Cyclin-dependent kinase 4 may be expressed as multiple proteins and have functions that are independent of binding to CCND and RB and occur at the S and G₂/M phases of the cell cycle

Abstract

Cyclin-dependent kinase 4 (CDK4) is known to be a 33 kD protein that drives G₁ phase progression of the cell cycle by binding to a CCND protein to phosphorylate RB proteins. Using different CDK4 antibodies in western blot, we detected 2 groups of proteins around 40 and 33 kD, respectively, in human and mouse cells; each group often appeared as a duplet or triplet of bands. Some CDK4 shRNAs could decrease the 33 kD wild-type (wt) CDK4 but increase some 40 kD proteins, whereas some other shRNAs had the opposite effects. Liquid chromatography–mass spectrometry/mass spectrometry analysis confirmed the existence of CDK4 isoforms smaller than 33 kD but failed to identify CDK4 at 40 kD. We cloned one CDK4 mRNA variant that lacks exon 2 and encodes a 26 kD protein without the first 74 amino acids of the wt CDK4, thus lacking the ATP binding sequence and the PISTVRE domain required for binding to CCND. Co-IP assay confirmed that this ΔE2 protein lost CCND1- and RB1-binding ability. Moreover, we found, surprisingly, that the wt CDK4 and the ΔE2 could inhibit G₁–S progression, accelerate S–G₂/M progression, and enhance or delay apoptosis in a cell line-specific manner in a situation where the cells were treated with a CDK4 inhibitor or the cells were serum-starved and then replenished. Hence, CDK4 seems to be expressed as multiple proteins that react differently to different CDK4 antibodies, respond differently to different shRNAs, and, in some situations, have previously unrecognized functions at the S–G₂/M phases of the cell cycle via mechanisms independent of binding to CCND and RB.

Introduction

Cyclin-dependent kinase 4 (CDK4) drives G₁-to-S phase progression of the cell cycle. Forming a holoenzyme with one of the 3 D-type cyclins (CCND1, CCND2, and CCND3) to phosphorylate a retinoblastoma (RB) protein is its canonical mechanism. CDK6 is a close sibling of CDK4 and also binds to a D-type cyclin to regulate the G₁ progression. CDK4 activity at the S-to-G₂/M phases of the cell cycle has also been reported^1-4 but has not yet received good recognition.

Besides CCND, 2 other groups of proteins can bind to CDK4/6 as well. One group includes p21cip1 and p27kip1, which help in assembling cyclin–CDK complex and internalizing the complex into the nucleus when they are present as a single molecule, but which inhibit the kinase activity when more molecules are present, i.e., overexpressed.^5,6 The other group is specific for CCND-CDK4/6 and includes p16ink4a (p16), p15ink4b, p18, etc.; they are also required for the assembling of the CCND-CDK4/6 complex but inhibit the kinase activity.^5,6

All CDK proteins share 3 common domains that are crucial for their binding to cyclins and inhibitors and for their kinase activity, i.e., (1) the ATP binding sequence and (2) the PSTAIRE domain, both at the N terminus, as well as (3) the kinase sequence in the middle region (Fig. 1).⁷ Like many other kinases, CDK4 protein has a typical 2-lobe structure, with its residuals 1–96 as the N-terminal domain and the remaining residuals (97–303) as the C-terminal domain.⁸ The N-terminal domain uses the PISTVRE sequence, which varies slightly among different CDKs and among animal species to engage CCND.^8,9 Human germline mutations and spontaneous mutations in human cancers have been found, but only at very low frequencies and mainly at codon 24.¹⁰ Site mutagenesis studies suggest that mutations in codons 22 and 24, both located in exon 2, significantly affect its binding to p16 and CCND1.¹¹

Figure 1. Expression of an exon 2-deleted CDK4 variant. Top panel: A 5′ part of hCDK4 (A) and mCDK4 (B) mRNAs with exon 2 underlined. The atg¹ in exon 2 and atg² in exon 3 are the start codons for the wt and the ΔE2, respectively. There are several in-frame atg or ctg start codons and tag or taa stop codons (underlined) upstream of atg.¹ Our wt mCDK4-HA construct starts from atg¹ whereas our ΔE2-HA construct has atg¹ and its downstream 55 nt (in gray) deleted. The ATP binding, PSTAIRE and kinase sequences in the hCDK4 protein are underlined (C). The ΔE2 protein lacks the first 74 amino acids (in green). Middle panel: In agarose gel, RT-PCR products from 67RN mouse breast cancer cells with the F109 and R1026 primers appear as three bands (A), which are confirmed by sequencing to be the wt mCDK4 (band-a), a wt/ΔE2 heterodimer (band-b), and the ΔE2 that lacks exon 2 (band-c), respectively. The 3 bands are also detected in a panel of mouse cell lines, with HPRT as a loading control (B), and in several normal mouse organs (C). Cisplatin slightly increases the ΔE2 level but decreases the wt level, causing a reciprocal change, in the NMuMG mouse benign mammary epithelial cells and several mouse breast cancer cell lines treated with (+) or without (−) cisplatin (D). Bottom panel: RT-PCR detects 3 bands in human MB231 cells, and sequencing confirms that the top and the bottom bands are the wt CDK4 and the ΔE2, respectively, whereas the middle band is a wt/ΔE2 heterodimer (A). MB231 cells sorted for the CCND1 or the vector were cultured with 5 or 10% serum or were deprived (0) from serum for 2 d. The lower band is the ΔE2, with HPRT as a loading control (B). MCF7 (C), L3.6pL (D), and AsPC-1 (E) cells sorted for CCND1 (D1), its K112E mutant or the vector were cultured with 10% serum or were deprived from serum (0) for 2 d. Expression of the Flag-tagged and the endogenous (Endo) D1 is confirmed. Because the ΔE2 can be detected only with more PCR cycles, sometimes PCR was also run with fewer (26) cycles to ensure the amplification of the wt CDK4 within the linear portion. RT-PCR with F258E1/3 and R1429 primers detects the ΔE2 and a much higher ratio of the ΔE2-to-wt hCDK4 in SKBR3 cells (F).

Although some cyclins such as CCND1 and CCNE1^6,12 have been known to have functions that are independent of their partner CDKs, so far none of the CDK members has been known to function independently of a cyclin or of its kinase activity. In this study we provide, for the first time, evidence showing the existence of such mechanisms for CDK4 in some situations.

