The classification of mRNA expression levels by the phosphorylation state of RNAPII CTD based on a combined genome-wide approach
- 5.4k Downloads
Cellular function is regulated by the balance of stringently regulated amounts of mRNA. Previous reports revealed that RNA polymerase II (RNAPII), which transcribes mRNA, can be classified into the pausing state and the active transcription state according to the phosphorylation state of RPB1, the catalytic subunit of RNAPII. However, genome-wide association between mRNA expression level and the phosphorylation state of RNAPII is unclear. While the functional importance of pausing genes is clear, such as in mouse Embryonic Stem cells for differentiation, understanding this association is critical for distinguishing pausing genes from active transcribing genes in expression profiling data, such as microarrays and RNAseq. Therefore, we examined the correlation between the phosphorylation of RNAPII and mRNA expression levels using a combined analysis by ChIPseq and RNAseq.
We first performed a precise quantitative measurement of mRNA by performing an optimized calculation in RNAseq. We then visualized the recruitment of various phosphorylated RNAPIIs, such as Ser2P and Ser5P. A combined analysis using optimized RNAseq and ChIPseq for phosphorylated RNAPII revealed that mRNA levels correlate with the various phosphorylation states of RNAPII.
We demonstrated that the amount of mRNA is precisely reflected by the phased phosphorylation of Ser2 and Ser5. In particular, even the most "pausing" genes, for which only Ser5 is phosphorylated, were detectable at a certain level of mRNA. Our analysis indicated that the complexity of quantitative regulation of mRNA levels could be classified into three categories according to the phosphorylation state of RNAPII.
KeywordsTranscription Start Site Phosphorylation State Gene Body Splice Junction Ser5 Residue
List of abbreviations
RNA polymerase II
transcription start site
transcription end site
Transcription factor II H
Negative elongation factor
DRB sensitivity including factor
Cellular function is accomplished by the accurate, regulated transcription of genes in the genome. The quantity of transcribed mRNA of protein-coding genes varies, and the regulation of transcription is carried out by a wide variety of nuclear factors on the chromatin structure. One of the key regulatory mechanisms is the control of the activation of RNA polymerase II (RNAPII) .
RNAPII transcribes all protein-coding genes and many non-coding genes, and the activity of RNAPII correlates with the phosphorylation state of RPB1, the large catalytic subunit of RNAPII . RPB1 has an unusual C-terminal domain (CTD) that consists of repeats of the heptapeptide consensus sequence N-Tyr1-Ser2-Pro3-Thr4-Ser5-Pro6-Ser7-C, of which there are 52 copies in mammals . The amino acids in these repeats are potential targets for modification, such as phosphorylation and glycosylation. During transcriptional regulation, free hypophosphorylated RNAPII is recruited to gene promoters. RNAPII's escape from the promoter requires TFIIH, a general transcription factor that mediates phosphorylation of CTD Ser5 . After promoter escape, RNAPII can move downstream of the transcription start site (TSS) ; however, pausing factors, such as NELF and DSIF, prevent productive elongation of mRNA . This phenomenon is known as promoter proximal pausing . Productive elongation of mRNA is coupled with phosphorylation of the CTD Ser2 residue . The influence of promoter proximal pausing of RNAPII may contribute to the control of gene expression levels [9, 10, 11]. It is possible that full length mRNA cannot be detected because of pausing, and that a wide variety of expression levels, including high expression, are regulated by pause site entry and escape of RNAPII . Recent studies revealed that RNAPII could bind to the promoter region of inactive genes in human fibroblasts , as well as in ES cells . Additionally, in mouse ES cells, Ser5 phosphorylated and Ser2 unphosphorylated RNAPII accumulates around the TSSs in bivalent genes . These genes, as differentiation markers, can be detected at low levels, despite their association with pluripotency . High throughput sequencing technology and cDNA analysis have emerged as revolutionary tools in recent years, but whether these sequencing data come from active transcription or pausing state genes, and the genome-wide phosphorylation status of RNAPII in vivo, have not been studied. Several genes in which RNAPII is in the pausing state play key role in differentiation ; therefore, understanding the correlation of RNAseq and RNAPII phosphorylation state is very important. To evaluate the phosphorylation status of RNAPII for all genes identified with RNAseq, we have to exclude free RNAPII, in which Ser2 and Ser5 residues are not phosphorylated, and distinguish actively transcribed genes, for which both of Ser2 and Ser5 residues are phosphorylated, from pausing state genes, for which Ser5 residues are only phosphorylated. Evaluation of the relationship between the phosphorylation state of RNAPII and mRNA expression level will permit the identification of those genes that are actively transcribed and those that are pausing.
