Genome-wide analysis of mRNAs bound to the histone stem-loop binding protein

Abstract

The replication-dependent histone mRNAs are cell-cycle-regulated and expressed only during S phase. In contrast to all other eukaryotic mRNAs, the histone mRNAs end in a highly conserved 16-nucleotide stem-loop rather than a poly(A) tail. The stem-loop is necessary and sufficient for the post-transcriptional regulation of histone mRNA during the cell cycle. The histone mRNA 3' stem-loop is bound by the stem-loop binding protein (SLBP) that is involved in pre-mRNA processing, translation, and stability of histone mRNA. Immunoprecipitation (IP) of RNA-binding proteins (RBPs) followed by microarray analysis has been used to identify the targets of RNA-binding proteins. This method is sometimes referred to as RIP-Chip (RNA IP followed by microarray analysis). Here we introduce a variation on the RIP-Chip method that uses a recombinant RBP to identify mRNA targets in a pool of total RNA; we call this method recombinant, or rRIP-Chip. Using this method, we show that recombinant SLBP binds exclusively to all five classes of histone mRNA. We also analyze the messages bound to the endogenous SLBP on polyribosomes by immunoprecipitation. We use two different microarray platforms to identify enriched mRNAs. Both platforms demonstrate remarkable specificity and consistency of results. Our data suggest that the replication-dependent histone mRNAs are likely to be the sole target of SLBP.

FIGURE 1.

Strategy for identifying SLBP targets using the rRIP-Chip method. (A) Histone mRNAs are not polyadenylated but instead end in a conserved stem–loop. The message undergoes a single processing step in the nucleus that involves the SLBP and the U7 snRNP. The mature message ending in a conserved stem–loop structure is then transported to the cytoplasm where it is loaded onto polyribosomes. (B) Recombinant purified SLBP was used to identify mRNA targets. Total RNA is purified from either isolated polyribosomes or whole cells. Purified RBP is mixed with the purified RNA and RNA–protein complexes isolated using either an antibody against SLBP or an antibody against a tag in the recombinant protein. RNA is then purified from the bound and unbound fractions. Purified RNA is converted to either Cy3- or Cy5-labeled cDNA by reverse transcription primed with random hexamer primers and hybridized to whole genome microarrays.

FIGURE 2.

The replication-dependent histone mRNAs are bound by recombinant SLBP in vitro as determined by rRIP-Chip. We used purified recombinant, histidine-tagged SLBP and purified polyribosomal RNA to test the rRIP-Chip method described in the text and Figure 1. We performed four independent IP experiments to identify SLBP targets using an anti-SLBP antibody. Enriched mRNAs were identified by assigning each element on the microarray a percentile rank in each of the four experiments. These were subsequently used to calculate the median percentile rank for all four experiments. (A) The distribution of the median percentile ranks for the four αSLBP IPs are graphed. The distribution shows a small tail at the high percentile ranks (bin size = 0.005794) and shows a bimodal distribution. We have defined putative SLBP targets as those that fall to the right of this trough with a percentile rank above 98.83% (the distribution of these genes is colored red in the histogram inset). (B) The distribution of the median percentile ranks from the four mock IP experiments and the single no antibody IP are shown. There is not an obvious tail evident in the data. (C) The data for the genes selected as SLBP targets are displayed ordered by their percentile ranks. The log₂ of the Cy5/Cy3 ratio for each gene in each experiment is displayed with the data centered on its median value. Expression values above the median have been color-coded yellow and those below the median have been color-coded blue.

FIGURE 3.

The replication-dependent histone mRNAs are bound to the endogenous SLBP in vivo. Purified polyribosomes were incubated with anti-SLBP. We repeated the SLBP IPs independently five different times. In each case, a mock IP was performed in parallel without any SLBP antibody or anti-SLBP preincubated with the antigenic peptide, resulting in a total of 10 negative controls. For each IP, the precipitated mRNAs were assigned a percentile rank. The percentile rank was calculated based on the enrichment of each mRNA precipitated from polyribosomes, relative to the unprecipitated RNA from the polyribosomes. We subsequently used the five percentile rank values to calculate the median percentile rank for each precipitation. (A) Shown is the distribution of median percentile ranks in the anti-SLBP experiment. The distribution is skewed toward a single tail of the normal distribution, indicating enrichment of particular mRNAs (bin size = 0.007454). We have defined SLBP targets as those that fall to the right of the trough of the bimodal distribution. This cutoff, which corresponds to a percentile rank of 98.4% (and is colored red in the inset), identifies 36 genes (0.19% of the genes on the array) as significantly enriched. (B) The distribution of the median percentile ranks is shown from the 10 mock IP experiments from polyribosomes, which shows a normal distribution and illustrates that very few genes are enriched in the negative control experiments (bin size = 0.007326). (C) Genes identified as targets of the SLBP are displayed using Java Treeview, where expression values above the mean have been color-coded yellow and those below the mean have been color-coded blue. The genes are ordered by their median percentile rank in the IP of the endogenous SLBP. The 18 most highly enriched genes are histones. Of those genes selected, a subset of genes (n = 15) are not histone genes.

FIGURE 4.

