eIF5A-Independent Role of DHPS in p21CIP1 and Cell Fate Regulation

Deoxyhypusine synthase (DHPS) catalyzes the first step of hypusination of the elongation translation factor 5A (eIF5A), and these two proteins have an exclusive enzyme–substrate relationship. Here we demonstrate that DHPS has a role independent of eIF5A hypusination in A375 and SK-MEL-28 human melanoma cells, in which the extracellular signal regulated kinase 1/2 (ERK1/2) pathway is deregulated. We found that RNA interference of DHPS induces G0/G1 cell cycle arrest in association with increased p21CIP1 expression in these cells whereas eIF5A knockdown induces cell death without increasing p21CIP1 expression. Interestingly, p21CIP1 knockdown switched DHPS knockdown-induced growth arrest to cell death in these cells, suggesting a specific relation between DHPS and p21CIP1 in determining cell fate. Surprisingly, ectopic expression of DHPS-K329R mutant that cannot hypusinate eIF5A abrogated DHPS knockdown-induced p21CIP1 expression in these cells, suggesting a non-canonical role of DHPS underlying the contrasting effects of DHPS and eIF5A knockdowns. We also show that DHPS knockdown induces p21CIP1 expression in these cells by increasing CDKN1A transcription through TP53 and SP1 in an ERK1/2-dependent manner. These data suggest that DHPS has a role independent of its ability to hypusinate eIF5A in cells, which appears to be important for regulating p21CIP1 expression and cell fate.


Introduction
Deoxyhypusine synthase (DHPS) is the only known enzyme that catalyzes the first step of hypusination of the elongation translation factor 5A (eIF5A), which is the only DHPS substrate known to date [1,2]. Given this exclusive enzyme-substrate relationship, it is generally expected that DHPS and eIF5A mediate overlapping contexts of cellular processes, and although DHPS mRNA has been shown to be a natural antisense transcript for the scaffold protein MORG1 [3], no mutually independent function of DHPS and eIF5A has been reported at protein levels. Hypusination is a highly specific post-translational modification that has been detected with only eIF5A [4][5][6]. eIF5A hypusination is mediated by two step enzyme reactions that sequentially occur. First, DHPS conjugates the aminobutyl moiety of spermidine to a specific lysine residue of eIF5A to produce deoxyhypusine-eIF5A [7][8][9]. Second, deoxyhypusine hydroxylase (DOHH) reduces deoxyhypusine-eIF5A to irreversibly produce hypusine-eIF5A [10,11]. Cells express two functionally redundant highly homologous eIF5A isoforms, eIF5A1 and eIF5A2 [12]. eIF5A1 is the major isoform ubiquitously expressed in almost any tissue type, whereas eIF5A2 expression is limited to specific tissue or tumor types [13,14]. Hypusination enables eIF5A to promote translational elongation of the polypeptide chains stalling due to disruptive sequences such as polyproline-motifs and sequentially charged residues [15][16][17] as well as translational termination on a global scale [15,18,19]. As determined in yeast Saccharomyces cerevisiae and cells of higher organisms, eIF5A hypusination is a fundamental process for eukaryotic Using RNA interference, we determined the effects of DHPS depletion in A375 and SK-MEL-28 that exhibit high MEK/ERK activity due to BRAF V600E mutation. We found that knockdown of DHPS for 5 days by lentiviruses expressing two independent shRNA, which was sufficient to curtail eIF5A hypusination in cells, consistently increased p21 CIP1 expression in association with decreased phosphorylation of Rb in these cells ( Figure 1A). Rb is a key cell cycle regulator that sequesters the S-phase transcription factor E2F1 [28]. However, DHPS knockdown did not affect the levels of E2F1 and another CDKI p27 KIP1 , while not inducing p16 INK4a expression or notably increasing the cleavage of Poly ADPribose polymerase (PARP) and lamin A, the apoptotic cell death makers [29,30], in these cells ( Figure 1A). Consistent with these changes, DHPS knockdown substantially decreased cell proliferation rates without increasing death rates in these cells ( Figure 1B). Moreover, DHPS knockdown increased the G0/G1 phase cell population ( Figure 1C). These data demonstrate that DHPS depletion can induce cell cycle arrest which is associated with p21 CIP1 expression.