Results

CDK4 mRNA may use different start codons

Open reading frame (ORF) analysis reveals that human CDK4 (hCDK4) mRNA has 2 in-frame ATG and 1 CTG translation start codon upstream of the canonical ATG in exon 2 (referred to as ATG1) (Fig. 1). Translation from these upstream start codons should end at one of the 3 stop codons (TAG or TAA) before the ATG1, but reading through these upstream stop codons should add 62, 38, or 30 amino acids (AA), respectively, to the N terminus (Fig. 1). On the other hand, if the translation reads through the 1137–1139th TGA canonical stop codon, it should stop at the 1275–1277th TAA, appending to the C terminus 45 AA, or 46 AA if the TGA is not nonsense but instead encodes an AA, either arginine, cysteine, serine, tryptophan, pyrrolysine, or selenocysteine.^13,14

Similarly, mouse CDK4 (mCDK4) mRNA has an upstream CTG at its 33–35th nt, which initiates an in-frame ORF that ends at the 144–146th TAG (Fig. 1). Reading through this upstream TAG should add 47 AA to the N terminus of the mCDK4 initiated from the canonical ATG (referred to as ATG1) at the 174–176th nt in exon 2 (Fig. 1). On the other hand, if the translation reads through the 1275–1277th TGA canonical stop codon, it should stop at the 1203–1205th TAA, appending to the C terminus 39 AA, or 40 AA if the TGA also encodes an AA.

Both hCDK4 and mCDK4 mRNAs have many in-frame ATGs downstream of the ATG1. If the translation is initiated from one of them, it will produce a CDK4 with N-terminal deletion (Table S1), as seen in the RB1,¹⁵ c-MYC,¹⁶ and many other genes.

There is an exon-2 deleted CDK4 variant

We retrieved expression sequence tags (EST) of CDK4 from the AceView browser (www.aceview.org) of the NCBI and obtained 17 hCDK4 and 7 mCDK4 mRNA variants, besides the wild-type (wt) one (Fig. S1). While some hCDK4 variants are supported by only one EST, others are supported by as many as 17 ESTs. There is a total of 54 hCDK4 and 245 mCDK4 ESTs (Table S2).

Using NCBI Blast (http://blast.ncbi.nlm.nih.gov/) and UCSC Blat (http://genome.ucsc.edu/) browsers to align mRNA with genomic DNA, we identified 2 CDK4 pseudogenes in the mouse, but not in the human. One mouse pseudogene locates at the 1460057–1461349th base-pair (bp) region of the mouse X chromosome, with about 87% identity to the 35–1355th nucleotide (nt) region of the mCDK4. The other is the 8594976–85951085th bp region of chromosome 1 that is reverse-complementarily matched, with about 99% identity, to the 33–1360th nt region of the mCDK4. The pseudogenes are non-coding, as no long ORF is found. RT-PCR failed to detect their expression in major organs of the mouse and in the mouse cell lines we studied. We thus conclude that in both human and mouse, none of the mRNA variants is derived from pseudogene. However, because the mouse pseudogenes triplicate the gene copy number, complete removal of genomic DNA residual from the RNA sample becomes important but more difficult when RT-PCR is used to determine the mCDK4 expression, as we explained recently.¹⁷

Reverse transcription (RT) of the RNA from 67NR mouse breast cancer cells followed by polymerase chain reactions (PCR) with the F109 and R1026 primers (Table S3) yielded 3 bands in agarose gel (Fig. 1). T–A cloning these bands followed by sequencing revealed that the top band (band a) was the wt mCDK4 whereas the bottom band (band c) was a variant lacking the whole 234-bp exon 2 (Fig. 1), coined herein as ΔE2, although the AceView assigned it to the mCDK4-d group which alone was supported by 141 ESTs (Table S2; Fig. S1). Some cloned plasmids from the middle band (band b) were the wt mCDK4, while some others were the ΔE2, indicating that this band was a wt/ΔE2 heterodimer. Such heterodimers formed between 2 similar sequences often occur randomly in PCR, according to our experience.^18-20 The ΔE2 was detectable in all mouse cell lines and major mouse organs we studied, but the level was low, with the brain and testis manifesting the highest ΔE2-to-wt ratio (Fig. 1).

RT-PCR of RNA from MDA-MB231 (MB231) human breast cancer cells with the F86 and R1026 primers also yielded 3 bands in gel (Fig. 1). Cloning and sequencing these bands showed that the top band was the wt, the bottom band was an hCDK4 variant lacking the 237-bp exon 2 (ΔE2), and the middle one was a mixture of the two. The ΔE2, assigned to the hCDK4-d group in AceView (Fig. S1), was also detected in AsPC-1 pancreatic cancer cells and MCF7 breast cancer cells, but it did not seem to be detectable in L3.6pL pancreatic cancer cells (Fig. 1), in SKBR3 breast cancer cells, and in commercial RNA samples from a panel of normal human organs used previously.²⁰ However, RT-PCR with a forward primer (F258E1/3; Table S3), which had its last nt at the exon 3 and thus preferred the ΔE2 to the wt, clearly detected the ΔE2 in SKBR3 cells and gave a high ΔE2-to-wt ratio (Fig. 1). This suggests that the high abundance of the wt competes out the ΔE2 by depriving it of the primers during RT-PCR with the shared primers, leading to a prejudice against the ΔE2. Moreover, the ΔE2 fragment was 237-bp shorter and had fewer total nucleotides, thus was less visible in the gel as we described before.¹⁷

Due to the lack of the ATG1 in exon 2, translation of the ΔE2 in both mouse and human should be initiated from ATG2 at the exon 3 (Fig. 1), producing a 25.9 kD CDK4 protein with the first 74 AA deleted (Fig. S2; Table S1).

There may be multiple CDK4 protein isoforms

Western blot with the sc-260 polyclonal antibody raised against the mCDK4 C terminus (epitope in Table S4) detected the 33-kD wt mCDK4 as the dominant protein in many mouse cell lines as expected (Fig. 2A). However, a band slightly below it was also detected in some mouse cell lines, and sometimes there were some even smaller bands detected (bands in lane 1, below arrows, in Fig. 2A), as reported previously in hamsters.²¹ A band slightly higher than 33 kD and another band around 40 kD were also detected in some mouse cell lines. A tiny amount of the wt mCDK4 or mCDK4-like protein was detected in CDK4^−/− mouse embryonic fibroblasts (MEF; Fig. 2A) and RT-PCR could also detect the mCDK4 mRNA in this MEF line (Fig. 2B). It remains unclear whether a leaky scanning occurs during translation, since in this MEF a reversely oriented Neo cassette was inserted into intron 1,²² but it did not interrupt the ORF initiated from ATG1. Another CDK4^−/− mouse line is available in which the mCDK4 was knocked out with a different strategy,²³ but we were unable to maintain the MEF from this line.

