Critical reanalysis of the methods that discriminate the activity of CDK2 from CDK1

ABSTRACT

Cyclin dependent kinases 1 and 2 (CDK1 and CDK2) play crucial roles in regulating cell cycle progression from G1 to S, through S, and G2 to M phase. Both inhibition and aberrant activation of CDK1/2 can be detrimental to cancer cell growth. However, the tools routinely employed to discriminate between the activities of these 2 kinases do not have the selectivity commonly attributed to them. Activation of these kinases is often assayed as a decrease of the inhibitory tyrosine-15 phosphorylation, yet the antibodies used cannot discriminate between phosphorylated CDK1 and CDK2. Inhibitors of these kinases, while partially selective against purified kinases, may lack selectivity when applied to intact cells. High levels of cyclin E are often considered a marker of increased CDK2 activity, yet active CDK2 targets cyclin E for degradation, hence high levels usually reflect inactive CDK2. Finally, inhibition of CDK2 does not arrest cells in S phase suggesting CDK2 is not required for S phase progression. Furthermore, activation of CDK2 in S phase can rapidly induce DNA double-strand breaks in some cell lines. The misunderstandings associated with the use of these tools has led to misinterpretation of results. In this review, we highlight these challenges in the field.

Introduction

Cyclin-dependent kinases (CDKs) are a family of serine/threonine kinases whose sequential activation and inactivation ensures unidirectional progression through the cell cycle. CDK activity depends on association with a particular cyclin, whose expression oscillates at an appropriate time throughout the cell cycle, and on various post-translational modifications, resulting in phosphorylation of a myriad of substrates to regulate cell cycle progression. Passage through the G1 restriction point relies primarily on CDK4/6 and their association with cyclin D. Entry into S phase requires CDK2 and its association with cyclin E. In S phase, CDK2 dissociates from cyclin E and binds to cyclin A and phosphorylates a different set of substrates. Cyclin A also binds CDK1 in G2 phase, while the association of CDK1 with cyclin B is the primary driver for entry into mitosis. Completion of mitosis requires the degradation of cyclin B.¹

CDKs are targets of interest for anticancer drug development as uncontrolled activation of CDKs can accelerate tumor proliferation and enhance chromosomal instability.¹ Many studies have sought selective and effective inhibitors of CDKs, with inhibitors of CDK4/6 having recently been approved by the FDA.^2-5 In contrast to inhibition, we and others have recently shown that aberrant and uncontrolled activation of CDK2 and CDK1 can also be detrimental to cancer cells.^6-9

Our recent studies were designed to determine why some cancer cell lines are hypersensitive to inhibition of Checkpoint kinase 1 (Chk1).⁶ Chk1 phosphorylates and inactivates the CDC25 phosphatases, thereby preventing their ability to dephosphorylate and activate CDK1 and CDK2. In a subset of cell lines, inhibition of Chk1 resulted in rapid activation of CDC25A, phosphorylation of histone H2AX (the phosphorylated form is known as γH2AX), and DNA double-strand breaks in S phase cells, but whether sensitivity was due to activation of CDK1 or CDK2 became a challenge. Our studies identified many concerns for the tools commonly used to discriminate the activity of CDK1 from CDK2.⁶ These concerns are discussed here.

Phosphotyrosine-specific antibodies do not discriminate CDK2 from CDK1

In addition to binding cyclins, each CDK is modified by phosphorylation. Wee1 and Myt1 kinases inactivate CDK1/2 by phosphorylating them on the inhibitory sites, tyrosine 15 (Y15) and threonine 14 (T14) respectively.^10,11 Activation of these CDKs results from dephosphorylation at these sites by a member of the CDC25 family of phosphatases (CDC25A, B and C). Consequently, the activation of CDK1 and/or CDK2 is frequently assessed by the loss of this inhibitory phosphate on Y15.^6,12–25 Unfortunately, the commonly used antibodies cannot discriminate between phosphorylated CDK1 and CDK2 because the tyrosine phosphorylation site resides in the middle of a 13 amino acid conserved sequence (Table 1). Furthermore, this sequence is also conserved in the rarely studied CDK3. The related sequence in CDK5 differs by only 2 amino acids, whereas 2 other related kinases, CDK8 and CDK19, have 4 differences over this region.