A variety of techniques have been developed to quantify and analyze gene expression levels, such as northern blotting, RT-qPCR, SAGE, and microarrays. Recently emerged deep sequencers enable the analysis of mRNA expression with much less bias compared with previous technologies, by reading tens of millions of tags in a single run (RNAseq) . RNAseq can clarify the amount of previously identified transcripts , identify novel transcripts , and analyze tissue-specific alternative splicing . RNAseq is 1,000 times more sensitive than microarrays for quantifying transcripts, and appears to be the best currently available tool for the evaluation of mRNA . However, RNAseq has its own limitations. One such limitation is the need for reference sequences. The deep sequencer examines 25-200 bp short fragments, unlike previous technologies, and sequences tens of millions of fragments in a single run. These fragments, also known as 'reads', are mapped to a reference transcriptome to identify gene expression. However, because the transcriptomes are incomplete, even for well-studied species such as human and mouse, analysis of RNAseq data is restricted by the reference sequence, and requires another calculation to identify novel transcripts. TopHat  does not depend on a reference transcriptome, and provided a new way to evaluate novel transcripts, including new splicing sites. In addition, Cufflinks  can map reads to a reference genome and identify all transcripts quantitatively per kilobase of nucleotides and considers splicing. The weak point of quantification by these mapping techniques is the comparatively short sequence tag used to map to the reference genome. Success in mapping a sequence depends on the structure of the mRNA; it may have homologs that have a common structure, which may introduce bias to the statistical results. Therefore, to overcome these biases, it is necessary to use not only unique information, where one tag is mapped to one genomic region, but also multiple hit information, where one tag is mapped to two or more genomic regions. In TopHat the parameter 'Max multihits' controls how many regions one tag is allowed to map to, thereby optimizing mapping efficiency. However, a detailed evaluation of the influence of this parameter setting on the identification of mRNA has not been performed.
Thus, we used a deep sequencer to clarify how various mRNA expression levels are controlled, by analyzing the regulation of RNAPII through CTD phosphorylation. We categorized gene expression by identifying the phosphorylation control of RNAPII for all genes. In addition, by combining these data with genome-wide gene expression data that were obtained from RNAseq using the optimized 'Max multihits' parameter, we clarified the correlation between various mRNA expressions and RNAPII phosphorylation.
Results and Discussion
The accuracy of RNASeq is improved by permitting a small number of 'multihits'
The distribution of phosphorylated RNAPII
The composition of the CTD and the reactivity of each antibody
RNASeq can detect the expression of most genes, even in the "pausing" genes
Among 7,918 genes in which Ser2P is positive, 6,860 genes (87%) were assumed to be Ser5P positive too. On the other hand, among the 11,590 genes in which Ser5P is positive, 6,860 genes (59%) are assumed to be Ser2P positive. This result indicates that Ser5 and Ser2 of RNAPII have to be sequentially phosphorylated for active transcription, as described previously . However, 1,058 genes (13%) are Ser2P positive only. When these genes are observed in the UCSC genome browser (University of California, Santa Cruz) (Figure 3B), Ser2P single positive genes appear in the comparatively gene-dense areas. Moreover, when the ChIP-qPCR data were verified (Figure 3C), for instance, for SOX15, which is judged to be a Ser2P single positive, more Ser2P was identified around the TSS than around the TES, although the amount of Ser2P did gradually increase towards the TES (Figure 2C). These results suggest that Ser2P single positive genes are false positives caused by the influence of surrounding genes or non-annotated transcripts in these regions. RNAPII with an unphosphorylated CTD is first recruited to a promoter region and is then released when its Ser5 is phosphorylated. Active transcription is then initiated when Ser2 is phosphorylated; however, RNAPII keeps running until its termination, even if transcription ends . This results in the deterioration of the resolution of ChIPseq and it may be one factor that causes false positives in gene-dense areas. To overcome this limitation, we set a criterion in which we scored a peak as positive only when the peak extended over the gene body. Although this may affect the detection of RNAPII that is in the state of promoter proximal pausing, Ser5-phosphorylated RNAPII that is pausing around the TSS seems to be sufficiently detected when using this condition (Figure 3A-C).