eIF5A Depletion Induces Cell Death without Increasing p21 CIP1 Levels in A375 and SK-MEL-28 Cells
As determined by Western blotting using an antibody that reacts with both eIF5A1 and eIF5A2 isoforms, RNA interference of eIF5A1 substantially decreased eIF5A levels in SK-MEL-28 and A375 cells (Figure 2A). In these cells, the effects of eIF5A depletion were in stark contrast to the effects of DHPS depletion. eIF5A knockdown in these cells, mediated by two independent lentiviral shRNA constructs, robustly increased the cleavage of lamin A and PARP but did not increase p21 CIP1 expression (Figure 2A). Consistent with the death markers, eIF5A knockdown substantially increased cell death in the cultures of . β-tubulin is the control for equal protein loading. Empty virus was used as a control. (bottom) Densitometry of p21 CIP1 . (B) Time-course trypan blue exclusion assays to determine viability of cells infected as described in (A). (C) Propidium iodide staining and flowcytometry to determine cell cycle phases of cells treated as described in (A). Data are mean ± SD of biological triplicates. * p < 0.05, ** p < 0.005 and *** p < 0.001 by two-tailed Student t-test.

eIF5A Depletion Induces Cell Death without Increasing p21 CIP1 Levels in A375 and SK-MEL-28 Cells
As determined by Western blotting using an antibody that reacts with both eIF5A1 and eIF5A2 isoforms, RNA interference of eIF5A1 substantially decreased eIF5A levels in SK-MEL-28 and A375 cells (Figure 2A). In these cells, the effects of eIF5A depletion were in stark contrast to the effects of DHPS depletion. eIF5A knockdown in these cells, mediated by two independent lentiviral shRNA constructs, robustly increased the cleavage of lamin A and PARP but did not increase p21 CIP1 expression (Figure 2A). Consistent with the death markers, eIF5A knockdown substantially increased cell death in the cultures of these cells, as determined by the trypan blue exclusion analysis ( Figure 2B). Importantly, Figure 1. DHPS knockdown increases p21 CIP1 levels and induces growth arrest without causing cell death in A375 andSK-MEL-28 cells. (A) Western blots of total lysates of cells infected for 5 days with pLKO.1 viruses expressing shRNA targeting different regions of DHPS mRNA (shDHPS #1 and shDHPS #2). β-tubulin is the control for equal protein loading. Empty virus was used as a control. (bottom) Densitometry of p21 CIP1 . (B) Time-course trypan blue exclusion assays to determine viability of cells infected as described in (A). (C) Propidium iodide staining and flowcytometry to determine cell cycle phases of cells treated as described in (A). Data are mean ± SD of biological triplicates. * p < 0.05, ** p < 0.005 and *** p < 0.001 by two-tailed Student t-test. in these cells ( Figures 1A and 2A). Therefore, the different effects of these treatments are not due to different eIF5A hypusination levels or altered ERK1/2 activity. These data suggested that DHPS and eIF5A may have distinct effects on cell fate and p21 CIP1 regulation in cells. Data are mean ± SD of biological triplicates. * p < 0.05, ** p < 0.005, and *** p < 0.001 by two-tailed Student t-test.  Figure 1A) and ( Figure 2A) normalized to β-tubulin. Data are mean ± SD of biological triplicates. * p < 0.05, ** p < 0.005, and *** p < 0.001 by two-tailed Student t-test.
The contrasting effects of DHPS and eIF5A knockdowns were not limited to A375 and SK-MEL-28 cells. Similar as these cells, DHPS knockdown suppressed cell proliferation without inducing cell death whereas eIF5A knockdown induced cell death responses, albeit at varied degrees, in the KRAS-mutant human pancreatic cell lines, MIA-PaCa-2 and PANC-1 (Supplemental Figure S1A-C). However, unlike in A375 and SK-MEL-28 cells, DHPS depletion did not increase p21 CIP1 levels in these cells (Supplemental Figure S1A), suggesting that cellular responses to DHPS knockdown are not entirely identical across different cell types.