Figure 2. CDK4 protein multiplicity on western blots. (A) Western blots with sc-260 and sc-601 antibodies detect a protein smaller than the wt CDK4 (arrowhead vs. arrow) in several human and mouse cell lines. When a less amount of lysate was loaded, several smaller proteins are also discerned clearly in the M8 mouse cells on the sc-260 blot. (B) RT-PCR with the F314 (in exon 2) and R1026 primers detects a mCDK4 mRNA in the CDK4^−/−MEF, with HPRT as control. (C) 293T cells were cultured with 10% serum or were serum-deprived for 2 d. MB231 cells were untreated (NT) or treated with cisplatin, 5FU or G418. Cell lysates were blotted with four CDK4 antibodies. Loading more lysate from the G418-treated cells allows the 40-kD bands to be discerned clearly. Arrowhead and arrow indicate the wt hCDK4 and a smaller isoform, respectively. (D) Lysates from MCF7 and MB231 cells were treated with alkaline phosphatase (AK) or the corresponding buffer (Con), and then blotted with CDK4 antibodies. Arrows indicate some proteins with changed abundance.

The sc-601 polyclonal antibody raised against the hCDK4 C erminus (Table S4) had a lower affinity to the wt mCDK4, when sc-601 and sc-260 were compared and when human and mouse cell lines on the same blot were compared (Fig. 2A). Conversely, sc-260 could also detect the 33-kD wt hCDK4 and several larger human proteins, but the affinities were lower compared with sc-601 (Fig. 2A). However, sc-601 had a higher affinity than sc-260 to a protein around 40 kD and detected a protein slightly below 33 kD in some mouse cell lines, both of which appeared also in the CDK4^−/−MEF (Fig. 2A).

Western blots using not only the sc-601 and sc-260 but also the DCS35 N-terminal and the DCS156 C-terminal monoclonal antibodies (Table S4) all detected 2 groups of proteins around 40 and 33 kD, respectively, in different human cell lines, with HEK293T (293T) as an example (Fig. 2C). Each of these 2 groups often appeared as a duplet or a triplet of bands but varied among cell lines and among blots with different antibodies. Sometimes, the 33 kD wt hCDK4 was not even the dominant protein (Fig. 2C).

De-phosphorylation of protein lysates from MCF7 and MB231 cells with alkaline phosphatase changed the abundance of some proteins (Fig. 2D), suggesting that different phosphorylation statuses may contribute to the protein multiplicity. Since phosphorylation of CDK2 at some sites results in a faster migration on SDS-PAGE,^24,25 de-phosphorylation of CDK4 at some, but not every, site²⁶ may retard the migration.

LC-MS/MS identifies CDK4 at about 24–28 kD, but not at 40 kD

We ran a larger 10% SDS-PAGE for a longer time of electrophoresis to better separate proteins from HEK293 cells. Western blot of the left part of the gel using sc-601 antibody identified a 40-kD protein, a 33-kD protein, and probably 3 proteins at 24–28 kD (Fig. 3). The excised 3-mm stripe of the gel at 24–28 kD contained 1096 protein groups that were detectable and identifiable by liquid chromatography–mass spectrometry/mass spectrometry (LC-MS/MS). Three hCDK4 peptide fragments were identified, of which one was within the 74-AA region encoded by exon 2 and one at the C terminus, both unique to CDK4, while the third one was in the middle region of hCDK4 fragment and was not unique, as it also appeared in other proteins (Fig. 3). The 40-kD stripe of the gel contained 968 protein groups that were detectable and identifiable, but none of them was matched to hCDK4.

Figure 3. LC-MS/MS results. Lanes 2, 3 and 10 of 10% SDS-PAGE were loaded with pre-stained protein markers of molecular weight, while lanes 1 and 4–9 with cell lysates (100 µg/lane). After electrophoresis, the gel was excised along the thick-dashed line. The left part of the gel was subjected to western blot with the sc-601 CDK4 antibody, which identifies a 40-kD protein, a 33-kD protein, and 3 proteins (arrows) between 24–28 kD. The bands shown on the X-ray film as well as the protein markers on lanes 3 and 10 were used to guide the excision, with a ruler, of the boxed areas in the right part of the gel. Three CDK4 fragments, underlined in the wt hCDK4 sequence, were identified by LC-MS/MS from the 24–28 kD stripe of the gel, of which 2 are unique to CDK4, while one is also contained by other proteins, as shown in the table. Also shown in the table, the 3 fragments together cover about 12% of the 303-AA hCDK4 sequence and receive a general score of about 53 in the LC-MS/MS analysis. The first 74 AAs encoded by exon 2 are italicized in gray.

RNAi may cause reciprocal changes between the 33 and 40 kD proteins

We transfected 293T and Hela cells with 14 hCDK4 siRNA duplexes or shRNA constructs. While some shRNAs did not work efficiently, some others changed the protein abundance in a way that differed between the 2 cell lines, between blots with sc-601 and sc-260 antibodies, and among different proteins detected (Fig. 4). Detected by sc-601 and sc-260, siRNAs decreased the dominant 33-kD wt CDK4 as expected but, in the meantime, increased some other proteins slightly above 33 kD and at 40 kD in Hela cells, together causing a reciprocal change (Fig. 4A). Unexpectedly, the siRNAs did not have obvious effects in 293T cells (Fig. 4B), although, like Hela, 293T could also be transfected at a high efficiency (as shown later in transfections with other constructs). Some shRNAs (such as sh363) actually slightly increased the 33-kD wt CDK4 but decreased a 40-kD protein in Hela cells (Fig. 4C), causing the opposite reciprocity compared with the siRNAs or other shRNAs (Fig. 4A vs. 4C). Again, 293T cells were somewhat refractory to the shRNAs (Fig. 4D). These antibody, cell line, and target sequence specificities, although unexpected, were highly reproducible in many repeats of the transfections.

Figure 4. Effects of siRNAs and shRNAs. Western blot of Hela cells (A) 48 h after transfection with a CDK4 siRNA coctail detects a decrease in the 33-kD wt CDK4 but increases in a protein slightly above 33-kD and another protein at 40-kD, but the siRNAs have no obvious effects in 293T cells (B). Transfection of Hela cells with 3 CDK4 shRNAs (9876, 365 and 363) actually slightly increases the 33-kD wt CDK4, while the shRNA363 also decreases a 40-kD protein when detected with sc-601 antibody (C). Again, 293T cells were somewhat refractory to the shRNAs although sc-260 seems to detect some decrease in the 33-kD wt CDK4 (D). β-actin was used as loading control.

Stop codon suppression redistributes some putative CDK4 isoforms

Treatment of MB231 cells with G418, a suppressor of translation stop codon,^27-29 decreased the abundance of the 40-kD proteins, as detected with sc-601 (lanes 3 vs. 6 in Fig. 2C). Because of the decrease, more protein lysate was loaded into the gel to make it clearer that G418 also redistributed the abundance of the 40-kD proteins, leading to the clear appearance of a triplet, especially when detected with sc-260, although DCS35 and DCS156 could only detect its top and bottom band, respectively (Fig. 2C). The 33-kD wt hCDK4 might only be slightly increased, considering that more lysate was loaded, while the abundance of a band immediately above it was decreased (Fig. 2C). However, treatment of Hela, HEK293, and 293T cells with G418 slightly increased the abundance of the 40-kD proteins (Fig. 5A). A band above the 33-kD wt hCDK4 was also slightly increased in Hela and HEK293 cells as detected by sc-601 and sc-260 (Fig. 5A). The abundance of the 33-kD wt hCDK4 was slightly increased, as detected by most antibodies (Fig. 5A).

