Balance between senescence and apoptosis is regulated by telomere damage–induced association between p16 and caspase-3

Telomerase activation protects cells from telomere damage by delaying senescence and inducing cell immortalization, whereas telomerase inhibition mediates rapid senescence or apoptosis. However, the cellular mechanisms that determine telomere damage–dependent senescence versus apoptosis induction are largely unknown. Here, we demonstrate that telomerase instability mediated by silencing of sphingosine kinase 2 (SPHK2) and sphingosine 1-phosphate (S1P), which binds and stabilizes telomerase, induces telomere damage–dependent caspase-3 activation and apoptosis, but not senescence, in p16-deficient lung cancer cells or tumors. These outcomes were prevented by knockdown of a tumor-suppressor protein, transcription factor 21 (TCF21), or by ectopic expression of WT human telomerase reverse transcriptase (hTERT) but not mutant hTERT with altered S1P binding. Interestingly, SphK2-deficient mice exhibited accelerated aging and telomerase instability that increased telomere damage and senescence via p16 activation especially in testes tissues, but not in apoptosis. Moreover, p16 silencing in SphK2−/− mouse embryonic fibroblasts activated caspase-3 and apoptosis without inducing senescence. Furthermore, ectopic WT p16 expression in p16-deficient A549 lung cancer cells prevented TCF21 and caspase-3 activation and resulted in senescence in response to SphK2/S1P inhibition and telomere damage. Mechanistically, a p16 mutant with impaired caspase-3 association did not prevent telomere damage–induced apoptosis, indicating that an association between p16 and caspase-3 proteins forces senescence induction by inhibiting caspase-3 activation and apoptosis. These results suggest that p16 plays a direct role in telomere damage–dependent senescence by limiting apoptosis via binding to caspase-3, revealing a direct link between telomere damage–dependent senescence and apoptosis with regards to aging and cancer.

Telomerase is composed of a catalytic reverse transcriptase (TERT) 3 and telomeric template RNA, which plays a key role in the elongation and maintenance of telomeres at chromosome ends (1)(2)(3)(4). Shelterin complex proteins associate with telomere repeats for protection from inducing DNA damage response to overcome replicative end protection problems (5)(6)(7)(8). Telomerase is highly expressed in stem cells, whereas it is inactivated due to epigenetic silencing of TERT upon differentiation in most somatic cell types (9,10). However, telomerase is reactivated in the majority of cancer cells, inducing cell immortalization and tumorigenesis (11)(12)(13)(14). Recent studies revealed a clinical relevance of TERT activation in cancer, as mutations within the core promoter region of the TERT gene provide binding sites for E-twenty-six transcription factors, leading to cancerspecific telomerase reactivation (15)(16)(17).
Activation of telomerase protects cells from telomere damage, delaying senescence and the aging process (18,19), whereas targeting telomerase induces tumor suppressor responses such as senescence or apoptosis (20,21). However, signaling mechanisms that distinctly induce senescence versus apoptosis in noncancerous versus cancerous cells in response to telomerase inactivation and telomere damage induction are largely unknown. Sphingosine 1-phosphate (S1P) is a bioactive sphingolipid, generated by sphingosine kinase 1 or 2 (SPHK1 or SPHK2), which exerts pro-oncogenic signaling via engaging with its G protein-coupled receptors (S1PR1-5) in an autocrine or paracrine manner, and/or via directly binding to regulate its intracellular target proteins without receptor signaling (22)(23)(24)(25). Intracellular targets of SPHK2-generated nuclear S1P include HDAC1/2 (26) and TERT (27). Our previous study described a role for SPHK2-generated S1P in the stabilization of telomerase via directly binding to hTERT involving its Asp 684 residue (27).
These studies also suggested that S1P binding mimicked the phosphorylation of hTERT at Ser 921 , which protected hTERT from ubiquitination and proteasomal degradation (27). However, roles and mechanisms by which targeting SPHK2/S1P and TERT induces senescence versus apoptosis with respect to telomere damage signaling and tumor suppression are largely unknown. Thus, in this study, we have outlined experiments to uncover signaling mechanisms that determine responses for the induction of senescence versus apoptosis via telomere damage-mediated cellular stress.

Inhibition of SPHK2/S1P attenuates hTERT abundance and tumor growth
To determine the roles of SPHK2/S1P in the regulation of tumor suppression via controlling telomerase and telomere damage, we measured the effects of silencing and/or pharmacologic inhibition of SPHK2/S1P signaling using shRNA-dependent knockdown or small-molecule inhibitor of SPHK2, ABC294640 (Fig. S1A) (28 -30),on the growth of A549 xenograft-derived tumors in SCID mice. The shRNA-mediated knockdown of SPHK2 (ϳ75% mRNA reduction compared with Scr-shRNA-expressing tumors, measured by qPCR) almost completely prevented A549 -xenograft-tumor growth compared with Scr-shRNA/A549 -xenograft controls after 21 days (Fig. 1, A and B). Similarly, inhibition of SPHK2 using ABC294640 (100 mg/kg, 7 days/week treatment for 21 days) suppressed A549 -xenograft-tumor growth compared with vehicle-treated controls (Fig. 1A). Interestingly, ABC294640 treatment had no additional effect on A549 -xenograft-tumor growth suppression in response to shRNA-mediated SPHK2 knockdown, supporting its role in SPHK2 inhibition for tumor suppression (Fig. 1A). Suppression of A549 -xenograft-tumor growth in response to SPHK2 knockdown or inhibition was consistent with decreased hTERT protein abundance and apoptosis measured by IHC using anti-hTERT antibody and TUNEL assay, respectively, compared with control tumor tissues ( Fig. 1, C and D). In addition, SPHK2 inhibition or knockdown resulted in increased telomere damage in A549xenograft-tumors compared with controls, measured ex vivo by immunofluorescence using anti-TRF2 and anti-␥-H2AX antibodies in telomere damage-induced foci (TIF) assay ( Fig.  1E) (31). Moreover, there was no significant difference in the staining of senescence-associated-␤-gal in shSPHK2 tumors compared with shScr controls (Fig. 1F). Similar data were also obtained when A549-luciferase-derived tumors were grown Tumor growth was measured every other day for up to 21 days by calipers. After the growth study was completed, the tumors were removed and measured ex vivo (lower panel). Two-way ANOVA with Tukey's post hoc test was performed. shScr is versus shScr ϩ ABC (*, p Ͻ 0.001, F ϭ 22.76, DF ϭ 27); shScr is versus shSK2 (**, p Ͻ 0.001, F ϭ 35.48, DF ϭ 9); and shSK2 versus shSK2 ϩ ABC (***, p ϭ 0.997). B, shRNA-mediated knockdown of ISPHK2 mRNA in A549 xenograft-derived tumors was confirmed using qPCR after the tumor growth measurements were completed. Data are means Ϯ S.D. from three independent experiments, analyzed by Student's t test (n ϭ 3, *, p ϭ 0.039; **, p ϭ 0.014). C and D, effects of shRNA-mediated knockdown and/or ABC294640-dependent inhibition of SPHK2 on hTERT expression (C) or apoptosis (D) measured using IHC or TUNEL staining (data are means Ϯ S.D. from three independent experiments, analyzed by Student's t test, n ϭ 3, *, p ϭ 0.025; **, p ϭ 0.042; ***, p ϭ 0.66), respectively, were measured in A549 xenograft-derived tumors. Scale bars represent 100 m. E and F, telomere damage (E) in A549 xenograft-derived tumors expressing Scr-or SPHK2-shRNAs in the absence/presence of ABC294640 were measured using TIF assay. Quantification of images was performed using ImageJ (right panel). Data are means Ϯ S.D. from three independent experiments, analyzed by Student's t test (n ϭ 3, *, p ϭ 0.006; **, p ϭ 0.013; ***, p ϭ 0.003). F, senescence in A549-derived tumors expressing Scr-or SPHK2-shRNAs in the absence/presence of ABC294640 were measured using SA-␤-gal staining. Scale bars represent 100 m.

