Cloning and sequence analysis of CDK12S.
In a previous study, the expression profile of various protein kinases in rat cortical neurons was analyzed using degenerate RT-PCR and a nucleotide sequence identified as a novel CDC2-like kinase was isolated (18
). Using RACE, we obtained a contiguous nucleotide sequence of 4,381 bp encoding an open reading frame of 1,258 amino acids. A polyadenylation signal was found at bp 4231 (Fig. ). Functional domains present in the deduced amino acid sequences were predicted by MotifScan search (http://hits.isb-sib.ch/cgi-bin/PFSCAN?
) and included a bipartite nuclear localization signal, an RS domain, and a CDC2-like serine/threonine kinase domain (Fig. ). According to the characteristics of the sequence and the following data, we named the protein the short form of cyclin-dependent kinase 12 (CDK12S
, accession number NM_138916). The sequences of the human and mouse CDK12S
were then acquired by database comparison and were confirmed by RT-PCR and DNA sequencing. The deduced protein sequences of CDK12S
are highly conserved among these species, with 90.7% identity between human and mouse, 91.8% identity between human and rat, and 96.5% identity between mouse and rat.
FIG. 1. Protein, mRNA, and genomic structures of CDK12S and CDK12L. (A) Protein domains of CDK12S and CDK12L. The amino acid sequences of rat CDK12S and CDK12L were analyzed by the MotifScan program. CDK12S and CDK12L share 1,249 amino acids of sequence. They (more ...)
A search of the nonredundant sequence database revealed that the cDNA sequences of the human Crkrs
(NM_016507) and human CDK12S
genes are identical at their 5′ ends (16
). This suggests that they are alternatively spliced isoforms of the same gene. To confirm that the alternative splicing event is conserved, we cloned rat Crkrs
cDNA, which encodes a protein of 1,484 amino acids, from E14.5 brain by RT-PCR. Alignment of the rat CDK12S
cDNA sequences and comparison of these two sequences to rat genomic sequences indicates that CDK12S
cDNAs are transcripts of the same gene, CDK12
, and differ at the 3′ end (Fig. ).
There are 14 exons in the CDK12
gene. The CDK12S
transcript skips the 5′ splice site at the end of exon 13 and contains exon 13′. As there is an in-frame stop codon in exon 13′, this results in a shortened open reading frame for CDK12S
. In contrast, crkrs
transcript splices out exon 13′ and connects exon 13 and 14, resulting in a longer open reading frame. For clarity, we renamed CrkRS CDK12L
. Thus, the rat CDK12S
and rat CDK12L
proteins share the same 1,249 amino acids of the amino termini, but have an additional 9 and 235 distinct amino acid residues, respectively, at the carboxyl termini. An identical exon organization for the CDK12
gene is observed in the human and rodent genomes, albeit with different lengths of introns in the three species (Fig. ). The CDK12
gene is located on human chromosome 17q12 (AC009283) and on mouse chromosome 11 (AL591205). The closest homologous protein to human CDK12 is human CDC2L5 (22
), which also contains a CDC2-like protein kinase domain and an RS domain.
Expression analyses of rat CDK12.
Previously, Ko et al. detected CDK12 expression by Northern blotting using adult human tissues. Its expression in embryonic tissues was not analyzed. Moreover, the sizes of CDK12L and CDK12S transcripts were not determined. Thus, the expression of rat CDK12L and CDK12S transcripts in various tissues during embryonic development was analyzed using a probe (probe 1) that recognizes both transcripts. When total RNAs derived from E14.5 tissues were analyzed, two transcripts of CDK12 with sizes of 9.3 and 6.8 kb were detected (Fig. ). The CDK12 mRNAs were ubiquitously expressed in E14.5 tissues, including brain, spinal cord, heart, lung, liver, gut, and limb (Fig. ). We also examined the temporal expression pattern of CDK12 during the development of the nervous system. The CDK12 mRNAs were expressed in E10.5 neural tube, which was the earliest time point examined (Fig. ). The levels of CDK12 expression in rat brain are more abundant in the embryonic stages, gradually decrease as development proceeds, and became barely detectable at the adult stage (Fig. ).
FIG. 2. Expression of CDK12 in embryonic rat tissues. (A) We used 5 μg of total RNA from various E14.5 rat tissues analyzed by Northern blotting using probe 1 depicted in Fig. . Two major transcripts, 6.8 kb and 9.3 kb (arrowheads), were (more ...)