Table 1.

Similarity of the conserved sequence within different members of the CDK family.

Cyclin-dependent Kinase	Molecular weight	N-terminal sequence (mismatched bases shown in lower case in the 13 amino acid conserved sequence)
CDK1	34 kDa	MEDYTKI EKIGEGTYGVVYK GRHKT
CDK2	34 kDa	MENFQKV EKIGEGTYGVVYK ARNKL
CDK3	34 kDa	MDMFQKV EKIGEGTYGVVYK AKNRE
CDK5	35 kDa	MQKYEKL EKIGEGTYGtVfK AKNRE
CDK8	53 kDa	M+23aa cKvGrGTYGhVYK AKRKD
CDK19	57 kDa	M+23aa cKvGrGTYGhVYK ARRKD

“23aa” reflects the additional 23 amino acids between the start methionine and the conserved sequence.

This problem for the lack of selectivity of the antibodies is perpetuated by many companies who advertise their antibodies as being specific to phosphorylated CDK1 or CDK2 (Table 2). In a few cases cases, their product data sheet does mention potential cross-reactivity (Table 2, Antibodies 1–5), but it seems this information is ignored by many investigators. Furthermore, several papers have used 2 different antibodies, each purported to be selective for either CDK1 or CDK2, yet obtained identical data.^12,13,26

Table 2.

Commercially available antibodies targeting phosphotyrosine (Y15) on CDK1 and CDK2. The first 5 antibodies are noted as having cross reactivity; the subsequent antibodies (sorted as CDK2, then CDK1/cdc2) provide no information on cross reactivity.

	Antibody descriptor	Company	Catalog #	Additional comments from data sheets
1	Anti-Cdk2 Y15	Abcam	ab76146	Sequence analysis shows that the immunogen of ab76146 shares 100% identity with human CDK1, thus, would predict cross-reactivity
2	Phospho-Cdc2 p34 Y15/T14	Santa Cruz Biotech	sc-12340	Ab also detects Cdc2, Cdk2, Cdk3
3	Phospho-Cdc2 p34 (Y15)	Santa Cruz Biotech	sc-7989	Ab also detects Cdc2, Cdk2, Cdk3
4	Phospho-cdc2 (Tyr15) (10A11)	Cell Signaling Technology	cst-4539	Based on sequence similarity, the antibody may cross-react with CDK2 and CDK3
5	Phospho Cdc2 (Tyr15)	Cell Signaling Technology	cst-9111	Antibody detects endogenous levels of cdc2, CDK2 and CDK5 only when phosphorylated at tyrosine 15
6	Anti-CDK2 (Phospho Y15). Clone FQS3344Z	Creative Diagnostics	DCABH-9682	No information about potential cross-reactivity
7	Anti-CDK1 (phospho Y15) [EPR7875]	Abcam	ab133463	No information about potential cross-reactivity
8	Anti-CDK1 (phospho Y15)	Abcam	ab47594	No information about potential cross-reactivity
9	Anti-CDK1 (phospho Y15)	Abcam	ab192207	No information about potential cross-reactivity
10	Anti-CDK1 (phospho Y15)	Abcam	ab69829	No information about potential cross-reactivity
11	Anti-CDK1 (phospho Y15)	Creative Diagnostics	CPBT-68207RC	No information about potential cross-reactivity. Cited references do not use this antibody
12	Anti-CDK1 (phospho Y15)	Creative Diagnostics	DPAB-DC2312	No information about potential cross-reactivity. Cited references do not use this antibody
13	Anti-Cdc2 (phospho Y15)	Abnova	PAB9677	No information about potential cross-reactivity. Cited references do not use this antibody
14	Anti-Cdc2 (phospho Y15)	Abnova	PAB12614	No information about potential cross-reactivity. Cited references do not use this antibody
15	Anti-Cdc2 (phospho Y15)	Abnova	PAB25833	No information about potential cross-reactivity
16	Anti-Cdc2 (phospho Y15)	R&D Systems	AF888	No information about potential cross-reactivity
17	Anti-Cdc2 (phospho Y15)	St. John's Laboratory	STJ90214	No information about potential cross-reactivity
18	Anti-Cdk1 (phospho Y15)	Elabscience	EAP1724	No information about potential cross-reactivity. Figure used was the same for western blot and IF as MBS2518967 (presumably same antibody)
19	Anti-Cdc2 (phospho Y15)	Elabscience	ENP0054	No information about potential cross-reactivity
20	Anti-Cdk1 (phospho Y15)	US Biologicals	210053	No information about potential cross-reactivity
21	Anti-Cdk1 (phospho Y15)	US Biologicals	210054	No information about potential cross-reactivity
22	Anti-Cdk1 (phospho Y15)	MyBioSource	MBS128566	No information about potential cross-reactivity
23	Anti-Cdk1 (phospho Y15)	MyBioSource	MBS2518967	No information about potential cross-reactivity. Figure used was the same for protein gel blot and IF as EAP1724 (presumably same antibody)
24	Anti-Cdk1 (phospho Y15)	Biorbyt	orb129556	No information about potential cross-reactivity
25	Anti-Cdc2 (phospho Y15)	Biorbyt	orb159475	No information about potential cross-reactivity.
26	Anti-Cdc2 (phospho Y15)	Biorbyt	orb177883	No information about potential cross-reactivity.
27	Anti-Cdc2 (phospho Y15)	Signalway Antibody	11244	No information about potential cross-reactivity.