Figure 5. Effects of G418 and serum starvation on hCDK4 proteins detected by western blot. The abundance of most CDK4-antibody detected proteins is slightly increased in the Hela, HEK293, and 293T cells treated (+) with G418 for 2 days, compared with the untreated (−) counterpart (A). BxPC3, HPAC, Panc-1, and Panc-28 pancreatic cancer cells express mainly a 33-kD protein, which is decreased by serum withdrawal for 2 days (B). β-actin is used as loading control.

Serum starvation and chemotherapeutics regulate CDK4 expression

Serum starvation did not obviously affect the 33-kD wt hCDK4 in HEK293 and 293T cells although it slightly decreased it in the Hela cells (Fig. 2C; Fig. S3A). However, sc-601, sc-260 and DCS156, but not DCS35, detected a decrease of the wt hCDK4 by serum starvation in MCF7, MB231, AsPC-1, and L3.6pL cell lines (Fig. S3B–E). Actually, the decrease was more evident in the pancreatic cancer cell lines BxPC3, HPAC, Panc-1, and Panc-28 (Fig. 5B). Moreover, serum starvation also changed the abundance of other isoforms detected by some antibodies.

Cisplatin, a chemotherapeutic agent, has been reported to decrease CDK4,³⁰ but in our RT-PCR assay it slightly increased the wt mCDK4 mRNA in NMuMG benign mouse mammary epithelial cells and 5 mouse breast cancer cell lines, with a proportional increase in the ΔE2 level in most of these cell lines (Fig. 2D). Because most of the putative variants in the AceView have deletions, we put more attention on smaller proteins. Treatment of MB231 cells with cisplatin and 5-fluorouracil (5FU), another chemotherapeutic agent, increased the abundance of a band immediately below the 33-kD wt hCDK4 (arrow in Fig. 2C). In contrast, withdrawal of MCF7 cells from serum (Fig. S3B) or treatment of the cells with gemcitabine (Fig. S4), also a chemotherapeutic agent, decreased its abundance, as detected by sc-601 and sc-260. Unexpectedly, however, only gemcitabine decreased the level of the wt hCDK4 protein (Fig. S4).

CCND1 may increase or decrease CDK4 expression

Serum deprival of MCF7 and AsPC-1 cells decreased the mRNA level of the wt hCDK4, but this did not occur in MB231 and L3.6pL cells (Fig. 1). However, ectopic expression of a CCND1 cDNA decreased the mRNA level of the wt CDK4 in all these cell lines in the presence of serum, but, interestingly, CCND1 counteracted the inhibition of the wt CDK4 by serum starvation and even induced its level in L3.6pL and AsPC-1 cells (Fig. 1). The ΔE2 levels remained low in MCF7, MB231, and AsPC-1 cells and did not obviously respond to serum starvation or ectopic CCND1 (Fig. 1). The K112E mutation of CCND1, which abolishes the ability to activate CDK4 or CDK6,⁶ attenuated the effect of CCND1 on CDK4 expression, although the attenuation varied among cell lines (Fig. 1).

At the protein level, CCND1 also counteracted the decrease of the wt CDK4 by serum starvation in MCF7, AsPC-1, and L3.6pL cells, and even slightly induced it (Fig. S3B, D, and E). However, this counteraction was not discerned in MB231 cells (Fig. S3C) and was less evident for the K112E mutant. Some of the putative isoforms detected by different antibodies were also changed by CCND1, but the changes varied among cell lines (Fig. S3B–E).

The ΔE2 loses the CCND1- and RB1-binding ability

Lack of exon 2 leads to the deletion of the ATP binding sequence and the PSTAIRE domain at the N terminus (Fig. 1). Bioinformatic analysis suggests that this deleted region also contains 7 potential phosphorylation sites, i.e., the serines at the 4th, 28^th, and 52nd AA, the threonines at the 3rd, 19^th, and 53rd AA, and the tyrosine at 6th AA of hCDK4 (Table S5). Also lost may include one potential trans-membrane helix (Table S6) and the ability to bind to most CDK4 partner proteins, although it may retain some ability to interact with RB1 (Fig. S5). The tertiary structure might be changed as well (Fig. S6).

Transfection of 293T cells with our wt CDK4-HA (wt-HA) or ΔE2-HA construct followed by western blot with the sc-2937 HA antibody detected both ectopic proteins (Fig. 6A). Interestingly, although the ΔE2-HA protein was calculated as 29.6 kD (Fig. S2), it migrated at a similar position as the 37.4-kD wt-HA, for which the reason is unknown, but loss of some phosphorylation sites may be one. Western blots with another HA antibody (sc-7392) and a third one (sc-7392-HRP) tagged with a horse radish peroxidase also yielded the same results (not shown). Notably, a duplet of endogenous proteins around 33 kD was also detected by all 3 HA antibodies, as seen in non-transfected and vector-transfected 293T cells (Fig. 6A). The DCS156 CDK4 antibody could detect the wt-HA and the ΔE2-HA proteins in the 293T transfectants as well, although the affinity was lower than that to the endogenous wt hCDK4 (Fig. 6B).

Figure 6. Co-IP detection of CDK4 binding to CCND1 and RB1. (A) Lysates from the wt-HA, ΔE2-HA or vector transfected 293T cells were IP with 718 CCND1 antibody and then blotted with sc-2937 HA antibody. The wt-HA, but barely the ΔE2-HA, is detected. The HA antibody also detects a nonspecific protein or protein duplet (arrowhead) in non-IP lysates of transfected and untransfected cells. A larger amount of RB1 protein is detected in the CCND1 immunoprecipitate from the wt-HA transfectant than from the ΔE2 and the vector counterparts. (B) DCS156 CDK4 antibody detects the wt-HA and the ΔE2-HA proteins in the tranfectants, besides the endogenous wt CDK4 and some smaller isoforms (arrow) in untransfected (UnT) 293T and MB231 cells. (C) The sc-7392 HA antibody detects multiple HA-tagged proteins in the 718 and the DCS6 CCND1 immunoprecipitates from the wt-HA transfected 293T cells. A/G: protein A/G as an IP control. HC, heavy chain of mouse immunoglobulin (IG) detected because sc-7392 is also a mouse origin. (D) DCS35 CDK4 antibody detects some proteins (arrows) larger or smaller than the 33-kD wt CDK4, including one close to 40-kD, in sc260 and sc-601 CDK4 immunoprecipitates from MCF7, but not 293T, cells. PG, protein G; HC and LC, heavy chain and light chain of IG, respectively. (E) The sc-601 CDK4 antibody detects the 33-kD wt CDK4 and the 40-kD duplet in 293T cells included as control. sc-601 also detects the wt CDK4 in the D1–718 and the D1-DCS6 immunoprecipitates of construct transfectants. However, it detects mainly the bottom band of the 40-kD duplet and in only the D1–718, but not the DCS6, precipitate. m-IG-A/G: mouse IG and protein A/G control. (F) sc-601 detects the top band of the 40-kD duplet in the 718 CCND1 precipitates from MCF7, M231 and 293T cells. However, in the D1-DCS6 precipitates, the top band was detected only from MCF7 and MB231 cells, but not from 293T cells. Both sc-601and DCS35 CDK4 antibodies also detect some proteins smaller than the 33-kD wt CDK4 in both immunoprecipitates.