p16 -caspase-3 complex and senescence
orthotopically in the lungs of SCID mice after tail vein injection and inhibition of SPHK2 using ABC294640 decreased lung tumor growth (ϳ95%) and hTERT abundance compared with vehicle-treated controls (Fig. S1, B-D).
To assess the clinical relevance of hTERT regulation by SPHK2/S1P signaling, we then determined SPHK2 abundance with regard to hTERT expression by IHC using commercially available tumor microarrays (TMAs) containing lung tumor tissues obtained from patients with NSCLC (n ϭ 48) and their adjacent pathologically noncancerous lung tissues (n ϭ 48). The data showed that SPHK2 is significantly overexpressed (ϳ2.5-fold) in the majority of NSCLC tumors compared with their controls (46/48 tumors, p Ͻ 0.05) (Fig. S2, A and B). The data also showed that hTERT was up-regulated (ϳ2.8-fold) in the majority of NSCLC tumors (47/48, p Ͻ 0.05) (Fig. S2B). Importantly, the Spearman correlation analysis showed that there is a significant (n ϭ 48, p Ͻ 0.001) association between SPHK2 and hTERT expression in these tissues (Fig. S2C). Similar correlation between SPHK2 and hTERT was detected in the majority of tumor tissues obtained from patients with small cell lung cancer (Fig. S2D). Thus, these data suggest that targeting clinically relevant and up-regulated SPHK2/S1P in tumors inhibits hTERT abundance, which is associated with increased telomere damage and apoptosis, leading to tumor growth suppression in vivo.
Increased telomere damage in response to genetic loss of SphK2 was also detected in SphK2 Ϫ/Ϫ compared with WT-MEFs (Fig. 3A). Moreover, shRNA-mediated SPHK2 knockdown resulted in telomere damage in A549 cells compared with Scr-shRNA-transfected controls (Fig. 3B). As a control, we also detected the effects of hTERT knockdown on telomere damage in A549 cells in the absence/presence of ABC294640. shRNAmediated hTERT knockdown and SPHK2 inhibition using ABC294640 induced telomere damage at a similar level (ϳ2.5fold) compared with controls (Fig. 3C). However, inhibition of SPHK2 had no additional effect on telomere damage in response to hTERT knockdown (Fig. 3C). Moreover, shRNAmediated knockdown of SphK2 or its pharmacologic inhibition using ABC294640 resulted in telomere damage and subsequent caspase-3 activation in A549 cells in response to G 0 /G 1 arrest during serum starvation (Fig. S3, A and B and C and D). In addition, shRNA-mediated knockdown of SphK2 (Fig. S4A) decreased S1P and dihydro-S1P without causing large changes in C 16 -ceramide or dihydro-C 16 -ceramide compared with Scr-shRNA-transfected cells (Fig. S4B). These data suggest that telomere damage and/or caspase-3 activation in response to SphK2/S1P knockdown is regulated independently of cell cycle status and/or changes in ceramide and dihydroceramide accumulation in A549 cells.
However, SphK2 inhibition using ABC294640 mediated telomere damage without detectable caspase-3 activation in primary noncancerous human lung fibroblasts (generation 2, Fig. S5), suggesting that caspase-3 activation downstream of telomere damage might be regulated differentially in A549 lung cancer versus no cancerous lung fibroblasts. Interestingly, inhibition of SPHK2/S1P signaling using ABC294640 had no effect on p53-induced DNA damage as measured by co-localization of p53 and ␥-H2AX using immunofluorescence in A549 cells (Fig. 3D). Cisplatin exposure was used as a positive control for p53-mediated DNA damage, which increased the co-localization of p53 and ␥-H2AX in A549 cells compared with vehicletreated controls (Fig. 3D). These data suggest that targeting SphK2/S1P signaling results in caspase-3 activation due to selective telomere damage and not due to the activation of general DNA damage response.

