tissues, some gene products are required for the maintenance of the basal cellular function
and are constitutively found in all human cells. These genes are called housekeeping genes
(HK genes) [1]. HK genes can be used to calibrate measurements of gene expression [2].
They might also help to define the minimal gene complement needed for a human cell [1].
Several attempts have been made recently to define the complete set of HK genes [3, 4].
Microarrays are often used to identify sets of genes that are expressed either ubiquitously
or in specific tissues or conditions. However, the technique is technically demanding and
prone to artifacts, so independent evidence is often required to confirm the results. In
principle, identifying the set of HK genes using microarray data is straightforward; one
need only look for genes that are expressed in all tissues and all experimental conditions.
Employing such an approach has so far resulted in two lists of HK genes [3, 4]. However,
problems in probe design, measurement noise and other artifacts introduce inevitable errors
in such lists. Because a northern blot experiment for each gene in each tissue is impractical,
an independent test is needed to validate any list of HK genes. Here, we report a validation
test that uses a recently discovered property of highly expressed genes.
The transcription process is both slow and costly; it takes 50 milliseconds [5, 6] and
two ATP molecules [7] approximately to transcribe a nucleotide. This might be expected
to provide selective pressure to make genes as short as functionally possible. The more
copies of a gene required for the organism, the stronger this pressure should be. The first
demonstration of this principle [8] showed that genes with a large number of expressed
sequence tags (ESTs) in public libraries (and hence most mRNAs) have a significantly
shorter average intron length than those with fewer ESTs.
Here, an implication of this principle is used to validate a set of HK genes. The HK
genes, which are transcribed in all somatic cells and under all circumstances, are by nature
highly expressed, and therefore should be selected to have shorter introns. We used a
recently published database of microarray experiments [9] to identify a set of HK genes. As
a further validation step, we checked the Gene Ontology (GO) annotation of these genes.
We compared the structure of the HK genes with all other genes, and not only the introns,
but all parts of the HK genes were found to be, on average, shorter than other genes. In
particular, the untranslated regions and the translated proteins are all shorter in the HK

2
 assumed
housekeeping

Nuber of genes
1000 genes
background

500

0
0 5 10 15 20 25 30 35 40 45
Number of tissues expressed in

FIG. 1: The distribution of 7500 RefSeq genes represented on the microarray as a function of the
number of tissues they express in. Each bin gives the number of genes expressed in M out of 47
different tissues. The M=47 bin corresponds to the housekeeping genes, expressed in all tissues.

genes.

I. ASSIGNMENT OF HOUSEKEEPING GENES

A recently published database provides microarray expression data for Affymetrix U95A
chip, containing 12,600 probes, and hybridized to 101 different samples [9] from 47 different
human tissues and cell lines. These samples are mainly from the normal human physio-
logical state, and therefore this dataset provides a description of the normal mammalian
transcriptome.
We calculated the distribution of the number of different tissues in which a gene is ex-
pressed. Discarding probes for which the associated gene was not represented in the RefSeq
database [10], and unifying all probes measuring the same gene (ignoring the potential
differences among splice variants) yielded probes representing 7500 human genes. The ex-
periments measuring replicates of the same biological condition were averaged to reduce the
measurement noise, resulting in 47 data points per probe. We considered that a probe was
expressed in a certain condition if its average reading was above a certain cutoff value. The
results were not sensitive to the exact cut-off value, and we chose 200 standard Affymetrix
averagedifference units, considered to be a conservative cut-off value for determining gene
presence [9]. This is also the trimmed average expression level in each tissue in accordance
with the standard Affymetrix normalization procedure [11, 12]. Thus, our HK genes are
expressed in all tissues at an above-average level.
A histogram (Fig. 1) of the number of genes expressed in exactly M of the 47 tissues

3
 0 20000 40000 60000
Total introns length

FIG. 2: A Histogram of the total length of introns. Green bars, HK genes; blue bars, non-HK
genes.