Immunoprecipitation (IP) of cell lysates from both transfectants with the sc-718 polyclonal CCND1 antibody followed by blotting with the sc-2937 HA detected the wt-HA protein, but barely detected the ΔE2-HA (Fig. 6C). Co-IP assay with the DCS6 monoclonal CCND1 antibody for IP and the sc-7392 HA for blotting yielded similar results (Fig. 6C). These results confirm that in contrast to its wt counterpart, the ΔE2-HA protein has basically lost the ability to bind to CCND1.

Similar to the endogenous CDK4, the wt-HA proteins pulled down by the two CCND1 antibodies were expressed in multiple forms (Fig. 6C). Moreover, the RB1 protein in the sc-718 CCND1 precipitates was more abundant in the wt-HA transfectants than in the vector and the ΔE2-HA counterparts, but the RB1 levels were similar in the ΔE2-HA and the vector transfectants (Fig. 6A), suggesting that the ΔE2-HA protein could barely bind to RB1.

The 40-kD proteins may be associated with CCND1 in some situations

IP of MCF7 and 293T cell lysates with sc-260 or sc-601 CDK4 antibody followed by blotting with the DCS35 CDK4 antibody detected the 33-kD proteins as expected (Fig. 6D). Interestingly, the dominant wt CDK4 was the top band of the 33-kD duplet in MCF7 cells but the bottom band in 293T cells (Fig. 6D). Several proteins larger than 33-kD in MCF7 cells were also detected by DCS35 in the 2 CDK4 precipitates (arrows in Fig. 6D), indicating that they are reactive to sc-260 and sc-601 in the native configuration as well.

The sc-601 and DCS35 CDK4 antibodies could detect the endogenous wt CDK4 protein in both sc-718 and DCS6 CCND1 immunoprecipitates as expected (Fig. 6E and F). The top-band of the 40-kD duplet was also detected with both sc-601 and DCS35 in these 2 CCND1 immunoprecipitates from MCF7 and MB231 cells (Fig. 6F). In 293T cells, however, the 40-kD duplet, mainly its bottom band, could be detected only in the D1–718 immunoprecipitate, but not in the monoclonal D1-DCS6 precipitate (Fig. 6E and F). Why the 40-kD proteins/CCND1 association can only be detected in some cell lines with some specific antibodies remains unclear, but it may be related to different posttranslational modifications in different cells, since an earlier study even failed to detect the wt CDK4 in the DCS6 immunoprecipitate.³¹

Both wt and ΔE2 act also at the S–G₂/M to regulate cell growth and death

MTT assay showed that the wt-HA increased the number of viable HEK293 cells 2 to 4 d post-transfection as expected (Fig. 7), but this ability of the ΔE2-HA was decreased, since it was often insignificant (Fig. 7 and data not shown). Unexpectedly, however, both the wt and the ΔE2 slightly but significantly decreased the number of viable Hela cells at late time points (Fig. 7). The number of viable 293T cells was not affected by the wt-HA but was slightly decreased by the ΔE2 (Fig. 7).

Figure 7. Effect of the wt-HA and ΔE2-HA on the viability of HEK293, Hela and 293T cells at 48, 72, or 96 h after transfection, determined by MTT assay and presented as % of the empty vector. (A) significantly different from vector cells (P < 0.05; t test); (B) significantly different from both the vector and the ΔE2 cells (P < 0.05; t test).

To determine at which cell cycle stage CDK4 acts, we arrested the transfected HEK293 cells at G₀-G₁ phases by serum withdrawal for 48 h. Two hours after release by serum replenishment, the wt and the ΔE2 transfectants already had a higher G₂/M fraction, while a larger number of vector-transfected cells were still at the S phase (Table 1). However, although CDK4 is supposed to drive G₁ progression, to our surprise the G₁ fraction was not smaller in the wt than in the ΔE2 transfectants and, more surprisingly, was slightly larger in these cells than in the vector counterpart, while the apoptotic rate reflected by the sub-G₁ fraction was similar between the 2 construct transfectants but was much lower in these cells than in the vector transfectant (Table 1). As expected, serum replenishment corrected some of these changes hours later (Table 1). We conclude that both wt and ΔE2 can decrease apoptosis and drive S–G₂/M progression, but unexpectedly delay G₁-S progression of HEK293 cells (Fig. 8).

Table 1. Effects of CDK4 variants on cell cycle distribution.

Cell line	Time	Construct	%G1	%G2-M	%S	%Sub-G1
HEK293		Vector	35	3.7	61.2	13.4
	2h	wt CDK4	40.5	8.5	51	2.8
		ΔE2	40.9	10	49.1	3.8
		Vector	39.5	4	56.5	23.7
	6h	wt CDK4	35.2	10.2	54.7	3.9
		ΔE2	37.2	11.7	51.2	4.4
Hela		Vector	52.4	11.6	36	22.9
	2h	wt CDK4	62.5	12	25.5	29.9
		ΔE2	59.3	10.3	30.4	33.3
		Vector	48	10.1	41.9	26.7
	6h	wt CDK4	58.4	12.4	29.3	30.7
		ΔE2	56.6	13.9	29.6	27.7
293T		Vector	37.3	13.4	49.3	6.2
	2h	wt CDK4	35.3	21.3	43.4	12.9
		ΔE2	35.4	15.5	49.1	13
		Vector	11.1	10.3	78.6	10.4
	4h	wt CDK4	2.8	16.4	80.8	9.9
		ΔE2	3.9	10.8	85.3	8.7

Figure 8. Illustration of hypothesis on the novel functions of the wt CDK4 and the ΔE2. In a situation of insufficient growth stimuli such as in serum starvation, both the wt CDK4 and the ΔE2 may behave differently from the normal situation in a cell-specific manner: In cells that are non-malignant or are malignant but have functional RB1, such as in HEK293 cells, the wt CDK4 and ΔE2 are similarly potent in accelerating the S-to-G₂/M but in inhibiting both G₁-to-S progression and apoptosis. Also when lacking sufficient growth stimuli, in those cells that are malignant and/or have inactivated RB1, such as in 293T or Hela cells, both CDK4 isoforms not only delay the G₁-to-S progression but also enhance cell death. However, the wt CDK4 promotes cell death with no effect on S-to-G₂/M, whereas E2 promotes S-to-G₂ progression with no effect on cell death. Moreover, in a situation wherein CDK4 activity is inhibited by such as NPCD, both isoforms not only accelerate the G₁-to-S and the S-to-G₂/M progression but also inhibit apoptosis of the cells that are normal or have functional RB1. However, they delay the G₁-to-S and promote the S-to-G₂/M progression, without affecting the viability, of those cells that are malignant or are RB1-inactive. Whether it is the functional status of RB1 or the malignant status of the cell that is a determinant for these novel functions and mechanisms remains to be explored.