SphK2/S1P inhibition leads to telomere damage through attenuation of S1P-hTERT association and phosphomimicking of telomerase
To determine the involvement of hTERT in the regulation of telomere damage and subsequent caspase-3 activation, we performed studies in GM847 cells, which do not express endogenous TERT, but contain telomeric template RNA (27). In GM847 cells, ABC294640 increased caspase-3 activity ϳ3-fold compared with controls, and ectopic expression of WT-hTERT abrogated this ABC294640-mediated caspase activation (Fig. 4,  A and B). However, ectopic expression of a mutant hTERT with p16 -caspase-3 complex and senescence D684A conversion, with decreased S1P binding, was not protective for caspase-3 activation in response to ABC294640 (Fig.  4, A and B). Moreover, expression of the phosphomimic mutant of hTERT (S921D) with D684A conversion, which does not bind S1P (Fig. S3C), was able to protect ABC294640-mediated caspase activation (Fig. 4, A and B). Expression of WT-hTERT and mutant hTERT proteins was confirmed using Western blotting (Fig. 4D).
These data were also consistent with the effects of ectopic expression of WT-hTERT, but not D684A-hTERT with altered S1P binding, on the protection of telomere damage in response to SPHK2/S1P targeting using ABC294640 compared with vector-transfected controls (Fig. 5, A and B). However, phosphomimic mutant of hTERT restored telomere damage in response to ABC294640 even in the absence of S1P binding due to D684A conversion in cells that express D684A/S921D double mutation compared with controls, expressing S921A or S921D single mutations, which still bind S1P and protected telomere damage (Fig. 5, A and B). Similarly, although ectopic expression ofWT-hTERTpreventedABC294640-mediatedcelldeath,measured by trypan blue exclusion assay, inhibition of S1P-hTERT binding by D684A mutation restored cell death in response to ABC294640 compared with controls ( Fig. 5C). However, phosphomimic S921D/D684A-hTERT mutant expression also prevented ABC294640-mediated cell death (Fig. 5C). Thus, these data suggest that S1P binding to hTERT involving Asp 684 protects telomere damage and caspase activation by mimicking phosphorylation of hTERT at Ser 921 . These data also suggest that targeting SPHK2/S1P signaling by ABC29640 induces telomere damage due to attenuation of S1P-hTERT association, leading to inhibition of phosphomimicking of hTERT, hTERT instability, and subsequent caspase-3 activation.
To determine the effects of TCF21 activation in the regulation of lung tumor suppression in response to SPHK2/S1P/ hTERT targeting in vivo, we measured the growth of A549derived xenografts expressing Scr-or TCF21-shRNAs in the absence/presence of ABC294640. Treatment with ABC294640 (100 mg/kg at every 7 days for 21 days) inhibited the growth of Scr-shRNA-expressing tumors compared with vehicle-treated controls (Fig. 6A). shRNA-mediated knockdown of TCF21 prevented ABC29464-mediated tumor suppression compared with controls ( Fig. 6A). After growth measurements, the remaining tumors were extracted from these mice, and the effects of ABC294640 in the presence/absence of TCF21 knockdown on hTERT mRNA, caspase-3 activation, and apoptosis were measured. ABC294640 reduced hTERT mRNA abundance in the presence/absence of TCF21 knockdown (Fig.  6B), suggesting that TCF21 induction is regulated downstream of hTERT down-regulation in response to targeting SPHK2/ S1P signaling. Moreover, ABC294640 resulted in caspase-3 activation (measured by IHC) and apoptosis (measured by TEM), which were prevented in response to shRNA-mediated knockdown of TCF21 (Fig. 6, C and D). shRNA-mediated knockdown of TCF21 was confirmed in A549 xenograft-derived tumors ex vivo by qPCR (Fig. 6E). sh-RNA-mediated knockdown of TCF21 had no detectable effects on telomere damage in response to ABC294640-mediated SPHK2 inhibition compared with control tumors (Fig. 6, F and G). Thus, these data suggest a role for TCF21 induction in caspase-3 activation and apoptosis in response to the inhibition of SPHK2/ S1P/hTERT axis and telomere damage by ABC294640, leading to tumor suppression. A, effects of genetic loss of SphK2 in inducing telomere damage were measured using TIF assay in MEFs isolated from SphK2 Ϫ/Ϫ compared with WT mice by immunofluorescence. Images were quantified using ImageJ (right panel). Data are means Ϯ S.D. from three independent experiments, analyzed by Student's t test (n ϭ 3, *, p ϭ 0.001). Scale bars represent 100 m. B and C, effects of shRNA-mediated knockdown of SPHK2 (B) or hTERT in A549 cells on telomere damage in the absence/presence of ABC294640 (C) were measured using TIF assay by immunofluorescence. Images were quantified using ImageJ (upper right panel). Data are means Ϯ S.D. from three independent experiments, analyzed by Student's t test (n ϭ 3, *, p ϭ 0.03). shRNA-mediated knockdown of hTERT mRNA was confirmed using qPCR compared with Scr-shRNA-transfected controls (lower left panel); data are means Ϯ S.D. from three independent experiments, analyzed by Student's t test (n ϭ 3, *, p ϭ 0.047); and TIF colocalization was quantified. Data are means Ϯ S.D. from three independent experiments, analyzed by Student's t test (n ϭ 3, *, p ϭ 0.030; **, p ϭ 0.003; ***, p ϭ 0.108). Scale bars represent 100 m. D, effects of ABC294640 or cisplatin on p53-dependent DNA damage was measured by immunofluorescence using anti-␥H2AX and anti-p53 antibodies in A549 cells compared with vehicle (DMSO)-treated controls. Images were quantified using ImageJ (right panel). Data are means Ϯ S.D. from three independent experiments, analyzed by Student's t test (n ϭ 3, *, p ϭ 0.031; **, p ϭ 0.605). Scale bars represent 100 m.