p21 CIP1 Depletion Switches DHPS Depletion-Induced Growth Arrest to Cell Death
We previously showed that p21 CIP1 -mediated cell cycle arrest can suppress death responses in MEK/ERK-dependent tumor cells [31,32]. Given this and the data above, we asked whether p21 CIP1 has a role in determining cell fate in the face of DHPS depletion. Indeed, we found that blocking p21 CIP1 expression can switch DHPS depletion-induced cell cycle arrest to cell death responses. Knockdown of p21 CIP1 using lentiviruses expressing two independent shRNA consistently increased the cleavage of PARP and lamin A, and dead cell populations in A375 cultures only when combined with DHPS knockdown ( Figure 3A,B). Consistent with these effects, DHPS knockdown increased the cleavage of PARP and lamin A, and induced cell death in CDKN1A (encoding p21 CIP1 )-deleted HCT116 (p21 −/− ) cells, but not in the parental cells ( Figure 3C,D). Nevertheless, DHPS knockdown did not increase p21 CIP1 levels in HCT116 cells, further supporting the notion that DHPS-mediated p21 CIP1 regulation is cell type specific. p21 CIP1 depletion did not alter the effects of DHPS knockdown on eIF5A hypusination in these cells when compared with their parental cells ( Figure 3A,C), excluding potential interference by uneven eIF5A hypusination. These data demonstrate that p21 CIP1 can regulate cellular responses to DHPS depletion.

Hypusinating Activity Is Not Necessary for DHPS to Regulate p21 CIP1
Because the contrasting knockdown effects of DHPS and eIF5A may suggest an eIF5Aindependent role for DHPS, we examined the effects of the spermidine analog GC7 in A375 and SK-MEL-28 cells. GC7 is a competitive inhibitor of DHPS catalytic activity [33], and we expected that GC7 would help us determine whether DHPS knockdown-induced p21 CIP was due to depleted enzyme activity or protein. Our time-course analysis revealed that GC7, used at levels sufficient to inhibit eIF5A hypusination, did not increase but decreased the basal levels of p21 CIP1 in A375 and SK-MEL-28 cells ( Figure 4A). As an orthogonal approach, we also examined the effects of GC7 on p21 CIP1 expression induced upon activation of ∆Raf-1:ER in HEK293 cells. ∆Raf-1:ER is the CR3 catalytic domain of Raf-1 fused to the ligand binding domain of the estrogen receptor for tamoxifen-controlled activity [34]. In these cells, GC7 treatment substantially inhibited ∆Raf-1:ER-induced p21 CIP1 expression without affecting ERK1/2 phosphorylation ( Figure 4B). These data indicated that inhibition of DHPS catalytic activity cannot mimic DHPS depletion in cells, leading us to hypothesize that DHPS has a role independent of its catalytic activity to regulate p21 CIP1 in cells.
Lys329 of DHPS is necessary for the formation of an enzyme-imine intermediate required for eIF5A hypusination, and replacing this residue with Arg hinders the intermediate formation [35]. To test the aforementioned hypothesis, we generated a DHPS mutant in which Lys329 is replaced with Arg (DHPS-K329R) and determined whether this mutant can regulate p21 CIP1 levels in cells similarly as wild type DHPS. First, we confirmed that DHPS-K329R is catalytically deficient in HEK293 cells in which DOHH and eIF5A were co-expressed with either DHPS or DHPS-K329R to establish a proper stoichiometry between these proteins. Our data showed that, unlike wild type DHPS, DHPS-K329R cannot increase eIF5A hypusination in this assay ( Figure 4C). Expression of this mutant and wild type DHPS did not affect ERK1/2 phosphorylation induced by ∆Raf-1:ER activation ( Figure 4C). Next, we determined whether DHPS-K329R, which was engineered to avoid RNA interference, can restore p21 CIP1 suppression in DHPS-depleted A375 cells. Indeed, when ectopically expressed at similar levels, DHPS-K329R was similarly effective as wild type DHPS for abrogating the effect of DHPS knockdown to increase p21 CIP1 levels in cells ( Figure 4D). These data demonstrate that DHPS can negatively regulate cellular p21 CIP1 levels independently of its hypusinating activity, thus its substrate eIF5A. We also attempted to determine the ability of DHPS-K329R to rescue cells from DHPS knockdown, but prolonged overexpression of this mutant as well as wild type DHPS suppressed cell proliferation, hampering our efforts. It may be possible that unbalanced stoichiometry between DHPS, DOHH and eIF5A is anti-proliferative. (D) Trypan blue exclusion assays to determine viability of HCT116 cells infected as described in (C). Data are mean ± SD of biological triplicate and duplicate. * p < 0.05, ** p < 0.005, and *** p < 0.001 by two-tailed Student t-test.