To distinguish which transcripts correspond to CDK12S and CDK12L mRNAs in Northern blots, we designed probe 2 and probe 3, which specifically hybridize to the CDK12S and CDK12L mRNAs, respectively (Fig. ). In the Northern blots using CDK12S-specific probe 2, only the 6.8-kb band could be observed (Fig. ). Surprisingly, CDK12L-specific probe 3 could hybridize both the 9.3-kb and 6.8-kb bands, a pattern identical to that observed by probe 1 (Fig. ). The data indicate that two transcripts, with sizes of 9.3 and 6.8 kb, encode CDK12L; and a transcript with a size of 6.8 kb encodes CDK12S.
To verify that CDK12S and CDK12L transcripts can be translated to proteins and to examine endogenous protein expression of CDK12 in rat embryonic tissues by Western blotting, rabbit and mouse antisera (rαCDK12 and mαCDK12) were raised against peptide sequences 1040 to 1212 of rat CDK12 (peptide 1 in Fig. ) such that the antisera recognize both CDK12S and CDK12L. Mouse antisera (mαCDK12L) against amino acids 1297 to 1402 (peptide 2 in Fig. ) were also generated to detect only CDK12L. The specificity of the antibodies was determined by Western blotting using lysates of HEK293T cells transfected with a control vector (pEF), phCDK12S or phCDK12L. The latter two vectors were able to express human CDK12S and CDK12L protein, respectively.
In cells transfected with pEF, both mαCDK12 and mαCDK12L
identified a 200-kDa band, suggesting that this band corresponds to CDK12L
(Fig. , lanes 1 and 4). mαCDK12, but not mαCDK12L
, bound an additional 180-kDa band, suggesting that this band corresponds to CDK12S
(Fig. , lane 1). The data also indicated that HEK293T cells express CDK12S
endogenously. The specificity of antibodies was further validated when CDK12S
- or CDK12L
-overexpressing HEK293T cell lysates were analyzed by Western blotting (Fig. ). The sizes of CDK12S
are larger than those predicted (140.0 kDa for CDK12S
and 163.9 kDa for CDK12L
), possibly due to posttranslational modification such as phosphorylation (16
We then used these two antibodies to analyze the expression pattern of CDK12 in E14.5 rat tissues (Fig. ). Both CDK12S and CDK12L proteins were detected in E14.5 brain, spinal cord, heart, lung, liver, gut, and limb. The sizes of these two bands are identical to those found in HEK293T cells.
Interaction between CDK12, and cyclins L1 and L2.
It is possible that activation of CDK12 requires a cyclin. Examination of the cellular localization of known cyclins points to members of cyclin class L, cyclin L1 and cyclin L2, as good candidates for the cyclin partner of CDK12, as they contain an RS domain and are located in the nuclear speckles (6
). To test this hypothesis, we determined whether CDK12 interacts with the cyclin L members.
and EGFP-tagged cyclin L1α (the longest isoform of cyclin L1) were overexpressed in HEK293T cells. The subcellular localization of CDK12S
, recognized by rαCDK12, is in nuclear speckles and appears similar to the subcellular pattern of CDK12L
reported previously (Fig. ) (16
). As expected from previous studies (1
), EGFP-tagged cyclin L1α is located in the nuclear speckles (Fig. ) and is colocalized with CDK12S
(Fig. ). CDK12S
and cyclin L2α are also colocalized in nuclear speckles when overexpressed together in HEK293T cells (Fig. ).
FIG. 3. Interaction between CDK12 and cyclin L1 and cyclin L2. (A to C) HEK293T cells transfected with pCDK12S-Myc and pEGFP-cyclin L1α-Flag were stained with the rαCDK12 antiserum (A) and examined for EGFP fluorescence (B). The merged photograph (more ...)
To demonstrate the possibility that CDK12 can associate with cyclin L1 and cyclin L2, immunoprecipitation experiments were performed. Cell lysates containing overexpressed CDK12S or CDK12L and Flag-tagged cyclin L1α or Flag-tagged cyclin L2α were immunoprecipitated by an anti-Flag antibody. The immune complex was then subjected to Western blotting. Both CDK12S and CDK12L were coimmunoprecipitated with cyclin L1α or L2α (Fig. ). Similarly, when cell lysates containing overexpressed Myc-tagged CDK12S and Flag-tagged cyclin L1α were immunoprecipitated by an anti-Myc antibody, it was shown that cyclin L1α was coimmunoprecipitated with CDK12S (Fig. , lane 2).
FIG. 4. Mapping of interaction domains between CDK12 and cyclin L1. (A) Schematic structures of CDK12L, Myc-tagged CDK12S, various Myc-tagged CDK12 truncated constructs, Flag-tagged cyclin L1α, and Flag-tagged cyclin L1β. (B) Lysates of HEK293T (more ...)