p16 -caspase-3 complex and senescence Genetic loss of SphK2 results in accelerated senescence and aging via telomere damage and p16 induction without apoptosis in mice
To determine the physiological impact of loss of SphK2/S1P signaling on senescence and aging due to TERT alterations in a noncancerous model, we have generated subsequent generations of WT and SphK2 Ϫ/Ϫ mice (for six generations). SphK2 Ϫ/Ϫ mice exhibited accelerated senescence and aging phenotype measured by increased ␤-gal expression in their paws (37) at generation 6, and decreased weight of testes at generation 4 or 5 compared with WT and SphK1 Ϫ/Ϫ mice (generation matched controls) (Fig. 7, A and B). Examination of spleen, skin, and testes tissues at generation 6 using H&E staining also confirmed increased senescence and aging phenotype in SphK2 Ϫ/Ϫ compared with WT mice.
SphK2 Ϫ/Ϫ animals at generation 6 exhibited splenic atrophy with decreased white blood cells with no germinal centers present. It should be noted that the splenic atrophy is common in aged mice and can be accompanied with cachexia. There was decreased erythrocyte (red blood cell) turnover in the SphK2 Ϫ/Ϫ mice, which might be due to decreased numbers of erythrocytes in circulation. Moreover, we observed decreased thickness of the hypodermal adipose layer in the SphK2 Ϫ/Ϫ mice with multiple areas of inflammation, increased fibrosis, and abnormal follicle development. Also, there were fewer seminiferous tubules (hypoplasia) in SphK2 Ϫ/Ϫ mice compared with WT control testes tissues (Fig. 7C).
Intestestissues,therewasdecreasedmTERTabundance(measured by IHC) in SphK2 Ϫ/Ϫ compared with WT and SphK1 Ϫ/Ϫ mice (generation 6), whereas there was no apoptosis (measured by TUNEL staining) or TCF21 induction (measured by IHC) in these mice (Fig. 7D). In addition, telomere damage induction was detected in the testes of SphK2 Ϫ/Ϫ but not in WT or SphK1 Ϫ/Ϫ mice (Fig. 7E). Thus, these data indicate that loss of SphK2 results in accelerated senescence and aging starting at generation 4, which is consistent with decreased mTERT and induction of telomere damage (especially in testes tissues). However, accelerated senescence/ aging in mice appeared to be independent of TCF21 and/or apoptosis induction.
To further examine the effects of SphK2 loss, confirmed by Western blotting using anti-SphK2 antibody (38), in the induction of senescence, we have isolated MEFs at generations 4 -6 from WT, SphK1 Ϫ/Ϫ , and SphK2 Ϫ/Ϫ mice and measured senescence by measuring ␤-gal and p16 expression. Induction of senescence was detected in the generations 4 -6 (G4 -G6) of SphK2 Ϫ/Ϫ compared with WT and SphK1 Ϫ/Ϫ MEFs (Fig. 8, A  and B). In addition, increased telomere damage was detected in

Induction of p16 distinctly mediates telomere damagemediated senescence and prevents apoptosis in response to SPHK2/S1P/telomerase targeting
To determine whether alterations of SPHK2/S1P/hTERT signaling induces senescence in A549 lung cancer cells, which lack p16 expression (39), we measured the effects of downregulation or inhibition of SPHK2 on caspase-3 activation and ␤-gal induction in the absence/presence of ectopic expression of p16. shRNA-mediated knockdown of SPHK2 induced caspase-3, and ectopic expression of p16 attenuated this process compared with Scr-shRNA-expressing A549 cells (Fig.  9A). shRNA-dependent knockdown or ABC294640-mediated inhibition of SPHK2 had no detectable effect on the induction of senescence in A549 cells, whereas ectopic expression of p16 with/without SPHK2 targeting increased senescence in A549 cells compared with Scr-shRNA-expressing or vehicle-treated controls (Fig. 9B). Interestingly, although ectopic expression of p16 had no effect on the induction telomere damage (Fig. 9C), it completely prevented caspase-3 induction in response to SPHK2/S1P inhibition using ABC294640 compared with vehicle-treated controls (Fig. 9, D and E). In reciprocal studies, shRNA-dependent knockdown of p16 in H1341 human lung cancer cells (confirmed by qPCR, Fig. 9F) resulted in enhanced caspase-3 activation in response to SPHK2 inhibition by ABC294640 compared with Scr-shRNA-transfected controls (Fig. 9G). Lack of protein abundance of p16 in A549 and H157 human lung cancer cells compared with HeLa and H1341 cells, which express p16, is also confirmed by Western blotting (Fig.   9H). Thus, these data suggest that p16 expression determines the induction of senescence while inhibiting TCF21-dependent caspase-3 induction in response to SPHK2/S1P/TERT targeting and telomere damage in lung cancer cells.

Interaction between p16 and caspase-3 prevents caspase-3 activation and apoptosis in response to telomere damage
To define the mechanism by which p16 activation forces cells to undergo senescence but not apoptosis, we prepared studies to test a novel hypothesis that interaction between p16 and caspase-3 might limit caspase-3 activation and prevent apoptosis. This hypothesis was first tested in H1341 human lung cancer cells, which express p16. SPHK2 inhibition by ABC294640 mediated the association between endogenous p16 and caspase-3 in H1341 cells, and this interaction was attenuated in response to shRNA-mediated silencing of p16 compared with controls (Fig. 10A). Similar data were also observed when SPHK2 expression was knocked down using shRNA, which resulted in increased association between ectopically expressed WT-p16 and caspase-3 in A549 cells with/without ABC294640 exposure compared with Scr-shRNA/vector-transfected and vehicle-treated controls using PLA and immunofluorescence (Fig. 10, B and C). These data were also consistent when we measured the binding between recombinant and purified human p16 and caspase-3 proteins (40) using OpenSPR-localized surface plasmon resonance biosensor at various protein concentrations. The data showed that purified p16 interacted with caspase-3 (K D , equilibrium dissociation constant ϭ 1.82 ϫ One-way ANOVA with Tukey's post hoc analysis test was performed (WT versus SphK2 Ϫ/Ϫ (*, p Ͻ 0.0001) and SphK1 Ϫ/Ϫ versus SphK2 Ϫ/Ϫ (**, p Ͻ 0.0001) with F ϭ 18.91 and DF ϭ 76. C, spleen, skin, and testes tissues from WT and SphK2 Ϫ/Ϫ mice at generation 5 were isolated and examined after H&E staining. D, apoptosis was detected by TUNEL staining in testes tissues isolated from WT and SphK2 Ϫ/Ϫ mice at generation 4 (G4, upper panel). Protein abundance of mTERT and TCF21 were detected by IHC using anti-mTERT and anti-TCF21 antibodies in testes tissues isolated from WT, SphK1 Ϫ/Ϫ , and SphK2 Ϫ/Ϫ mice at generation 4 (G4, lower panels). E, telomere damage in testes isolated from G4 of WT, SphK1 Ϫ/Ϫ , and SphK2 Ϫ/Ϫ mice was measured using TIF assay. Scale bars represent 100 m. All mice were between 6 and 8 weeks old. p16 -caspase-3 complex and senescence suggest that p16 binding might prevent the activation of pro-caspase-3.
Next, using molecular simulation and docking, we generated a model in silico for protein-protein interaction between p16 and caspase-3 using their existing structures determined by X-ray crystallography (Fig. 10D). The model suggested that the interaction between p16 and caspase-3 might involve Ser 152 of p16 (turquoise) and Gly 251 of caspase-3 (gray), which is within the activation domain of caspase-3 (His 237 -Cys 285 ). To validate this model, we ectopically expressed WT-p16 and mutant-p16 with Ser 152 to Ala conversion, and we measured their association in p16-null A549 cells using anti-p16 and anti-caspase-3 antibodies by PLA compared with vector-transfected controls. The data showed that WT-p16 and caspase-3 association was induced in response to SPHK2 inhibition using ABC294640, whereas there was no detectable association between S152A-p16 in A549 cells (Fig. 10E). Moreover, ectopic expression of WT-p16 and not S152A-p16 largely protected cells from caspase-3 activation in response to ABC294640 exposure compared with vector-transfected controls (Fig. 10F). Ectopic expression of WT-p16 and S152A-p16 in A549 cells was confirmed using Western blotting compared with vector-transfected controls using anti-p16 and anti-actin antibodies (Fig.   10G). Importantly, interaction between p16 and caspase-3 was also detected in testes tissues isolated from Sphk2-deficient mice at generation 5, which expressed SA-␤-Gal, but not in testes isolated from matched WT or SphK1 Ϫ/Ϫ mice, which were also negative for SA-␤-Gal expression (Fig. 10H). Thus, these data suggest that interaction between p16 and caspase-3 involves the Ser 152 residue of p16 and that this interaction is key to prevent caspase-3 activation, which forces senescence while limiting apoptotic cell death in response to SPHK2/S1P inhibition, and subsequent telomere damage.