Hypusinating Activity Is Not Necessary for DHPS to Regulate p21 CIP1
Because the contrasting knockdown effects of DHPS and eIF5A may suggest an eIF5A-independent role for DHPS, we examined the effects of the spermidine analog GC7 in A375 and SK-MEL-28 cells. GC7 is a competitive inhibitor of DHPS catalytic activity [33], and we expected that GC7 would help us determine whether DHPS knockdown-induced p21 CIP was due to depleted enzyme activity or protein. Our time-course analysis revealed that GC7, used at levels sufficient to inhibit eIF5A hypusination, did (D) Trypan blue exclusion assays to determine viability of HCT116 cells infected as described in (C). Data are mean ± SD of biological triplicate and duplicate. * p < 0.05, ** p < 0.005, and *** p < 0.001 by two-tailed Student t-test. sion induced upon activation of ΔRaf-1:ER in HEK293 cells. ΔRaf-1:ER is the CR3 catalytic domain of Raf-1 fused to the ligand binding domain of the estrogen receptor for tamoxifen-controlled activity [34]. In these cells, GC7 treatment substantially inhibited ΔRaf-1:ER-induced p21 CIP1 expression without affecting ERK1/2 phosphorylation ( Figure  4B). These data indicated that inhibition of DHPS catalytic activity cannot mimic DHPS depletion in cells, leading us to hypothesize that DHPS has a role independent of its catalytic activity to regulate p21 CIP1 in cells.

MEK/ERK Activity Is Necessary for DHPS Knockdown to Increase p21 CIP1 Levels in Cells
We previously demonstrated that MEK/ERK activity is sufficient to induce p21 CIP1 [36] and that the MEK/ERK activity in BRAF-mutant tumor cells can be routed to induce p21 CIP1 expression [37]. We therefore determined whether the MEK/ERK activity is necessary for DHPS knockdown to increase p21 CIP1 levels in cells. First, we found that AZD6244, a highly selective MEK1/2 inhibitor [38], substantially attenuated p21 CIP1 expression in DHPS depleted SK-MEL-28 cells ( Figure 5A). Second, the combination of DHPS knockdown and ∆Raf-1:ER activation robustly increased p21 CIP1 levels in HEK293 cells without affecting other cell cycle or death effectors ( Figure 5B). Conversely, DHPS overexpression in LNCaP cells, a cell line that rapidly expresses p21 CIP1 in response to ∆Raf-1:ER activation [32,36], substantially abrogated the p21 CIP1 expression induced by ∆Raf-1:ER activation ( Figure 5C). Moreover, consistent with its effect to augment p21 CIP1 expression, the combination of DHPS knockdown and ∆Raf-1:ER activation more potently suppressed HEK293 cell proliferation than DHPS knockdown or ∆Raf-1:ER activation alone ( Figure 5D). However, this combination did not notably induce cell death ( Figure 5D), which is consistent with its lack of effects on PARP and lamin A cleavage ( Figure 5B). Together, these data suggest that DHPS negatively regulates MEK/ERK-induced p21 CIP1 expression.   [39]. We found that DHPS knockdown substantially increased CDKN1A (encoding p21 CIP1 ) mRNA levels in A375 and SK-MEL-28 cells (Figure 6A), and thus sought to determine the molecular mechanism underlying this increase. Although TP53