In our own experiments, we immunoprecipitated CDK2 and found that it was phosphorylated at Y15 using a purported anti-Y15 CDK1 antibody (Cell Signaling Technology, cst-9111) confirming the lack of specificity.⁶ Several publications have noted the lack of specificity of these phospho-Y15 antibodies,^22-25 but clearly the majority of papers cited above (and many others) have misinterpreted results obtained with these antibodies. These concerns should also extend to antibodies that “selectively” detect phosphorylation on threonine-14.

CDK1 and CDK2 Inhibitors are Rarely Selective

The relative importance of CDK1 versus CDK2 in cell cycle regulation is frequently implicated on the basis of their sensitivity to inhibitors that are suggested to be selective. CDK inhibitors can be classified into broad-spectrum inhibitors (for example, flavopiridol, roscovitine, purvalanol A, dinaciclib)^2,27–29 and selective inhibitors (for example, CVT-313, RO-3306,).^28,30,31 It is important to note that the term ‘selective’ is misleading as all these drugs inhibit more than one member of the CDK family. For example, purvalanol A is selective for CDK1, CDK2 and CDK5 at low concentrations, but can also inhibit CDK4 at higher concentrations.^32-34 Similarly, dinaciclib has been used to selectively implicate CDK2, yet it inhibits CDK5 at the same concentrations, and CDK1 and CDK9 at only slightly higher concentrations.^35,36 Ro-3306 has been reported to be 10-fold more selective for CDK1 than CDK2 when assayed in an in vitro kinase assay,³¹ and this selectivity has generally been extrapolated to cells. However, our data shows that it inhibits both CDK1 and CDK2 at similar concentrations in cells.⁶ CVT-313 is reported to be 10-fold selective for CDK2 over CDK1 in kinase assays,³⁰ and our experiments demonstrate some selectivity also occurs in cells.⁶ Perhaps the only means to date to get highly selective inhibition of CDK2 is by mutating the gatekeeper amino acid in the ATP pocket, which makes the mutant kinase selectively sensitive to a bulky ATP mimetic.³⁷ The same strategy could be used to generate selective inhibition of CDK1.