Discussion
Here, our data revealed that SPHK2-generated S1P binds and stabilizes TERT involving its Asp 684 residue via mimicking TERT phosphorylation at Ser 921 , which protects telomere damage, and results in delayed senescence and protection from apoptosis. These data also suggested that targeting/inhibition of the SPHK2/S1P/telomerase axis results in telomere damage, which signals TCF21-dependent caspase-3 activation in cancer cells or immortalized MEFs, or p16-dependent senescence and accelerated aging phenotype in noncancerous lung fibroblasts, primary MEFs (nonimmortalized), and noncancerous testes tissues in subsequent generations of SphK2 Ϫ/Ϫ mice (Fig. 11). Mechanistically, this study demonstrated that interaction Figure 8. Genetic loss of SphK2 results in accelerated p16-dependent senescence but not apoptosis via increased telomere damage in noncancerous fibroblasts. A, senescence in testes tissues obtained from WT and SphK2 Ϫ/Ϫ mice at generations 4 -6 was measured by SA-␤-gal staining. Scale bars represent 100 m. B, abundance of p16 mRNA was measured by qPCR in MEFs isolated from WT, SphK1 Ϫ/Ϫ , and SphK2 Ϫ/Ϫ mice at generation 6 in the absence/presence of Scr-or p16-shRNAs. Data are means Ϯ S.D. from three independent experiments, analyzed by Student's t test (n ϭ 3, *, p ϭ 0.0402; **, p ϭ 0.0019). C and D, effects of p16 knockdown on telomere damage (C). Data are means Ϯ S.D. from three independent experiments, analyzed by Student's t test (n ϭ 3, *, p ϭ 0.006; **, p ϭ 0.013) or caspase-3 activation (D), and were measured using immunofluorescence in MEFs obtained from WT, SphK1 Ϫ/Ϫ , and SphK2 Ϫ/Ϫ mice at generation 6. Scr-shRNA transfected SphK2 Ϫ/Ϫ MEFs were used as controls. Images were quantified using ImageJ (right panels). Data are means Ϯ S.D. from three independent experiments, analyzed by Student's t test (n ϭ 3, *, p ϭ 0.001).

p16 -caspase-3 complex and senescence
between p16 and caspase-3, involving Ser 152 of p16, prevents caspase-3 activation in response to telomere damage, forcing cells to selectively induce senescence but not apoptosis. Our previous study revealed that S1P-TERT binding, involving the Asp 684 residue, stabilizes TERT via phosphomimic function of S1P at Ser 921 to regulate hTERT stability (27). Here, the data further suggested that whereas targeting SPHK2/S1P induces hTERT degradation, resulting in telomere damage and apoptosis, expression of the phosphomimic mutant of hTERT with S921D conversion even in the absence of S1P binding due to D684A mutation prevented telomere damage in response to SPHK2 inhibition. It will be interesting to define whether the phosphomimic function of S1P upon binding to other proteins, such as HDAC1/2 (26) and prohibitin 2 (41), or TRAF2 (22) and peroxisome proliferatoractivated receptor-␥ (42), play any roles in the regulation of their biological functions.
Interestingly, our current data also suggested that TCF21-dependent caspase-3 activation plays a key role in tumor suppression downstream of telomere damage in response to SPHK2/ S1P targeting. TCF21 was reported as a tumor suppressor, whose expression is suppressed by epigenetic methylation in lung tumor tissues and cells (43,44). It was shown recently that TCF21 is reactivated by the expression of long noncoding RNA TARID, which enhanced its demethylation via GADD45A (45). However, it remains unknown whether SPHK2/S1P/TERT inhibition results in TCF21 activation through TARID/GADD45A.
In these studies, targeting SPHK2/S1P signaling using molecular (shRNA) or pharmacologic (ABC294640) tools resulted in consistent telomere damage in various cell types. Although ABC294640 is known to have off-target effects, especially inducing the accumulation of dihydroceramides due to inhibition of dihydroceramide desaturase (46,47), knockdown of SphK2 using shRNA had no effect on dihydroceramide accumulation (Fig. S4B). Thus, these data suggest that telomere damage evoked by SphK2 silencing or inhibition appears to be independent of ceramide/dihydroceramide alterations. However, the effects of shRNA-mediated knockdown of SphK2 or its inhibition using ABC294640 on S1P versus ceramide and/or  (lower right panel). B and C, effects of stable expression of p16 on senescence in response to shRNA knockdown (B) or ABC294640-mediated inhibition (C) of SPHK2 were measured by SA-␤-gal staining in A549 cells compared with Scr-transfected and/or vehicle-treated controls. D and E, effects of stable expression of p16 on telomere damage (D). Images were quantified by ImageJ. Data are means Ϯ S.D. from three independent experiments and analyzed by Student's t test (n ϭ 3, *, p ϭ 0.006; **, p ϭ 0.002) or caspase-3 activation (E) in A549 cells, and in the absence/presence of ABC294640 they were measured using TIF assay or caspase-3 activation assay by immunofluorescence and confocal microscopy. Images were quantified by ImageJ (right panels). Data are means Ϯ S.D. from three independent experiments and analyzed by Student's t test (n ϭ 3, *, p ϭ 0.0006; **, p ϭ 0.0005; ***, p ϭ 0.0004). F, shRNA-mediated knockdown of p16 was confirmed using qPCR in H1341 cells compared with Scr-shRNAtransfected controls. Data are means Ϯ S.D. from three independent experiments, analyzed by Student's t test (n ϭ 3, *, p ϭ 0.00005). G, caspase-3 activation was measured by immunofluorescence using anti-caspase-3 antibody that detects active caspase-3 in H1341 with/without p16 knockdown compared with Scr-shRNA-transfected controls in the absence/presence of ABC294640 (right panel). Effects of ectopic expression of p16 on caspase-3 activation in the presence/absence of ABC294640 was measured in A549 cells compared with vector-transfected controls (left panel). H, expression levels of p16 protein in different lung cancer cell lines were confirmed by Western blotting using anti-p16 antibody (upper panel). Actin was used as a loading control (lower panel).