At 2 h post-serum replenishment, both wt- and ΔE2-transfected Hela cells manifested a decreased S fraction but increased G₁ and sub-G₁ fractions without a change in the G₂/M, compared with the vector transfectant, but the higher apoptotic rate at the sub-G₁ vanished hours later (Table 1). These results suggest that both CDK4 constructs not only hinder the serum-starved Hela cells from progressing into the S phase, but also enhance the starvation-induced apoptosis, both probably contributing to the decreased viability seen in MTT assay (Fig. 8).

Unlike HEK293 and Hela cells, 293T did not manifest the G₁ arrest by the 2 CDK4 constructs at 2 h post-release, and accelerated S-G₂/M progression was observed only in the wt-HA transfectant, but both constructs showed a similar enhancement of the apoptotic rate at the sub-G₁ (Table 1). After another 2 h, both constructs started to accelerate G₁ progression, resulting in a lower G₁ fraction, while the enhancement of apoptosis quickly vanished, but the acceleration of the S-G₂/M progression by the wt CDK4 was sustained compared with the vector transfectant (Table 1). Hence, a quicker cell turnover caused by the wt-HA may be a reason why MTT assay did not detect a change in 293T viability, whereas increased apoptosis in the absence of accelerated proliferation may explain a decreased viability seen in the MTT assay of the ΔE2 expressing 293T cells.

Like its wt, the ΔE2 also affects the response to CDK4/6 inhibitor

Treatment of HEK293 cells with NPCD, a CDK4 and CDK6 inhibitor,^20,32-34 significantly increased apoptosis at the sub-G₁ fraction as expected. However, both the wt- and the ΔE2- transfected HEK293 cells manifested a much lower apoptotic rate in association with increased G₂/M and decreased G₁ fractions (Table 2). We conclude that the 2 constructs not only drive HEK293 proliferation by accelerating the G₁-to-S and S-to-G₂/M progression but also render the cells resistant to NPCD-induced cell death. However, in the same situation, both constructs arrested Hela cells at the G₁ but accelerated the S-to-G₂/M progression, collectively resulting in a decreased S fraction (Table 2). Thus, although the effects of the 2 constructs are cell line specific, in both cell lines the ΔE2 has effects similar to those of the wt CDK4 in response to NPCD, suggesting that either these effects of CDK4 are irrelevant to its kinase activity or NPCD also acts independently of its inhibition of CDK4 kinase activity.

Table 2. CDK4 effects on the response to NPCD.

Cell line	Construct	%G1	%G2-M		%S	%Sub-G1
	Vector	35.9	7.8	56.3		31.8
HEK293	wt CDK4	27.7	12.2	59.7		10.2
	ΔE2	29.1	12.6	58.3		10.9
	Vector	33.7	5.7	60.6		42.7
Hela	wt CDK4	67.8	13.3	18.9		42.5
	ΔE2	66.8	15.9	17.3		46.6

Discussion

There are many transcriptional, posttranscriptional, and translational mechanisms that allow an individual gene to yield multiple proteins, although it is currently unclear whether protein splicing is also a mechanism in multi-cell organisms.^35,36 Many types of post-translational modifications, including less discussed S-nitrosylation³⁷ and many non-enzymatic reactions³⁸ can also affect protein mobility in SDS-PAGE. Indeed, so often western blots detect multiple bands, although most publications only present the anticipated band without really proving that the unexpected bands are, in fact, artifacts. We prefer to present all bands in case some of them are later proved to be genuine.

CDK4 may have multiple isoforms

All 4 CDK4 antibodies we used can detect multiple bands around 33–40 kD on immunoblots of both human and mouse cells, in part due to different phosphorylation statuses, since de-phosphorylation by alkaline phosphatase can alter the abundance of some of them. Actually, a few other antibodies also result in similar data (not shown). Many published studies also show CDK4 as a duplet^22,39 or as multiple bands,²¹ sometimes even expressed from a CDK4-HA construct,²⁶ although no comments are given. Our LC-MS/MS analysis confirms that some proteins below 33 kD are CDK4 isoforms, which likely lack part of the sequence. Many CDK4 ESTs also show sequence deletion. However, since both N- and C-terminal fragments are identified between 24–28 kD, there may be several smaller isoforms that lack different parts of the sequence, but whether the ΔE2 is one of them requires more studies.

Of the 40-kD duplet, the top band can be detected in 2 CCND1 precipitates from MCF7 and MB231 cells (Fig. 6E and F), although in 293T cells, mainly the bottom band in only the D1–718 immunoprecipitate is detected (Fig. 6F). The results suggest that 40-kD proteins/CCND1 association may be detected in some cell lines with some antibodies. However, LC-MS/MS fails to identify CDK4 in the 40-kD stripe of the gel, explanations for which include (1) the 40-kD proteins are nonspecific; (2) the CDK4 proteins were missed in the 2-mm gel stripe; (3) the abundance of the CDK4 proteins is insufficient for LC-MS/MS detection, although it is sufficient for western blot; or (4) they degrade much quicker than other proteins during the LC-MS/MS procedure. If the 40-kD proteins are not CDK4, their binding to multiple CDK4 antibodies in western blot and IP assays as well as their possible binding to CCND1 in some situations suggest their similarity in the structure to CDK4.

Some of the 17 CDK4 mRNA variants in AceView may contribute to the protein multiplicity. In addition, translation from an upstream start codon^40,41 by reading through the upstream stop codons to extend the N terminus may also be a mechanism, as seen in many other genes.⁴² Reading through the canonical stop codon can also extend the C terminus. At least in yeast, stress can increase the use of upstream start codons or downstream stop codons.⁴³ We observe that a stop codon suppressor G418 can redistribute the abundance of some bands detected, although whether it is really the N-terminal or C-terminal extension or both that yield the sluggish bands remains unknown, in part because N- and C-terminal extensions add similar numbers of AA to the protein. On the other hand, those proteins smaller than 33-kD may result from using a downstream start codon (Table S1) via many mechanisms,⁴⁴ including leaky scanning, reinitiation, internal ribosome entry site (IRES),^40,41 and cap-independent translational enhancers,⁴⁵ as we discussed before.^18,46 Some other mechanisms that do not even involve 5′-to-3′ scanning may also be possible.⁴⁷ All these mechanisms can be affected by many factors such as serum starvation, while different antibodies have different affinities to different protein isoforms, which collectively hinder us from characterizing so many bands on immunoblots.