DHPS Knockdown Increases CDKN1A Transcription via TP53 and SP1 in A375 and SK-MEL-28 Cells
p21 CIP1 is regulated at multiple levels including transcription, translation, and posttranslation [39]. We found that DHPS knockdown substantially increased CDKN1A (encoding p21 CIP1 ) mRNA levels in A375 and SK-MEL-28 cells (Figure 6A), and thus sought to determine the molecular mechanism underlying this increase. Although TP53 is the major regulator of CDKN1A transcription [40,41], we previously reported that SP1 can also mediate MEK/ERK-induced CDKN1A transcription in tumor cells deficient of a functional TP53 [42]. Accordingly, we examined the effects of TP53 knockdown and SP1 knockdown on DHPS knockdown-induced p21 CIP1 expression in A375 (TP53 wild type) and SK-MEL-28 (TP53 L145R ) cells. As determined by Western blotting in A375 cells, TP53 knockdown and SP1 knockdown substantially attenuated p21 CIP1 expression induced by DHPS knockdown while increasing lamin-A cleavage ( Figure 6B). This effect is consistent with p21 CIP1 depletion effects in the background of DHPS knockdown. Moreover, DHPS knockdown increased the activity of the luciferase reporter, H2320, that contains 2320 base pairs of CDKN1A promoter in A375 cells, but this increase was abrogated upon truncation of the region that contain TP53 and SP1 responsive elements ( Figure 6C). Similarly, DHPS knockdown upregulated the activity of the luciferase reporter, S2260, that contains 2260 base pairs of CDKN1A promoter in SK-MEL-28 cells but this induction was abrogated upon disabling the SP1 binding sites by site-directed mutagenesis ( Figure 6D). These data suggest that DHPS knockdown increases p21 CIP1 expression by promoting CDKN1A transcription through TP53 or SP1 in these cells. can also mediate MEK/ERK-induced CDKN1A transcription in tumor cells deficient of a functional TP53 [42]. Accordingly, we examined the effects of TP53 knockdown and SP1 knockdown on DHPS knockdown-induced p21 CIP1 expression in A375 (TP53 wild type) and SK-MEL-28 (TP53 L145R ) cells. As determined by Western blotting in A375 cells, TP53 knockdown and SP1 knockdown substantially attenuated p21 CIP1 expression induced by DHPS knockdown while increasing lamin-A cleavage ( Figure 6B). This effect is consistent with p21 CIP1 depletion effects in the background of DHPS knockdown. Moreover, DHPS knockdown increased the activity of the luciferase reporter, H2320, that contains 2,320 base pairs of CDKN1A promoter in A375 cells, but this increase was abrogated upon truncation of the region that contain TP53 and SP1 responsive elements ( Figure 6C). Similarly, DHPS knockdown upregulated the activity of the luciferase reporter, S2260, that contains 2,260 base pairs of CDKN1A promoter in SK-MEL-28 cells but this induction was abrogated upon disabling the SP1 binding sites by site-directed mutagenesis ( Figure  6D). These data suggest that DHPS knockdown increases p21 CIP1 expression by promoting CDKN1A transcription through TP53 or SP1 in these cells.