Interestingly, RNAseq detected highly expressed genes not only in the state of active transcription (Ser5P+, Ser2P+), but also in the state of promoter proximal pausing (Ser5P+, Ser2P-), in the majority of FPKM > 0 genes. These results indicated that the phosphorylation of Ser5 and Ser2 correlates with gene expression in two stages. It also indicates that RNAseq, because of its high sensitivity, disregards the background epigenetic expression adjustment machinery associated with RNAPII phosphorylation. Some of the differentiation markers that were Ser5P single positive showed low mRNA expression in mouse embryonic stem cells . However, we should take note of the expression of differentiation markers, as interpreted by RNAseq, in stem cells, because some of these genes could be identified as a result of RNAseq's high sensitivity.
Among the 13,462 genes which RNAseq judged to have an FPKM value > 0, 11,156 genes (83%) are Ser2P and/or Ser5P positive. The remaining 2,306 genes (17%) with FPKM > 0 in RNAseq, were identified as neither Ser2P nor Ser5P in ChIPseq. Among 12,648 genes which ChIPseq judged to be Ser2P and/or Ser5P positive, 11,156 genes (88%) were FPKM > 0 genes in RNAseq.
To further investigate functional relationship among pausing/active genes and gene functions, we analyzed significant associations using Gene ontology  and Fishers' exact test(Additional File 3, Table S1). Hundreds of GO terms were calculated to be significant for active genes, and some of GO terms associated with mitochondorial genes were judged to be significant for pausing genes. Neither calculation seemed to give significant enrichment of specific genes, except for housekeepking genes.
Gene expression levels reflect the level of phosphorylation of RNAPII
RNAPII status can be classified into three categories for transcribed genes
When tags were summed for genes with FPKM > 0/Ser2P(+)/Ser5P(-), the number of Ser2P tags tended to be high (Additional File 4, Figure S3 B). However, for these genes, the tag count outside of the gene (X axis is more than 1 or less than 0) for Ser2P and Ser5P are also higher than for other gene categories, and this may indicate that they were picked up from the background noise generated by surrounding genes. The Ser5P tags showed a small peak around the TSS of these genes (Additional File 4, Figure S3 A), and when the background was excluded, the shape of the graph obtained from FPKM > 0/Ser2P(+)/Ser5P(-) genes was approximately the same as that from FPKM > 0/Ser2P(-)/Ser5P(-). These results suggest that the genes whose expression is confirmed by RNAseq can be classified into three categories by combining ChIPseq data concerning Ser2P/Ser5P: 'High pausing, High elongation (Ser5P+, Ser2P+)', 'High pausing, Low elongation (Ser5P+, Ser2P-)', and 'Low pausing, Low elongation (Ser5P-, Ser2P-)'.
Control of mRNA expression is correlated to phased phosphorylation of Ser2 and Ser5
Logistic regression analysis for RNAPII CTD phosphorylation Whole model test
Prob > ChiSq
Observations (or sum weights)
Prob > ChiSq
Intercept [S2P- S5P-]
log2(FPKM+1) [S2P- S5P-]
Intercept [S2P+ S5P+]
log2(FPKM+1) [S2P+ S5P+]
We studied the association between mRNA expression level and RNAPII phosphorylation state in Hela cells using a deep sequencer for RNAseq and ChIPseq analysis. During verification to improve the accuracy of RNAseq, we found that the correlation between RNAseq and past expression microarray data could be increased by adjusting the 'Max multihits' parameter. We optimized this parameter such that it minimized the risk of reading genes that are not simultaneously expressed. We also produced an antibody against the phosphorylated form of RNAPII, which allowed the genome-wide visualization of the state of RNAPII phosphorylation using ChIPseq. RNAseq and ChIPseq showed a very high correlation, and the existence of RNAPII on approximately 82% of genes that were detected with RNAseq was confirmed in ChIPseq. In addition, when we examined the relationship between the phosphorylation state of RNAPII and the level of mRNA expression, phosphorylation of both Ser2 and Ser5 of RNAPII was confirmed for almost all highly expressing genes. When only Ser5 of PNAPII was phosphorylated, low mRNA expression was detectable by RNAseq, in spite of the pausing state. Moreover, when tag counts of Ser5P were counted for genes identified only with RNAseq, the existence of a slightly higher level of Ser5P was detected compared with the negative control. This indicated that transcriptional adjustment is performed in two stages: promoter escape and active elongation. We also provide a hypothesis that gene expression can be classified into three groups according to the phosphorylation state of RNAPII.