shows a clear tendency for frequency to decrease as M increases. However, a substantial

number of genes (575), belong to the class of genes that are expressed in all tissues. Because
their number is far greater than expected based on the general trend described above, we
assumed this class to be rich in HK genes, and considered it to be the set of HK genes.
It is noteworthy that the genes in our HK list tend to have an average expression sig-
nificantly higher than other genes; the geometric mean expression of our HK genes is 1200
in Affymetrix average difference units, whereas that of other genes is 150. The difference
cannot be accounted for by the cutoff used to define the HK genes, and is not a result of a
bias due to inclusion of genes expressed in a few tissues only (data not shown).
Two additional tests were conducted to validate this set. First, a study of the GO
annotation [13] of these genes revealed the set is rich in metabolic proteins (24%) and
RNA-interacting proteins (19%, mostly ribosomal proteins). Second, we compiled a list
of 18 well-established HK genes commonly used for quantitative PCR calibration [14, 15],
and checked our list against it. We found 13 of the 18 genes in our list, and the other
five were not represented on the microarray (see Table in Supplementary Information at
http://www.compugen.co.il/supp info/Housekeeping genes.html).

II. LENGTH ANALYSIS OF HK GENES

Table 1 compares the lengths of various parts of the HK genes and the background genes.
The alignment data was taken from the UCSC genome browser (http://genome.ucsc.edu)
[16]. We excluded 322 genes that do not have a unique alignment, as well as 1242 genes that
were not expressed in any tissue (to avoid potential problems because of defective probes).

4
This left 532 HK genes and 5404 non-HK genes. The histograms in Fig. 2-4 compare HK
genes with the other genes by total intron length, 5’ UTR length and coding sequence length.
Remarkably, there was a statistically significant difference between HK and non-HK genes
in all aspects of gene structure. Average intron length is shorter for the HK genes than for
the background genes (2573 bp versus 5025 bp, respectively); total gene length is shorter
(21,050 bp versus 53,418 bp); average exon length is shorter (212 bp versus 240 bp); average
lengths of both 3’ and 5’ untranslated regions (UTRs) are shorter (5’: 135 bp versus 173

TABLE I: Human housekeeping genes are compact. Comparison of structure of housekeeping

(HK) genes versus non-HK genes. For each case the first line gives the average value, s.e.m, and
the second line gives the median. For the average intron and exon lengths, all introns and exons
belonging to the relevant set were included; the number appears in parentheses. The P-value was
calculated using the Mann-Whitney test. UTR, untranslated region.

HK genes (n=532) non-HK (n=5404) P-value

Average intron length 2573 ± 145(n=4353) 5025 ± 71(n=57447) 4 × 10−130
672 1365
Total intron length 21050 ± 1781 53418 ± 1425 7 × 10−28
9293 20804
Average exon length 212 ± 5(n=4885) 240 ± 2(n=62851) 9 × 10−5
672 1365
5’ UTR length 135 ± 8 173 ± 3 4 × 10−7
79 106
3’ UTR length 599 ± 30 846 ± 13 3 × 10−13
333 552
Coding sequence length 1211 ± 44 1770 ± 26 3 × 10−26
928 1322
Number of introns 8.2 ± 0.3 10.6 ± 0.2 6 × 10−7
6 8
Intron bps per coding bp 20 ± 2 31.8 ± 0.8 2 × 10−11
9.9 15.6

5
 0 250 500
5` UTR length

FIG. 3: A Histogram of the length of the 5’ untranslated regions (UTR). Green bars, HK genes;
blue bars, non-HK genes.

bp; 3’: 599 bp versus 846 bp); and, most notably, the translated proteins are shorter as well
(403 amino acids versus 590 amino acids). Accordingly, the number of introns bp per unit of
coding sequence length is lower for the HK genes (20 versus 32). We studied the structure
of each gene as a function of the number of tissues it is expressed in and verified that the
results are not due to bias of the non-HK genes by tissue-specific genes (data not shown).
The pronounced statistical characteristics of the HK gene set further supports their as-
signment as a unique set. Our findings confirm and extend previous research, showing that
the introns of highly expressed genes are shorter [5]. As mentioned above, the HK genes
expression levels are high, and the fact that they have to be expressed in all cells at all
times makes them even more costly to transcribe. Previously [8], the high abundance of a
certain gene in EST libraries was an indication the gene was highly expressed in the hu-
man body. It was pointed out [8], however, that this method is prone to bias due to the
inclusion of normalized and tumor libraries and overrepresentation of certain tissues. Our
approach overcomes this difficulty and confirms the previous result. Moreover, we find here
that UTRs and even the encoded proteins are shorter for the HK genes. The magnitude
of the difference is greater for the introns than for the exons and proteins (Table 1), which
makes sense because the coding sequences and the UTRs are less susceptible to change.
It should be mentioned that intronless genes were included in our analysis after verifying
that their inclusion or exclusion had no effect on the results. It also must be noted that the
UTRs are not always fully sequenced, and thus their actual lengths might be longer. This
bias was found to have no effect on the length of the coding sequences, and in any case the
effect would be the same for both HK and non-HK genes.