Not all shRNAs can degrade CDK4 mRNA

In Hela cells, some shRNAs decrease the 33-kD wt CDK4 but increase some 40-kD proteins, whereas some other shRNAs have the opposite effects. These opposite reciprocities suggest an interrelationship between the 33-kD wt CDK4 and some 40-kD proteins, but the underlying mechanisms remain unknown. The antisense strand of the CDK4 gene encodes another gene called TSPAN31. The mRNAs of these 2 genes have a sense-antisense relationship and overlap at their last 517 nt, as we outlined recently.⁴⁸ A question thus arises as to how the cell knows whether it is the sense or the antisense RNA that should be knocked down when we force the cell to express a double-stranded regulatory RNA, such as an siRNA, shRNA, or microRNA. This question is of importance, because peers, when using RNAi technology, often select those shRNAs or siRNAs that work “fine” without asking why many others do not work. Over 63% of RNA transcripts, including the CCND1 mRNA,⁴⁸ from the human genome are accompanied by antisense counterparts,^49,50 and the Univergene database of the NCBI has over 123 000 human antisense entries.⁵¹ Almost the whole human genome is transcribed, but all exons that encode proteins sum up to only 1.5% of the genomic sequence, with the hefty majority of the remaining transcripts mainly acting as regulatory RNAs.^17,52 It is now known that the sense strand of a shRNA and the miRNA* strand of a microRNA duplex are not useless as thought previously, but, instead, play regulatory roles.^53-55 In a situation wherein both Watson and Crick strands of DNA double helix encode genes and are transcribed, cells probably have intrinsic mechanisms, such as a different way of strand-sorting,^53,55 to prevent degradation of the reverse-complementary RNA by some regulatory RNAs. Certain regions of an mRNA sequence may be regulatory to other RNAs and, in the meantime, are protected from being targeted by some others. Moreover, expression of the 2 oppositely oriented genes may be under different controls, and the 2 genes may regulate each other positively or negatively. Any of these scenarios may explain why different siRNAs or shRNAs cause opposite reciprocities and why 293T responds differently from Hela to the RNAi we used. Therefore, selecting a favorable siRNA or shRNA may, sometimes, be picking up a bias.

CDK4 may have functions independent of binding to CCND and RB

The ΔE2 protein lacks not only the ATP binding sequence, but also the PISTVRE domain required for binding to the p16 and CCND. Co-IP assay confirms that it has lost the CCND1- and RB1-binding ability. However, it can still drive the S-G₂/M progression and, more surprisingly, is as potent as its wt in enhancing serum starvation-related apoptosis of 293T and Hela cells but inhibiting the death of HEK293 cells. It is also as potent as its wt in arresting the Hela, but not the HEK293, cells at the G₁ when treated with NPCD. Hence, no matter whether ΔE2 protein is naturally expressed or not, our results from the ΔE2 construct suggest that CDK4 may use some CCND- and RB-irrelevant mechanisms, not only in the well-known promotion of cell proliferation, but also in the unexpected inhibition of cell cycle progression and induction of cell death. Actually, a previous study on B-cell lymphoma has already indirectly suggested the possible existence of cyclin-independent functions of CDK4.^56,57 Warenius et al. have also reported that the hexapeptide (PRGPRP) at the 251–256th AA has killing effects on cancer cells.⁵⁸ Probably the growth arrest and apoptotic effects of the ΔE2 and the wt may be attributed to this C-terminal region. This novel function may not be unique to CDK4, because CDK8, another canonical growth-promoting CDK, has also be shown to be tumor suppressive in some situations,⁵⁹ which is generally congruent with our opinion that many genes actually have dual-functions.⁶⁰ Since our data suggest that these novel functions may occur at the S-G₂/M phases (Fig. 8), we probably should now put more attention to some earlier observations of CDK4 activities outside the G₁.^1-4

It needs to be mentioned that so far the novel functions and mechanisms are observed only in such situations wherein the cells are serum-starved and replenished or wherein the cells were treated with a CDK4 inhibitor (Fig. 8). Another caveat is that independence of CCND- and RB-binding does not mean without influence from these proteins. RB1 is functional in HEK293 cells but is inactivated in 293T cells by the SV40 T large antigen and in Hela cells by the E7 protein of HPV18 virus.⁶¹ Therefore, the opposite effect of the ΔE2 on apoptosis in these 2 sets of cell lines (Table 1) may be indirectly related to their difference in the functional status of the RB1, although a difference between a benign and a malignant status may be another explanation (Fig. 8), since 293T should be considered a malignantly transformed version of HEK293 as it can form colonies in soft agar and develop xenograft tumors in mice.^62,63

CCND1 may also inhibit CDK4 expression

Serum deprival is known to cause G₁ arrest. As shown herein, downregulation of CDK4 may be a major mechanism in some cells, such as in several pancreatic cancer cell lines (Fig. 5B). Some cell lines such as 293T and Hela are relatively refractory to serum starvation and do not manifest obvious change in CDK4 (Fig. 2C; Fig. S3A), probably because their RB1 is inactivated.⁶¹ CCND1 is known to act as a sensor for cells to receive extracellular growth stimuli, which, as we observed, may be exerted in part via sustaining the CDK4 level in a situation of insufficient growth stimuli, such as a low serum concentration, that mimics a poor blood supply in vivo. Unexpectedly, however, we also observe that CCND1 slightly decreases CDK4 mRNA and protein levels when the cells are cultured with sufficient serum. Thus, CCND1 may hold the rein of CDK4, preventing it from going either too low or too high, probably because a high CDK4 level may cause apoptosis as observed herein. This may also explain why sometimes CCND1 only changes the mRNA, but not the protein, level of CDK4, because a slight change of the mRNA level can already adjust the protein level. Mechanistically, the regulation by CCND1 seems to involve CCND1–CDK4 kinase activity, because K112E mutation attenuates the effects. This previously unrealized mechanism allows CDK4 to regulate itself.

Many chemotherapeutic agents like those used herein are known to inhibit cancer cell growth. We are surprised that these agents, except gemcitabine, do not decrease the wt CDK4, unlike serum starvation. Cisplatin even slightly induces it. Likely, CDK4 responds more sensitively to endogenous factors than to exogenous ones.

Conclusion

Using multiple antibodies in western blot, IP and Co-IP assays, we show that CDK4 may be expressed as multiple protein isoforms that have different affinities to different antibodies. LC-MS/MS analysis confirms the existence of some smaller CDK4 isoforms. We propose that multiple antibodies should be used in future studies of CDK4. We have cloned a ΔE2 CDK4 mRNA variant that encodes a protein lacking the ATP binding sequence and the PISTVRE domain at the N terminus, thus lacking the ability to bind to CCND1 and to phosphorylate RB. We observed surprisingly that the wt CDK4 and the ΔE2 could delay the G₁–S progression, accelerate the S–G₂/M progression, and enhance or inhibit apoptosis in a cell line-specific manner when the cells were treated with a CDK4 inhibitor or the cells were serum-deprived and then replenished. These results suggest that CDK4 has some novel functions in S–G₂/M phases in some situations, besides the canonical promotion of G₁ progression in routine cell culture, in part via mechanisms independent of binding to CCND and phosphorylating RB.