Hela cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum under a humidified atmosphere with 5% CO2 at 37°C.
Production of rat monoclonal antibodies
The RNA polymerase II (RNAPII) Ser2P and Ser5P antigens were synthesized based on their specific sequences, Ser2P: SPTSPSYSphPTSPSYSPTSPS and Ser5P: SPTSPSYSPTSphPSYSPTSPS (Sigma-Aldrich). A C-terminal cysteine residue that was not part of the sequence was introduced to allow coupling to the carrier protein maleimide-activated keyhole limpet hemocyanin (Thermo Scientific). The coupling reaction was performed according to the supplier's instructions. Rat monoclonal antibodies were generated based on the rat lymph node method established by Sado et al. . A 10-week-old female lzm rat (Japan SLC) was injected in the rear footpads with 500 μl of an emulsion containing 200 μg RNAP2 Ser2P or Ser5P peptide and Freund's complete adjuvant. After two weeks, the cells from the lymph nodes of the immunized rat were fused with mouse myeloma Sp2/0-Ag14 cells at a ratio of 5:1 in 50% polyethylene glycol (Merck) solution. The resulting hybridoma cells were plated onto 96-well plates and cultured in HAT selection medium [hybridoma SFM medium (Invitrogen); 10% fetal bovine serum; 10% BM-Condimed H1 (Roche); 100 μM hypoxanthine; 0.4 μM aminopterin; 1.6 μM thymidine]. At seven days post-fusion, the hybridoma supernatants were screened using an enzyme-linked immunosorbent assay (ELISA) against each antigen. Positive clones were subcloned and rescreened by ELISA (Table 1). To prepare hybridoma supernatants containing highly concentrated antibodies, the resulting positive clones, 3E7C7 for RNAP2 Ser2P and 1H4B6 for RNAP2 Ser5P, were cultured at a high cell density using a MiniPERM bioreactor (Vivascience).
BSA conjugated RNAPII Ser2P or Ser5P peptides (5 μg/mL) at dilutions ranging from 1:100 to 1:100000 in ELISA buffer [10 mM sodium phosphate pH7.0] were adsorbed on the surface of 96-well costar Serocluster 96 Well "U" Bottom Plates (Corning) by overnight incubation at 4°C. To avoid non-specific binding, the plates were blocked with 1% bovine serum albumin (BSA) in PBS. Hybridoma supernatants were applied to the plates and incubated for 1 h at room temperature and then washed three times with PBS. The plates were incubated for 30 min at room temperature with alkaline phosphatase-conjugated anti-rat IgG antibody (Sigma) at a dilution of 1:10000. After washing three times with TBS-T, immunoreactivity was visualized using a pNPP phosphatase substrate system (KPL).
Hela cells were washed twice with phosphate buffered saline (PBS), centrifuged, and then resuspended in 2 × SDS sample buffer. The samples were separated by SDS-PAGE and transferred to a nitrocellulose membrane with iBlot (Invitrogen). The membrane was blocked for 1 h in 5% (w/v) skimmed milk in Tris-buffered saline containing 0.05% (v/v) Tween 20 (TBST), then incubated with primary antibodies in solution 1 (TOYOBO). The blot was then incubated with horseradish peroxidase-labeled secondary antibodies and detected using the WestDura chemiluminescence kit (Pierce). The primary antibodies were anti-RNAPII Ser2P (3E7C7, hybridoma supernatant, 1:1000; Figure 2B and Table 1), anti-RNAPII Ser5P (1H4B6, hybridoma supernatant, 1:1000; Figure 2B and Table 1), and sc-899, the antibody against the N-terminus of RNAPII (1:1000; Figure 2B). Secondary antibodies were horseradish peroxidase-conjugated anti-rat IgG antibodies (1:5000; GE Healthcare).
Total RNA was isolated and reversed-transcribed with Takara Prime Script Reverse Transcriptase and an oligo dT primer, as previously described . Quantitative-PCR (Q-PCR) was performed using TaKaRa SYBR Premix Dimer Eraser. Q-PCR data are presented as the mean ± standard deviation of three independent experiments. Primer sequences are available upon request.