p16 -caspase-3 complex and senescence
dihydroceramide accumulation in nuclear membranes need to be determined.
It was interesting that SphK2 Ϫ/Ϫ mice showed accelerated telomere damage/senescence and aging phenotype after four generations. This was somewhat expected as mTERT Ϫ/Ϫ showed signs of telomere damage and aging phenotype after generation 4 also (48). Interestingly, although genetic loss of SphK2 induced senescence and aging due to telomerase instability and increased telomere damage without induction of apoptosis in SphK2 Ϫ/Ϫ compared with WT and SphK1 Ϫ/Ϫ knockout mice, down-regulation and/or inhibition of SPHK2/S1P/ TERT signaling resulted in TCF21-dependent caspase-3 activation downstream of telomere damage in lung cancer cells.
Our data further suggested that activation of p16 was the key to decide the fate of cells to succumb in senescence rather than apoptosis, as ectopic expression of p16 in p16-deficient cancer cells induced senescence by preventing caspase-3 activation. These data are consistent with a role of p16 in the protection of ATR4-mediated DNA damage in response to dysfunctional telomeres (49). Our data are also in agreement that, in addition to p53 signaling, p16 is another effector protein for the telomere damage signaling (50). Moreover, recent studies suggest that p16-expressing cells shorten a healthy life span (51) and that targeting p16 to clear senescent cells delays aging-related disorders detected in adipose tissue, skeletal muscle, and eye (52) or rejuvenates aged hematopoietic stem cells in mice (53).

Figure 10. Interaction between p16 and caspase-3 prevents apoptosis and induces senescence in response to telomere damage.
A, association between endogenous p16 and caspase-3 was measured using PLA in H1341 cells in the absence/presence of Scr-or p16-shRNAs. Quantification of PLA images was performed using the PLA software as described by the manufacturer. Data are means Ϯ S.D. from three independent experiments, analyzed by Student's t test (n ϭ 3, *, p ϭ 0.0349). Successful knockdown of p16 using shRNA was confirmed by qPCR in H1341 cells compared with Scr-shRNA-transfected cells (right panel). Data are means Ϯ S.D. from three independent experiments, analyzed by Student's t test (n ϭ 3, *, p ϭ 0.004). B and C, interaction and colocalization of p16 and caspase-3 in response to SPHK2 inhibitor ABC294640 (ABC) in the absence/presence of ectopic expression of WT-p16 (p16) and/or shRNA-dependent knockdown of SPHK2 (shSPHK2) were measured using PLA. Data are means Ϯ S.D. from three independent experiments, analyzed by Student's t test (n ϭ 3, *, p ϭ 0.045; **, p ϭ 0.094; ***, p ϭ 0.005) (B) or confocal microscopy and immunofluorescence (C) in A549 cells. Data are means Ϯ S.D. from three independent experiments, analyzed by Student's t test (n ϭ 3, *, p ϭ 0.484; **, p ϭ 0.001; ***, p ϭ 0.0009, ****, p ϭ 0.001). Vehicle-treated, vector-only, or Scr-shRNAtransfected cells were used as controls. Data represent at least three independent studies. Quantification of PLA images were performed using the PLA software as described by the manufacturer. Quantification of colocalization was performed using ImageJ. shRNA-mediated knockdown of SPHK2 and ectopic expression of WT-p16 were confirmed using qPCR (upper panels) and Western blotting, respectively (lower panels). Data are means Ϯ S.D. from three independent experiments, analyzed by Student's t test (n ϭ 3, *, p ϭ 0.049). D, molecular modeling for the predicted interaction between p16 and caspase-3 was generated using the already existing NMR structure of p16 and crystal structure caspase-3 as described under "Materials and methods." Ser 152 of p16 (turquoise) and Gly 251 of caspase-3 (gray) were among key residues for the interaction between these two proteins. E, association between p16 and caspase-3 was measured by PLA in A549 cells expressing WT-p16 or p16 S152A in the absence/presence of ABC294640. Vector-only-transfected and vehicle-treated cells were used as controls. Data are means Ϯ S.D. from three independent experiments, analyzed by Student's t test (n ϭ 3, *, p ϭ 0.010; **, p ϭ 0.009; ***, p ϭ 0.009). Quantification of PLA images was performed using the PLA software as described by the manufacturer. F and G, effects of ABC294640 exposure on caspase-3 activation were measured by confocal microscopy and immunofluorescence using anti-caspase-3 antibody that detects activated caspase-3 in A549 cells expressing WT-p16 or p16 S152A (F). Data are means Ϯ S.D. from three independent experiments and analyzed by Student's t test (n ϭ 3, *, p ϭ 0.0001; **, p ϭ 0.643; ***, p ϭ 0.0006). Vector-only-transfected and vehicle-treated cells were used as controls. Quantification of colocalization was performed using ImageJ. Expression of WT-p16 and p16 S152A was confirmed by Western blotting (G). Vector-transfected cell extracts were used as controls. H, association between p16 and caspase-3 was measured by PLA in testes tissues obtained from WT (SphK1 ϩ/ϩ /SphK2 ϩ/ϩ ), SphK1-deficient (SphK1 Ϫ/Ϫ ), or SphK2-deficient (SphK2 Ϫ/Ϫ ) mice at generation 6. Data are means Ϯ S.D. from three independent experiments, analyzed by Student's t test (n ϭ 3, *, p ϭ 0.002; **, p ϭ 0.001). Quantification of PLA images was performed using the PLA software as described by the manufacturer. Scale bars represent 100 m.

p16 -caspase-3 complex and senescence
These studies are consistent with our data, which provide a mechanistic link between p16 and inhibition of caspase-3 and apoptosis, supporting that targeting p16 induces clearance of senescent cells possibly by relieving caspase-3 from p16 for activation of caspase-3 and apoptosis. To this end, our data might have important clinical implications such that inhibition of p16 would switch senescent phenotype to apoptosis in tumors, improving tumor elimination and suppression, at least in part, in response to SPHK2/telomerase targeting.
Nevertheless, these data provide a novel link between the regulation of senescence versus caspase-3-dependent apoptosis, which is controlled by p16 signaling at least in response to telomere DNA damage evoked by SPHK2/S1P/TERT inhibition.