Materials and Methods

Cell lines and cell culture

Cell lines are detailed before²⁰ or in the literature. The CDK4^−/− MEF was provided by Dr Chenguang Wang at Thomas Jefferson University. For starvation studies, cells in duplet dishes were first cultured with DMEM containing 5% serum. When the cells reached about 60% confluence, the medium for one dish was replaced with a serum-free DMEM while the other with a 10% serum containing DMEM, followed by another 2-d culture.

cDNA constructs

A mCDK4 cDNA provided by Dr Chenguang Wang was cloned into a pMIG (MSCV IRES GFP) bicistronic retroviral vector with an IRES driven GFP (green fluorescent protein) as the second cistron. This cDNA, referred to as the wt mCDK4, lacks the 5′ untranslational region (5′UTR) before the ATG1 in exon 2, and its canonical TGA stop codon along with the whole 3′ UTR is replaced with an HA-tag sequence (Fig. S2). We deleted the ATG1 along with its downstream 55 nt to make the ΔE2-HA construct (Fig. 1). A human CCND1 cDNA and its K112E mutant, both containing an N-terminal Flag-tag and cloned into the pMIG vector, were provided also by Chenguang Wang.

CDK4 siRNA and shRNA

A cocktail of 4 CDK4 siRNA oligos (Cat # sc-29261) and a corresponding scrambled siRNA were purchased from Santa Cruz Biotech (www.scbt.com). A CDK4 shRNA plasmid used by Navarro et al.⁶⁴ was purchased from Addgene (www.addgene.org/25788). Another CDK4 shRNA plasmid was provided by Dr Greene.⁶⁵ Moreover, all eight CDK4 shRNA clones in the TRCN library (TRCN00000-00362, -00363, -00364, -00365, -10472, -18364, -09876, and -10520) that were claimed to having been validated, along with a control scrambled shRNA, were purchased from University of Minnesota Core Facility. All CDK4 target sequences are listed in Table S7.

Transfection, infection, and cell sorting

Transfection or infection of DNA or RNA and the ensuing fluorescent assistant cell sorting (FACS) for GFP were performed as described previously.²⁰ The FACS sorting resulted in a pure population of construct-expressing cells.

RT-PCR and T-A cloning

Total RNA was isolated with Trizol followed by DNase I digestion and then DNase inactivation as described previously.^18-20 RT was performed with random hexamers.^18,19 PCR was performed with primers listed in Table S3, using HPRT as the reference gene, as explained before.¹⁷ PCR products were separated in agarose gel, visualized by ethidium bromide staining and purified with the UltraClean Gel DNA Extraction Kit (Cat # G-1210–0300, IscBioExpress; www.bioexpress.com) or with a simple method we have described before.⁶⁶ T–A cloning was described before.^18,19

Western blot and Co-IP

As described before,^21,67 cell lysates were prepared with inhibitors of proteases, kinases and phosphatases. Proteins were fractioned in 8–12% SDS-PAGE and transferred onto a PVDF membrane. All primary antibodies were purchased from Santa Cruz Biotech and detailed in Table S2. For IP,²¹ 400 µg proteins were diluted in a volume of 400 µl lysate buffer and incubated with 0.2 µg indicated primary antibody at 4 °C for 4 h with rotation. Protein A/G-conjugated agarose beads (15 µl; from Santa Cruz Biotech) were then added, followed by another 4-h rotation. The complex of antigen-antibody-beads was quickly spun down and washed 4 times with cold lysate buffer. Proper controls included an aliquot of normal immunoglobulins from the same species (Table S4) to replace the primary antibody in the IP reaction. The pellets were suspended in 30 µl 2 × SDS reducing loading buffer and boiled for 4 min. The immunoprecipitates were used in western blot to detect desired proteins with indicated primary antibody.

LC-MS/MS

To better separate proteins, we ran, for a longer time of electrophoresis, a larger (10 × 8 cm) 10% SDS-PAGE loaded with HEK293 cell lysate in 2 panels (Fig. 3). The left panel of the gel was used in western blot, while the right panel was stained with Coomassie blue and then destained to visualize the proteins. A 2-mm stripe of the gel at 40-kD and a 3-mm stripe around 24–28 kD were excised out, which was performed with a ruler and guided by the protein markers and the corresponding bands on the X-ray film from the western blot to carefully avoid contamination from the 33-kD CDK4 (Fig. 3). The excised gel stripes were dehydrated, reduced, alkylated, and digested by trypsin as described in detail before.⁶⁸ The digested products were delivered to Q-Exactive mass spectrometry (Thermo Scientific). The acquired MS/MS data were searched using Mascot v2.3.01 in local server against human SwissProt_new xyzzy database (20255sequences). All searches were performed with tryptic specificity, allowing one missed cleavage. Oxidation (M) and Gln- > pyro-Glu (N-term Q) were considered as variable modifications. The mass tolerances for MS and MS/MS were 15 ppm and 20 mmu, respectively.

Treatment with chemotherapeutic agents, NPCD and G418

Desired cell lines cultured with DMEM containing 5–10% serum were treated with indicated chemotherapeutic agents or NPCD, as described before.^33,69,70 Saline was used as the non-treated control but for NPCD the solvent (DMSO) was used.³³ Cells were harvested at the indicated time points, followed by RT-PCR, western blot or MTT (3-[4, 5-Dimethylthiazol-2-yl]-2, 5-diphenyltetrazolium bromide) assays.^20,33 Desired cells were also treated with 100 µg/ml G418 dissolved in DMSO, using DMSO as control, followed by cell harvest 2 d later for western blot.

Cell cycle and cell death analyses

FACS was used as described previously.³³

In vitro de-phosphorylation assay

An aliquot of protein lysates from indicated cell lines was adjusted to pH8.5 with Tris-HCl buffer. A total of 250 μg proteins were then treated for 1 h at 37 °C with or without (as control) calf alkaline phosphatase (325 U, Sigma-Aldrich, Inc) as we described previously.⁴⁶

In silico analyses of protein structure and functions

Different software packages were used for these analyses as described in the corresponding supplementary figures and tables.

Statistical analyses

MTT assays were analyzed using t test. Cell cycle distributions were repeated 6 times and all showed the same trend. Because the absolute figures varied greatly among these experiments, especially in the serum replenishment studies, a representative experiment was presented.

10.4161/cc.26510

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Supplemental Materials

Supplemental materials may be found here: www.landesbioscience.com/journals/cc/article/26510