Libraries were generated by the modified Illumina protocol using the mRNAseq preparation kit. Briefly, 1 μg of total RNA was enriched for polyA RNA by two successive rounds of oligo(dT) selection. The polyA RNA was then fragmented, and first-strand cDNA synthesis was performed using random hexamer priming. Following second-strand cDNA synthesis, dsDNA was repaired using T4 DNA polymerase, Klenow enzyme, and T4 polynucleotide kinase (PNK) (New England Biolabs), followed by treatment with Klenow exo- to add an A base to the 3' end. After ligation of the Solexa adaptor using TaKaRa ligation Mix (TaKaRa), the adaptor-ligated DNAs were amplified using Solexa PCR primers for 18 cycles, and the amplified library was isolated from an agarose gel. The samples were purified using the QIAquick MinElute kit (Qiagen) at each preparation step.
RNASeq data analysis
For each sample, cDNA was sequenced (single 36 bp read) by an Illumina Genome Analyzer GAIIx. The base-called sequences were obtained using SCS2.7 from RNAseq image data. To calculate the total amount of the transcripts of each mRNA, a series of programs-Bowtie , TopHat (v1.1.4) , and Cufflinks (v0.9.3) -were used. Briefly, RNAseq reads were mapped against the whole reference genome (hg19) using Bowtie. The reads that did not align to the genome but were mapped to potential splice junctions by TopHat were considered to bridge splice junctions. The quantification of transcripts, with normalization for gene length, was performed by Cufflinks. All of the parameters, except 'Max multihits' (TopHat), were substituted with default options (TopHat: -g options as utilized as "multihits". Cufflinks: default suggested as -m 230 -s 20 -I 300000). The 'Max multihits' was set at 1, 2, 5, 10, 20, 40, 100, and 1000, and then the number of FPKM > 0 genes was determined (Figure 1A, Additional File 2, Figure S2). The Spearman's correlation coefficients with microarray data (Figure 1B), the percentage of splice sites that were included in the gene body (Figure 1C), and histograms of FPKM distributions at three 'Max multihits' values (Figure 1D) were plotted.
Gene ontology and Fishers' exact test
For the analysis, we used Funcassociate 2 , which is a web application tool http://llama.mshri.on.ca/funcassociate/ that finds significant Gene ontology terms from large-scale experimentation. All of the parameters were substituted with default options, i.e. Mode:unordered, Over/Under:over, Simulations:1000, and Significance Cutoff:0.05.
Chromatin Immunoprecipitation (ChIP)
ChIP assays were performed by modifying the Upstate Biotechnology protocol, as described previously  except adding 40 mM β-glycerophosphate and 1 mM sodium fluoride to immunoprecipitation buffer, utilizing rat monoclonal antibodies against RNAPII Ser2P (3E7C7, 5 μg; Figure 2B and Table 1) and RNAPII Ser2P (3E7C7, 5 μg; Figure 2B and Table 1). Relative recruitment (Figure 3C) was defined as the ratio of amplification of the PCR product relative to 1% of input genomic DNA. Q-PCR data are presented as the mean ± standard deviation of three independent experiments. We designed PCR primers for gene regions within the 3 kb downstream of the 5'-start of each gene and within the 3 kb upstream of the 3'-end of each gene because Ser5 phosphorylated RNAPII is positioned in the coding region at +2 to +4 kb from start site, as well as upstream of start site . Coding regions were used to prevent any effects from neighboring genes. Primer sequences are available upon request.
For ChIPseq, sample preparation was performed using the ChIP protocol described above. The ChIP DNA and the Input DNA ends were repaired using T4 DNA polymerase, Klenow enzyme, and T4 polynucleotide kinase (PNK) (New England Biolabs), followed by treatment with Klenow exo- to add an A base to the 3' end. After ligation of the Solexa adaptor using TaKaRa ligation Mix (TaKaRa), the adaptor-ligated DNAs were amplified using Solexa PCR primers for 18 cycles, and the amplified library was isolated from an agarose gel. The samples were purified using the QIAquick MinElute kit (Qiagen) at each preparation step. The purified library was used for cluster generation and sequencing analysis using the Genome Analyzer GAIIx (Illumina K. K.).