To study further the interaction between CDK12 and cyclin L1, we next mapped the domains involved in the interaction. Since CDK12S is sufficient to bind cyclin L1α and the amino acid sequence of CDK12S is included in CDK12L (except the last 9 amino acids), we used CDK12S as the starting template to study the interaction between CDK12 and cyclin L1. The Myc-tagged CDK12S or Myc-tagged truncated proteins (Fig. ) were coexpressed with Flag-tagged cyclin L1α in HEK293T cells. Immunoprecipitation with the anti-Flag antibody and Western blotting were then performed. The results showed that cyclin L1α protein was coprecipitated with CDK12S and the truncated CDK12 protein containing the kinase domain and the carboxyl-terminal domain (CDK12SKC, Fig. , left panel).
In the reverse immunoprecipitation experiment, CDK12S and the kinase domain (CDK12K), carboxyl terminus (CDK12SC) or CDK12SKC were coprecipitated with cyclin L1α protein by the anti-Flag antibody (Fig. , right panel). The cause for the discrepancy observed between these two immunoprecipitation experiments is unclear. One possibility is that binding of the anti-Myc antibody to CDK12K and CDK12SC during immunoprecipitation may generate steric hindrance that inhibits interaction between these proteins and cyclin L1α. Together, these results suggest that the kinase domain and carboxyl-terminal domain of CDK12S interact with cyclin L1α. In a control experiment, the anti-Flag antibody was unable to pull down CDK12S or truncated proteins using lysates of cells transfected with the CDK12 constructs but without pcyclin L1α-Flag (data not shown).
As many cyclins interact with CDK through the cyclin domain, we cloned a naturally occurring cyclin L1 isoform, cyclin L1β, which contains the cyclin domain but no carboxyl-terminal RS domain (Fig. ), and examined its interaction with CDK12 by immunoprecipitation. The results showed that cyclin L1β can interact with CDK12S (Fig. ). The kinase domain of CDK12 can also interact with cyclin L1β in the immunoprecipitation experiment (Fig. ).
CDK12 and cyclin L1 are regulators of alternative splicing.
It was reported that human CDK12L
immune complex contains a kinase activity that phosphorylates SF2/ASF in vitro (16
). We also found that immunoprecipitated rat CDK12S
can phosphorylate SF2/ASF, although CDK12S
does not directly phosphorylate SF2/ASF (see Fig. S1 in the supplemental material). As SF2/ASF is a member of the SR proteins, which are involved in the regulation of constitutive splicing as well as alternative splicing (9
), we tested whether CDK12 and cyclin L1 change splicing site selection of an E1a minigene in HEK293T cells.
The pre-mRNA of the E1a minigene can be processed into five mRNAs. Three major forms, 13S, 12S, and 9S, derive from selection of one of three 5′ splice sites, and two minor forms, 11S and 10S, involve additional splicing using an upstream 3′ splice site (Fig. ). We transfected the E1a minigene together with pCDK12S or pCDK12L and then quantified the intensity ratios of the 13S, 12S, and 9S products after RT-PCR and Southern blotting. We normalized the percentages of the three major bands to those in the control experiments, so that changes in the splicing patterns are easily visualized.
FIG. 5. Regulation of alternative splicing by CDK12 and cyclin L1. (A) A schematic diagram illustrating the splicing pattern of the E1a reporter gene. The alternative 5′ splice sites and splicing events that generate 13S, 12S, and 9S mRNAs are indicated (more ...)
In HEK293T cells, both CDK12S
decreased the 13S transcript and increased the 9S transcript (Fig. ). Moreover, this effect of CDK12 on E1a splicing pattern is dosage dependent (Fig. ). Similar dosage-dependent increases of the 9S transcript and decrease of the 12S and 13S transcripts were also observed when cyclin L1α and cyclin L2α were transfected into HEK293T cells (Fig. ; see Fig. S2B in the supplemental material). Changes of the splicing pattern of the E1a minigene were also observed when CDK12S
and cyclin L1α were overexpressed in P19 cells (Fig. , lanes 1 and 2). P19 cells are a mouse embryonic carcinoma cell line (8
FIG. 6. Changes in the alternative splicing pattern in cells with decreased expression of CDK12. (A and B) Proteins and RNAs of P19 cells transfected with 0.4 μg of pCEP4-E1a and 1.2 μg of pmU6, pCDK12si1, and pCDK12si2 were prepared. Proteins (more ...)