Previous experiments had demonstrated that roscovitine prevents γH2AX induced by Chk1 inhibitors but these experiments did not resolve whether the critical kinase was CDK1 or CDK2.^38,39 In our most recent experiments, we observed that low concentrations of CVT-313 (˜1 µM) effectively inhibited γH2AX, but that higher concentrations (˜10 µM) inhibited phospho-histone H3 (pHH3), a marker of CDK1 activity, and which only occurred in the G2/M phase.⁶ It is important to emphasize that the γH2AX induced by Chk1 inhibition occurred only in S phase and hence reflects a different population of cells than those exhibiting pHH3. These results suggested that γH2AX was a consequence of activation of CDK2. Surprisingly, Ro3306 (˜2.5 µM) was equally effective at preventing both γH2AX and pHH3, suggesting it is inhibiting both CDK1 and CDK2 in cells.

This surprising lack of selectivity of Ro3306 for CDK1 may relate to the different abundance of CDK1 and CDK2 in cells. It has been reported that CDK1 is >10-fold more abundant than CDK2 in cells.⁴⁰ Perhaps even more important is a difference in the activation pathway for these 2 kinases.²³ Activation of CDK1 initially requires its binding to cyclin A or B, and it is then phosphorylated on tyrosine-15 (Y15); CDC25 family of phosphatases then remove this inhibitory phosphate for its activation. In contrast, CDK2 is phosphorylated on tyrosine prior to binding to cyclin, and only a small fraction actually binds to cyclin. Dephosphorylation by CDC25 only occurs on that fraction bound to cyclin. Hence, very little dephosphorylation of CDK2 occurs at any particular time. We surmise this low level of active CDK2 is readily inhibited by Ro3306.

Genetic manipulations are often considered a better reflection of the importance of a particular pathway because of the apparent off-target effects of small molecule inhibitors. However, it is worth considering what might happen if one tries to discriminate CDK1 from CDK2 using, for example, siRNA. We found this approach to be problematic as suppression of CDK2 and CDK1 arrested cells in G1 and G2 respectively (unpublished observation), thereby preventing an assessment of whether these kinases are involved in response to the Chk1 inhibitor whose sensitivity requires cells to be progressing through S phase. Small molecule inhibitors can circumvent this problem because they can rapidly inhibit a pathway and avoid critical changes in cell cycle distribution. However, long-term treatment with the inhibitors still needs to be avoided, as they do arrest cell cycle progression.

Cells do not tolerate CDK2 activity in S phase

While comparing the activity of Ro3306 and CVT-313 we also assessed their ability to arrest cell cycle progression in undamaged cells.⁶ Using synchronized cells, we demonstrated that both inhibitors prevented progression of cells from G1 to S phase. However, when added to cells that had already entered S, both inhibitors arrested the cells in G2. Importantly, neither inhibitor prevented progression through S phase. This appears to contradict the generally accepted role for CDK2 in regulating S phase progression, as opposed to its accepted role in S phase entry.

Entry into S phase requires the activity of both CDK2 and the Dbf4-dependent CDC7 kinase.⁴¹ Formation of the pre-replication complex involves loading of minichromosome maintenance proteins MCM2-7. Initiation of DNA replication occurs as a consequence of Dbf4/Cdc7 phosphorylation of the MCM2-7 helicase and CDK2-mediated recruitment of Cdc45. Many more pre-replication complexes form than are required for normal DNA synthesis, but some of the dormant origins are required for initiation of replication in late S phase. The dormant origins can also fire when the DNA is damaged, providing a means to circumvent blocks to DNA synthesis. Consequently, CDK2 is believed to be required during a normal S phase to facilitate firing of late replication origins. Hence, it was a surprise to observe that neither CVT-313 nor Ro3306 induced S phase arrest.