ChIPSeq data analysis
Base-called sequences were obtained using SCS2.7 from ChIPseq image data. The sequence tags for RNAPII Ser2P and Ser5P and Input were aligned to the human genome (hg19) using Bowtie  software. Peak detection and identification of binding sites of RNAPII Ser2P and Ser5P were obtained by correcting from Input DNA using Peakseq software, as described previously . The box plot of RNAPII Ser2P and Ser5P enriched regions that were found in Peakseq when using the threshold of P-value < 0.05, Q-value < 0.05 is shown in Figure 3B. We defined RNAPII recruitment as positive if the box plot overlapped the gene body to create the Venn diagram (Figure 3A, Additional File 2, Figure S2). For the detection of the binding site of RNAPII Ser2P and Ser5P, all tags normalized to input by Peakseq were summed according to their shifted positions, with the definition that a gene length was 1, and along the horizontal axis. 0 indicates the TSS (Transcription start site) and 1 indicates the TES (Transcription end site) (Figure 2C, 5 and Additional File 4, Figure S3).
Any experimental procedure in this study does not contain any animal experiment.
The raw illumina sequencing data are available from the DNA Data Bank of Japan (DDBJ) with accession number [DDBJ: DRA000363].
Logistic regression analysis
Logistic regression was used to estimate the probability of each RNAPII CTP phosphorylation state across the FPKM range. A logistic probability plot for RNAPII state was also created. All calculations were performed using JMP v 8.0.2 (SAS, Cary, NC) running under Windows XP.
We thank Dr. H. Kimura, Dr. H Kurumizaka for advice, and Ms. Ito and Ms. Onishi for technical support. This work was supported in part by grants from the Ministry of Education, Culture, Sports, Science, and Technology of Japan, the Kaibara Morikazu Medical Science Promotion Foundation.
- 2.Kershnar E, Wu SY, Chiang CM: Immunoaffinity purification and functional characterization of human transcription factor IIH and RNA polymerase II from clonal cell lines that conditionally express epitope-tagged subunits of the multiprotein complexes. J Biol Chem. 1998, 273: 34444-34453. 10.1074/jbc.273.51.34444.CrossRefPubMedGoogle Scholar
- 11.Alder O, Lavial F, Helness A, Brookes E, Pinho S, Chandrashekran A, Arnaud P, Pombo A, O'Neill L, Azuara V: Ring1B and Suv39h1 delineate distinct chromatin states at bivalent genes during early mouse lineage commitment. Development. 2010, 137: 2483-2492. 10.1242/dev.048363.CrossRefPubMedPubMedCentralGoogle Scholar
- 19.Trapnell C, Williams BA, Pertea G, Mortazavi A, Kwan G, van Baren MJ, Salzberg SL, Wold BJ, Pachter L: Transcript assembly and quantification by RNAseq reveals unannotated transcripts and isoform switching during cell differentiation. Nat Biotechnol. 2010, 28: 511-515. 10.1038/nbt.1621.CrossRefPubMedPubMedCentralGoogle Scholar
- 21.Viegas MH, Gehring NH, Breit S, Hentze MW, Kulozik AE: The abundance of RNPS1, a protein component of the exon junction complex, can determine the variability in efficiency of the Nonsense Mediated Decay pathway. Nucleic Acids Res. 2007, 35: 4542-4551. 10.1093/nar/gkm461.CrossRefPubMedPubMedCentralGoogle Scholar
- 22.Scotto L, Narayan G, Nandula SV, Arias-Pulido H, Subramaniyam S, Schneider A, Kaufmann AM, Wright JD, Pothuri B, Mansukhani M, Murty VV: Identification of copy number gain and overexpressed genes on chromosome arm 20q by an integrative genomic approach in cervical cancer: potential role in progression. Genes Chromosomes Cancer. 2008, 47: 755-765. 10.1002/gcc.20577.CrossRefPubMedGoogle Scholar
- 29.Sado Y, Kagawa M, Kishiro Y, Sugihara K, Naito I, Seyer JM, Sugimoto M, Oohashi T, Ninomiya Y: Establishment by the rat lymph node method of epitope-defined monoclonal antibodies recognizing the six different alpha chains of human type IV collagen. Histochem Cell Biol. 1995, 104: 267-275. 10.1007/BF01464322.CrossRefPubMedGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.