Distribution of all replication-dependent histone genes found on the microarray. (A) Shown are histograms showing the distribution of log₂(Cy5/Cy3) for all genes on the microarray (blue) and for the replication-dependent histone mRNAs (green). In each case, the y-axis is the percentage of genes plotted at a given intensity value. The identity of each histone gene, along with accession numbers and the average log₂(Cy5/Cy3) is given in Supplementary Tables S1 and S2. In panel A, the average log₂ of the Cy5/Cy3 ratio distributions across all four αSLBP rRIP-Chip experiments is plotted (bin size = 0.1) as a percentage of all genes (blue; 33,125 genes) or as a percentage of only the replication-dependent histone genes (green; 33 genes). (B) The average log₂ (Cy5/Cy3) ratio across the five endogenous SLBP RIP-Chip experiments are plotted (bin size = 0.1) versus the percentage of all genes on the array (blue, N = 27,957) and the percentage of only the replication-dependent histone genes (green, N = 29). (C) Ten nanograms of a “synthetic histone mRNA pool” comprised of T7 transcribed histone coding regions (HIST1H1A, HIST1H2AB, HIST1H2BB, HIST1H3A, HIST1H4A) were labeled and hybridized to Stanford cDNA microarrays. The net Cy5 intensity for each replication-dependent histone gene on the array was extracted from the data and plotted along with the average percentile rank for that gene in both the RIP-Chip and rRIP-Chip experiments. A dashed line indicates the cutoff for an enriched gene. These data show that the histone genes identified as enriched in either the rRIP-Chip or the RIP-Chip experiment show significant intensities when hybridized to the positive control synthetic histone mRNA pool. In contrast, histone mRNAs that were not identified as enriched show hybridization signals approaching background. The exception is HIST3H2A, which shows high net Cy5 signal intensity, yet was not identified as enriched.

FIGURE 5.

Concordance between rRIP-Chip and mRNP complexes formed in vivo. (A) Genes identified as enriched in both the rRIP-Chip and RIP-Chip of the endogenous mRNP complexes are shown. Expression values above the mean are color-coded yellow and those below the mean are color-coded blue. The median percentile rank for each gene in the rRIP-Chip experiment and the endogenous RIP-Chip experiment are given, as is the mean of the two measurements. The overlap between the two data sets identifies histone genes as the primary class of mRNAs bound to the SLBP. (B) Ten nanograms of a synthetic histone mRNA pool were hybridized to cDNA microarrays to test for cross hybridization. Shown are the net Cy5 signal intensities for the three histone genes with the highest average percentile rank from panel A (HIST1H1C, HIST1H2AL, HIST1H2BC), the three nonhistone genes (HLA-DRA, BCHE, TCF15), and two genes not identified as enriched in either the rRIP-Chip or RIP-Chip experiments (ACTA2, TUBA3). The nonhistone genes show signal intensities above background (5–10 units) but lower than that of the replication-dependent histone mRNAs, when hybridized to the synthetic histone mRNA pool containing only the five different replication-dependent histone mRNA coding regions. This suggests their enrichment is likely a result of low-level cross hybridization.

FIGURE 6.

RIP-Chip analysis of the SLBP on Agilent oligonucleotide microarrays. We used Agilent 44,000 element oligonucleotide microarrays to analyze the RNAs bound to the endogenous SLBP and to compare the results to those obtained by RIP analysis of SLBP on the cDNA microarrays. The genes are ordered by mean percentile rank in the RIP-Chip experiment determined using the Agilent arrays. Only those genes above a percentile rank of 99.30% are shown. The results obtained on the Agilent platform indicate that histone genes are likely to be the primary target of the SLBP. It is notable that the nonhistone genes identified as significantly enriched in the rRIP-Chip and RIP-Chip treatments are not enriched in this analysis.

FIGURE 7.

Analysis of putative SLBP targets by RT-PCR. Shown are ethidium bromide stained gels analyzing HIST1H1A and putative false-positive genes. (A) The RT-PCR results confirming the IP of the HIST1H1A gene. As a negative control for nonspecific IP, we have also measured the expression of an unrelated cell-cycle-regulated gene that does not contain a histone stem–loop, the Forkhead Box M1 (FOXM1) gene. The FOXM1 mRNA is not precipitated in the IP experiment. (B) The expression levels of a subset of nonhistone, off-target genes identified as enriched in the endogenous IP experiment were tested by semiquantitative RT-PCR. We measured the expression level of the genes in HeLa total RNA and Stratagene Universal Human Reference RNA by semiquantitative RT-PCR. cDNA was generated by reverse transcription primed with random hexamer (pdN6) and the amount of cDNA for each gene measured by PCR. The left panel shows genes that are not expressed in HeLa cells. We find cyclin-dependent kinase-like 5 (CDKL5), Butyrycholinesterase (BCHE), Zinc finger protein 259 (ZNF259), KIAA0056 protein, and Spectrin, alpha, erythrocytic 1 (elliptocytosis 2) (SPTA1) are present in UHRR but not in HeLa cell total RNA, suggesting the genes are not expressed in HeLa cells. Glycerophosphodiester phosphodiesterase domain containing 2 (GDPD2), Lymphocyte antigen 9 (LY9), RAS and EF hand domain containing (RASEF), and Heat-responsive protein 12 (HRSP12) are shown in the right panel. Each of these genes was found to be expressed in both UHRR and HeLa total RNA. Despite their presence in the HeLa total RNA, they are found below the 50th percentile when the RIP-Chip experiment is analyzed on Agilent Oligonucleotide microarrays.