The primary goal of our recent investigation was to assess the sensitivity of a large panel of cell lines to the Chk1 inhibitor MK-8776, and to define the mechanism of sensitivity.⁶ We determined that 15% of cell lines were acutely sensitive to inhibition of Chk1 with DNA double-strand breaks appearing within 6 h in S phase cells. The majority of cell lines were resistant to Chk1 inhibition even though Chk1 was still inhibited in these cells. In these resistant cells, Chk1 inhibition failed to activate CDC25A, and therefore CDK2 was not activated. In contrast, sensitivity to Chk1 inhibition was attributed to the activation of CDC25A and CDK2. We concluded that cells did not tolerate activation of CDK2 in S phase, and this observation is consistent with the failure of CVT-313 to induce S phase arrest. CDK2 must be highly regulated in S phase, and is generally maintained inactive to preserve cell viability. We surmise that CDK2-mediated loading of CDC45 onto chromatin at the beginning of S phase maybe sufficient for cells to complete progression through S. We continue to investigate the potential role of CDK2 in response to S phase arrest mediated by DNA damage.

Elevated Cyclin E levels are a reflection of inactive CDK2

CDK1 and CDK2 phosphorylate many substrates throughout the cell cycle, and there maybe considerable overlap in their activity. We noted above that CDK2 signals to γH2AX in S phase, and CDK1 signals to pHH3 in G2/M, but neither H2AX nor HH3 are direct targets of CDK2 and CDK1, respectively. Selectivity for protein targets is somewhat controlled by the different cell cycle phases in which these kinases are active. However, under some circumstances, CDK1 can be activated aberrantly in S phase leading to premature mitosis.⁹ The number of CDK2 substrates that might be phosphorylated by CDK1 when it is active in S phase is unknown.

In seeking a selective target that reflected CDK2 activity, our attention was drawn to cyclin E. In G1/S-phase, cyclin E binds and activates CDK2, which in turn phosphorylates cyclin E on serine 399. This phosphorylation primes cyclin E for phosphorylation on threonine 395 by GSK3b thereby targeting it for degradation. This feedback loop limits overt activation of CDK2.^42,43 Our data has shown that cyclin E degradation can be used as a selective measure of Cdk2 activity in many cell lines. A few cell lines failed to degrade cyclin E when Cdk2 was active but this was attributed to inactive GSK3b.⁶ Treatment of these cells with LY294002, a PI3K inhibitor resulted in activation of GSK3b and this resulted in cyclin E degradation. In addition, when we incubated asynchronously growing cancer cells with CVT-313 at concentrations selective for CDK2 inhibition, cyclin E levels accumulated indicating that there is a basal level of CDK2-mediated degradation of cyclin E.⁶ However, there is an important consequence of these observations as high cyclin E is commonly thought to reflect active cyclin E/CDK2. However, our results clearly demonstrate that high cyclin E can be a consequence of inactive CDK2.

Concluding comments

Understanding the activities of cyclin-dependent kinases, especially CDK1 and CDK2 are of continuing interest due to their involvement in cancer. While inhibition of either CDK1 or CDK2 can arrest cell growth, their inappropriate activation has been shown to kill cancer cells. For example, inhibition of Chk1 activates CDK2 in a subset of cell lines resulting in DNA double-strand breaks in S phase cells.⁶ Similarly, activation of CDK1 during S phase can occur in some cells and results in premature mitosis.⁹ Unfortunately, the tools routinely employed to differentiate between the activities of these 2 kinases have presented an often unrecognized challenge. We have highlighted concerns for the use of antibodies and inhibitors that do not discriminate the activity of these 2 kinases, and for substrates such as cyclin E whose accumulation has been misconstrued as reflecting an increased activity of CDK2. We hope that by highlighting these concerns here, we will encourage critical reevaluation of past conclusions, and avoidance of similar errors in the future.

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Funding

The authors' research is supported by a research grant (CA117874) from the National Cancer Institute, as well as a Cancer Center Support Grant to the Norris Cotton Cancer Center (CA23108).