Cell lines and culture conditions
Cell lines were tested for mycoplasma contamination and authenticated by short-tandem repeat profiling (PowerPlex16HS) at the Genetica DNA Laboratories (Burlington, NC) in November, 2016. A549 cells (expressing K-ras mutant, WT p53) were grown in Dulbecco's modified Eagle's medium (Cellgro) with 10% fetal bovine serum (Atlanta Biologics) and 1% penicillin and streptomycin (Cellgro). H1299 (expressing WT K-Ras and lacking expression of p53 due to a homozygous gene deletion) and H1341 cells were grown in RPMI 1640 medium (American Type Culture Collection) with 10% fetal bovine serum and 1% penicillin and streptomycin. GM00847 cells were purchased from Coriell Cell Repositories. WT, SphK1 Ϫ/Ϫ , and SphK2 Ϫ/Ϫ MEFs were obtained from K. Argraves (Medical University of South Carolina). GM00847 cells and MEFs were cultured in Dulbecco's modified Eagle's medium and incubated at 37°C with 5% CO 2 . Primary human lung fibroblasts were purchased from Lonza, Inc., and cultured in fibroblast growth media as described (27).

Lung cancer-specific Q-PCR-based gene array
RT 2 profiler PCR array (PAHS-134C) from Qiagen was used for this purpose. Total RNA was isolated from shScr and shSPHK2 tumor tissues using the RNeasy kit (Qiagen) with DNase digestion step, and 1 g of purified RNA was used for cDNA synthesis using iScript cDNA synthesis kit (Bio-Rad). PCR array was carried out using Applied Biosystems (ABI 7300), and analysis was performed as per the manufacturer's instructions.

Immunohistochemistry and TMA
Lung adenocarcinoma TMA (LC1002) and small cell lung cancer TMA (LC10010) was purchased from US Biomax, and immunohistochemistry was performed with SPHK2 antibody (1:100) and hTERT antibody (1:50). Also, primary lung tumor tissues were obtained from the Medical University of South Carolina tumor biorepository core facility, and immunohistochemistry was performed by fixing with formalin and embedding with paraffin, and 5-m sections were made for immunohistochemistry. Pathology analysis and scoring was performed by a pathologist at the biorepository core facility, Medical University of South Carolina.

Western blotting
1 ϫ 10 6 A549 cells stably overexpressing either SPHK2 WT or SPHK2 G212E were plated in a 100-mm Petri dish overnight. The cells were harvested and centrifuged at 1300 rpm for 3.5 min and washed with 1ϫ PBS (pH 7.4) (Gibco). Cells were lysed with 1ϫ CHAPS lysis buffer (10 mM Tris-HCl (pH 7.5), 1 mM MgCl 2 , 1 mM EDTA, 0.1 mM benzamidine, 5 mM ␤-mercaptoethanol, 0.5% CHAPS, 10% glycerol), including a protease inhibitor mixture (Sigma) for 20 min on ice. Cell lysates were centrifuged at 12,000 ϫ g for 15 min at 4°C, and supernatants were used for Western blotting. SDS-PAGE was carried out using the Bio-Rad Criterion apparatus, followed by semi-dry transfer onto a polyvinylidene difluoride membrane. The membrane was blocked with 5% milk ϩ 0.1% Tween 20 in 1ϫ PBS (pH 7.4). Primary antibodies were used at 1:1000 dilution overnight at 4°C, followed by rabbit or mouse secondary antibodies (Jackson ImmunoResearch) conjugated with horseradish peroxidase at room temperature for 1 h. The rabbit secondary antibody was used at 1:2500 for detecting hTERT and at 1:5000 dilution for the rest of the antibodies. Actin was used as an internal control for Western blotting. The membranes were washed either with 1ϫ PBS with Tween 20 or 1ϫ TBS with 0.1% Tween 20 and developed using ECL plus chemiluminescence detection kit (GE Healthcare). 50,000 cells were plated in each of 4-well chamber slides overnight. Following treatment with ABC294640 (80 M for 8 h), the cells were fixed in 4% paraformaldehyde in 1ϫ PBS (pH 7.4) for 15 min at room temperature. The slides were washed three times with 1ϫ PBS (pH 7.4) and blocked with 1% goat serum in 0.3 M glycine, 1% BSA, and 0.1% Tween 20 in 1ϫ PBS (pH 7.4), for 2 h. The cells were then incubated with primary antibodies that recognize ␥-H2AX (5 g/ml; ab2893, Abcam) and TRF-2 (5 g/ml; IMG-124A, Imgenex) for 18 h. The slides were then washed three times with 1ϫ PBS (pH 7.4), and incubated with secondary antibodies containing Alexa 488, Alexa 594, and Cy5 fluorophores against ␥-H2AX and TRF-2, respectively, for 2 h at 20-25°C. The slides were then washed three times with 1ϫ PBS (pH 7.4), and 4Ј,6-diamidino-2-phenylin-p16 -caspase-3 complex and senescence dole containing mounting media was added and sealed with glass coverslips (colocalization resulted in either purple or yellow/orange images). Images were captured using IF-CM, and colocalization quantified using Pearson's correlation coefficient was used to calculate colocalization efficiency, and p Ͻ 0.05 was considered significant by ImageJ Fiji software (27).