As cyclin L1β, which contains mainly the cyclin domain, had lower activity in the E1a splicing assay (see Fig. S2A in the supplemental material), we examined whether it could potentiate the effects of CDK12. There was no change in the splicing pattern when 0.8 μg of pcyclin L1β was transfected into HEK293T cells (Fig. , lane 4). However, this dose of pcyclin L1β potentiated the effects induced by 2.8 μg of phCDK12S and phCDK12L (Fig. , compare lanes 5 and 6 to lanes 2 and 3).
To further confirm that CDK12 proteins are involved in alternative splicing, we knocked down mouse CDK12 in P19 cells, in which the expression of endogenous CDK12 is higher than that in HEK293T cells (data not shown). Two small interfering RNA constructs, pCDK12si1 and pCDK12si2, were transfected into P19 cells. Both constructs decreased endogenous CDK12 proteins by at least 40% (Fig. ). When the E1a minigene splicing assay was performed in P19 cells, fewer of the 13S transcripts and more of the 9S transcripts were detected in the control conditions (Fig. , lane 1) compared to the levels in HEK293T cells. When CDK12 was knocked down, there were increases in the 13S transcript and decreases in the 9S transcript levels (Fig. , lanes 2 and 3), a pattern opposite what was observed for the CDK12-overexpressed experiments. The small interfering RNAs also completely blocked any effect on the alternative splicing pattern generated by overexpression of CDK12S (Fig. , compare lane 5 to lane 2). Moreover, when the expression of endogenous CDK12 was suppressed, the effect of overexpressed cyclin L1α on alternative splicing was also partially inhibited (Fig. , compare lane 6 to lane 3), suggesting that the effects of cyclin L1 on alternative splicing, at least in part, are mediated by CDK12.
Effect of CDK12 on E1a splicing pattern is compromised by overexpression of SF2/ASF and SC35.
Many SR splicing factors have been shown to increase the 13S and 12S transcripts and to decrease the 9S transcript, a pattern opposite what was observed in the case of CDK12 in the E1a splicing assay (26
). We thus hypothesized that CDK12 may deplete the activities of SR proteins in the splicing machinery by indirect phosphorylation or sequestration of SR proteins and change the way in which the splicing machinery selects splice sites. If this hypothesis is correct, we reasoned that overexpression of SR proteins may counteract the effects of CDK12 on the E1a splicing pattern.
We examined this possibility by overexpressing five SR splicing factors, including SF2/ASF, SC35, SRp40, SRp55, and SRp75, with or without CDK12S in HEK293T cells for the E1a assay. In HEK293T cells, due to their already greater amounts of the 13S transcript, no further increase of the 13S transcript was observed by overexpression of SR proteins alone (Fig. , lanes 3 to 6), except the minor increase caused by overexpression of SF2/ASF (Fig. , lane 2). Among the five SR proteins tested, SF2/ASF partially inhibited and SC35 almost eliminated the effect of CDK12S in the E1a assay (Fig. , compare lanes 8 and 9 to lane 7). The other three SR proteins had no influence on CDK12S activity. Some of the HEK293T cell lysates used for the splicing assay were used in Western analyses to detect the overexpressed proteins. The results showed that SR proteins were expressed as expected and ectopic expression of SR proteins did not affect synthesis of CDK12S proteins (see Fig. S3 in the supplemental material). Moreover, no changes in the molecular weights of all the SR proteins tested were observed when SR proteins and CDK12S proteins were coexpressed.
FIG. 7. Inhibition of CDK12 alternative splicing activity by SR proteins. (A and B) Splicing assays of HEK293T cells transfected with 0.4 μg of pCEP4-E1a and 1.8 μg of the plasmids listed above the panels were performed. The splicing assays were (more ...)
Previous reports had shown that the RNA recognition motifs (RRMs), but not the RS domain, mediate the effect of SF2/ASF in the E1a splicing assay (4
). We thus tested whether the RRMs of SF2/ASF are sufficient to block the effect of CDK12S
in the E1a assay. As expected, the RRMs of SF2/ASF, but not the RS domain, changed the E1a splicing pattern (Fig. , lanes 2 and 3) and eliminated the effect of CDK12S
(Fig. , compare lane 7 to lane 6). However, neither the RRM nor the RS domain of SC35 offset the effect of CDK12S
, suggesting that both domains of SC35 were required to counteract CDK12S
(Fig. , lanes 9 and 10). Together, these data suggested that one of the mechanisms by which CDK12 regulates alternative splicing is mediated by blockage of the activities of SF2/ASF and SC35.