Discussion
In this report, we demonstrate that, although known for their exclusive enzymesubstrate relationship and thus expected to mediate largely overlapping cellular processes [5], DHPS and eIF5A depletion can result in distinct effects on cell fate. Mechanistically, our data suggest that DHPS has a novel eIF5A hypusination-independent function that can negatively regulate p21 CIP1 expression in cells by suppressing TP53-and SP1-mediated MEK/ERK activity-sensitive CDKN1A transcription (a model depicted in Figure 7). Given that DHPS does not interact with TP53 and SP1 (data not shown), we speculate that DHPS indirectly regulates CDKN1A transcription in this model, although the detailed molecular mechanism requires to be elucidated.
Our data provide compelling evidence supporting that DHPS can regulate cell fate independently of its activity to catalyze hypusination, thus of its only known substrate eIF5A. First, knockdown of DHPS and eIF5A produced different physiological effects in A375, SK-MEL-28, MIA-PaCa-2, PANC-1, and HCT116 cell lines. Whereas DHPS knockdown induced growth arrest, direct knockdown of eIF5A induced cell death accompanied by increased cleavage of lamin A and PARP, albeit at varied levels. In these cells, DHPS knockdown and eIF5A knockdown reduced eIF5A hypusination to similar degrees, and knockdown of one did not affect the expression of the other. Therefore, we suggest that the differential outcomes in cell fate were not a result of varying levels of eIF5A hypusination but probably a specific effect due to distinct roles of DHPS and eIF5A. Second, our data suggest a clear functional difference between DHPS and eIF5A in p21 CIP1 regulation. Whereas DHPS knockdown increased p21 CIP1 expression in A375 and SK-MEL-28 cells, eIF5A knockdown decreased it. Third, similar as eIF5A knockdown, GC7 decreased basal p21 CIP1 expression in these cells while exhibiting higher potency than eIF5A knockdown, which suggested that protein depletion but not hypusinating activity underlies the p21 CIP1 -inducing effect of DHPS knockdown. This notion is strongly supported by the data obtained with the catalytically disabled DHPS-K329R mutant. Given these contrasting effects on p21 CIP1 , we suspect that while DHPS maintains basal p21 CIP1 expression via an eIF5A-dependent mechanism, which fits into the context of the global effect of hypusinated eIF5A to promote protein translation [16], DHPS also regulates an eIF5A-independent mechanism to prevent p21 CIP1 overexpression caused by an abnormal signal such as hyper ERK1/2 activity (Figure 7). possible that DHPS has a role in promoting the cooperative effects of SP1 and TP53. How DHPS functionally interacts with TP53 and SP1 remains to be elucidated in future work. In summary, we report a previously unknown non-canonical DHPS function, which is independent of its activity to hypusinate eIF5A and regulates p21 CIP1 expression induced by MEK/ERK via TP53 and SP1. We propose that this function may be important for BRAF-mutated tumor cells to suppress p21 CIP1 overexpression, which can be triggered by aberrant MEK/ERK activity in these tumor cells if not prevented.
In summary, we report a previously unknown non-canonical DHPS function, which is independent of its activity to hypusinate eIF5A and regulates p21 CIP1 expression induced by MEK/ERK via TP53 and SP1. We propose that this function may be important for BRAF-mutated tumor cells to suppress p21 CIP1 overexpression, which can be triggered by aberrant MEK/ERK activity in these tumor cells if not prevented.
For lentivirus production, 293T cells were co-transfected with the lentiviral backbone and packaging vectors, as previously described [36]. Viral supernatants were collected after 72 h. Viral titers were determined by infecting the recipient cell lines with serially diluted viral supernatants, and then scoring cells expressing GFP at 48 h post-infection.

Analysis of Cell Viability and Cell Cycle
Cell viability was measured by trypan blue exclusion assay. The cell cycle analysis was performed as previously described [36]. Briefly, cells were washed with ice-cold 0.2% bovine serum albumin in phosphate-buffer saline, and then resuspended in 250 mM sucrose/40mM citrate buffer (pH 7.6) containing 0.5% dimethylsulfoxide. Cell nuclei were stained with propidium iodide and analyzed using the Guava EasyCyte flowcytometry system (Millipore Sigma, St. Louis, MO, USA). Gating targeted single cell nuclei within a normal size range. The cell-cycle parameters were determined from 5000 gated nuclei, and then analyzed with FCS EXPRESS 6 FLOW software (De Novo Software, Boulder, CO, USA).

Quantitative RT-PCR (qPCR) and Luciferase Reporter Assays
TRIzol ® reagent (Invitrogen, Waltham, MA, USA, 15596026) was used to isolate total RNA from cells. Reverse transcription was performed using Superscript III (Invitrogen, Waltham, MA, USA, 18080-044) and oligo-dT according to the manufacturer's instructions. Quantitative RT-PCR was performed by mixing the resulting cDNA with the SYBR GreenER qPCR Supermix Universal (Invitrogen, Waltham, MA, USA, 11762100) and the following primers: CTGGAGACTCTCAGGGTCGAA and CCAGCACTCTT AGGAACCTCTCA for p21 CIP1 ; AAGTTTGAGGACTGGCTGATG and CAGGGATGTGGT TCTTCTGG for DHPS; and GTCCTCTCCCAAGTCCACAC and GGGAGA CCAAAAGCCTTC AT for β-actin. qPCR signal was obtained using a Stratagene MX3005P instrument.
Membranes were then incubated with the appropriate antibodies overnight at 4 • C at the dilutions indicated as follows: phospho-ERK1/  [45]). SuperSignal West Pico and Femto chemiluminescence kit (Pierce, Waltham, MA, USA, 34094) were used for visualization of the signal. For densitometry, immunoblots were analyzed using Image Lab software (Bio-Rad, Hercules, CA, USA).

Statistical Analysis
Statistical significance was determined using the two-tailed unpaired Student's t-test using PRISM (Graph-Pad Software, La Jolla, CA, USA) of two data sets. p values of <0.05 were considered statistically significant.