6
 0 1000 2000 3000 4000
Total cds length

FIG. 4: A Histogram of the length of the coding region. Green bars, HK genes; blue bars, non-HK
genes.

It has been noted that codon usage bias in nonmammalian organisms is correlated with
the expression level and with the gene length [17, 18, 19]. These results led to the conjecture
of selective pressure on highly expressed genes resulting in shorter proteins [19]. However, no
evidence for this selection was found [18], possibly because of a lack of high quality databases
for these organisms. Recent works have suggested that there is no selection for codon usage
bias in humans [20], and thus our results demonstrate that the expression-length correlation
is not related to the expression-codon bias correlation.
It could be argued that selection towards shorter genes should have eliminated the introns
in highly expressed genes. However, it is known that introns do have important roles, such
as splicing regulation. Therefore, there is a balance between the advantageous contribution
of the introns and the selective pressure for shortening.
Finally, when we compared our results with two (largely overlapping) published sets of
HK genes, we found that roughly half of the genes in the intersection of those sets were
present in our set. We used the genomic structure to test the remaining genes, and found
a statistically significant difference between them and our HK gene set. The differences
between our results and those of earlier studies [3, 4] could be due to the fact that the
database we used was based on more advanced chip technology and included many more
different tissues, giving it more discriminative power to identify HK genes.
In conclusion, we have identified a set of HK genes. The set is publicly available at
http://www.compugen.co.il/supp info/Housekeeping genes.html and can be used for cali-
bration of microarrays, toxicity evaluation and quantitative PCR experiments. Furthermore,
we show that HK genes have shorter introns, UTRs and coding sequences, attesting to the

7
strong selection for compactness in these genes.

Acknowledgments

We thank Andrew Su for helpful discussion and for providing us with the RefSeq mapping.
Gady Cojocaru and Rotem Sorek are acknowledged for comments on the manuscript and
insightful discussion.

[1] Butte, A.J. et al. (2001) Physiol. Genomics 7, 95-96.

[2] Gibson, U.E. et al. (1996) Genome Res. 6, 995-1001.
[3] Warrington, J.A. et al. (2000) Physiol. Genomics 2, 143-147.
[4] Hsiao, L.L. et al. (2001) Physiol. Genomics 7, 97-104.
[5] Ucker,D.S. and Yamamoto, K.R. (1984) J. Biol. Chem. 259, 7416-7420.
[6] Izban, M.G. and Luse, D.S. (1992) J. Biol. Chem. 267, 13647-13655.
[7] Lehninger, A.L. et al. (1982) Biochemistry, 615-644.
[8] Castillo-Davis, C.I. et al. (2002) Nat. Genet. 31, 415-418.
[9] Su, A.I. et al. (2002) Proc. Natl. Acad. Sci. U.S.A. 99, 4465-4470.
[10] Pruitt, K.D. et al. (2000) Trends Genet. 16, 44-47.
[11] Lockhart, D.J. et al. (1996) Nat. Biotechnol. 14, 1675-1680.
[12] Wodicka, L. et al. (1997) Nat. Biotechnol. 15, 1359-1367.
[13] Gene Ontlology Consortium, (2001) Genome Res. 11, 1425-1433.
[14] Hamalainen, H.K. et al. (2001) Anal. Biochem. 299, 63-70.
[15] Lee, P.D. (2002) Genome Res. 12, 292-297.
[16] Karolchik, D. et al. (2003) Nucleic Acids Res. 31, 51-54.
[17] Akashi, H. (2001) Curr. Opin. Genet. Dev. 11, 660-666.
[18] Duret, L. and Mouchiroud, D. (1999) Proc. Natl. Acad. Sci. U.S.A. 96, 4482-4487.
[19] Moriyama, E.N. and Powell, J.R. (1998) Nucleic Acids Res. 26, 3188-3193.
[20] Urrutia, A.O. and Hurst, L.D. (2001) Genetics 159, 1191-1199.