Measurement of caspase-3 activation and apoptosis
Active caspase-3 was measured using cleaved caspase-3 antibody (2305-PC-020) from Trevigen Inc. Immunofluorescence was performed using Olympus FV10i microscope with 594and 488-nm channels for visualizing red and green fluorescence. Images were taken at ϫ60 magnification. At least three random fields were selected for images. Caspase-3/7 activity was measured in A549 cells and mouse embryonic fibroblasts by caspase-3 fluorometric assay kit (BF-1100, R&D Systems) as per the manufacturer's instructions. Moreover, A549 cells were pretreated with Z-DEVD-FMK (caspase-3-specific inhibitor, FMK-004, R&D Systems) at 10 M for 2 h prior to ABC294640 treatment and caspase-3 activity measurement. Also, caspase-3 activity was measured using caspase-3 fluorometric kit following pretreatment with 5 M S1P. TUNEL assay, DeadEnd TM Colorimetric TUNEL System (G7360, Promega), was used for measuring TUNEL-positive cells in mouse tissues as per the manufacturer's instructions.

Serum starvation and cell cycle analysis by flow cytometry
A549 cells (1 ϫ 10 6 ) were seeded in 100-mm dishes and incubated at 37⌼C for 16 h before being switched to serum-free media for 48 h. Then, cells were either treated with DMSO or 80 M ABC294640 for an additional 8 h. Cells were collected by trypsinization and washed twice in 1ϫ PBS. The cells were then fixed for 30 min at 4°C in cold 70% ethanol (added dropwise to the pellet while vortexing to ensure fixation of all cells and to minimize clumping). The pellets were spun at 850 ϫ g for 5 min and rinsed twice with 1ϫ PBS. The cells were then treated with RNase, and 5 g/ml propidium iodide was added before analysis by FACS.

Measurement of sphingolipids using MS by lipidomics
Sphingolipids in cells (1 ϫ 10 6 ) were measured using MS by the lipidomics shared resource facility at the Medical University of South Carolina, as we described previously (54). The lipid abundance was normalized to inorganic phosphate (P i ),

Ultrastructure analysis using TEM
Formalin-fixed shScr and shTCF21 tumors treated with ABC294640 or vehicle control were used. After post-fixation in 2% (v/v) osmium tetroxide, specimens were embedded in Epon 812, and sections were cut orthogonally to the cell monolayer with a diamond knife. Thin sections were visualized in a JEOL 1010 TEM (55,56).

Detection of senescence
A549 cancer cells and mouse embryonic fibroblasts (WT, SphK1 Ϫ/Ϫ , and SphK2 Ϫ/Ϫ ) at passage 6 were plated in 60-mm Petri dishes for 18 h followed by detection of senescence using senescence-associated ␤-gal assay kit as described by the manufacturer (Cell Signaling Technology) (27).

Measurement of senescence, aging, and tumor suppression in mice
Mouse testes tissues at different mating generations and tumor xenografts were fresh-frozen, and 5-m tissue sections were cut and stained for senescence using senescence-associated ␤-gal assay kit as described by the manufacturer's instructions (Cell Signaling Technology). SA-␤-gal activity was also measured for paw tissues of mice from indicated genotypes using the same kit. All animal experiment protocols were approved by the Institutional Animal Care and Use Committee at the Medical University of South Carolina (27, 54 -56).

Orthotopic lung tumor development in SCID mice
SCID mice were purchased from Harlan Laboratories. Ageand sex-matched mice (female SCID mice, n ϭ 7) were used for this experiment. Mice were injected (100 l) with 1 million A549-luciferase cells through the tail vein and were allowed to colonize lungs. Ten days after tail vein injection, the mice were anesthetized with isoflurane, and D-luciferin was administered by intraperitoneal injection, and luciferase bioluminescence was measured using Xenogen IVIS 200 Bioluminescence and Fluorescence Imaging System. Once luciferase expression was measured in the lungs of the animals, they were segregated into two groups; one group received vehicle control (50% DMSO and 50% PEG-200) and the other received ABC294640 (100 mg/kg body weight) for 16 days. At the end of the experiment, the mice were euthanized; lung tissues were collected, and Ki-67 and hTERT levels were measured by immunohistochemistry.

Xenograft-derived tumors in SCID mice expressing stable shSPHK2
A549 cells stably expressing SCR-shRNA or shRNA against SPHK2 (TRCN00000036973) were used for the experiment. Two million cells were implanted (100 l) into both flanks of SCID mice (n ϭ 4 female SCID mice per group/one tumor on each side of the flank), and when the tumors were palpable, the mice were either treated with vehicle control or with ABC294640 at 100 mg/kg body weight for 21 days. Tumor volume was measured every other day using digital calipers, and treatments were performed using oral gavage method every day. At the end of the experiment, the mice were euthanized, and tumor tissues were collected (27, 54 -56).

Xenograft-derived tumors in SCID mice expressing stable shTCF21
A549 cells stably expressing control shRNA or shRNA against TCF21 (TRCN0000015564, Open Biosystems) were used for the experiment. Two million cells (100 l) were implanted into both flanks of female SCID mice (n ϭ 4 mice per group/one tumor on each side of the flank), and when the tumors were palpable, the mice were either treated with vehicle control or with ABC294640 at 100 mg/kg body weight for 21 days. Tumor volume was measured every other day using digital calipers, and treatments were performed using oral gavage p16 -caspase-3 complex and senescence method every day. At the end of the experiment the mice were euthanized, and tumor tissues were collected (27, 54 -57).

Expression and purification of caspase-3
WT human caspase-3 was expressed in Escherichia coli and purified using nickel-nitrilotriacetic acid-agarose affinity resin, as described previously (58,59). Briefly, C-terminal His 6 -tagged plasmid pET23b-Casp3-His was purchased from Addgene (plasmids 11821 and 90090, respectively) and transformed into E. coli strain C41(DE3). A single colony was used to inoculate 50 ml of LB containing 50 g/ml ampicillin and grown overnight at 37°C. This culture was diluted 1:2000 into 4 ϫ 250 ml of LB media and incubated at 37°C until the A 600 reached 3-5. Rich media were exchanged to a minimal formulation suitable for high-density IPTG induction by spinning the cells gently at 5000 ϫ g for 10 min at 20 -25°C and then resuspending in an equal volume of minimal media. Cultures were incubated at 37°C until the A 600 increased by at least 1 unit, and the incubator was lowered to 18°C, and protein expression was induced with 1 mM IPTG for 16 h at 170 rpm. The cells were pelleted at 10,000 ϫ g for 20 min and were resuspended in lysis/binding buffer (buffer A: 20 mM HEPES (pH 7.5), 1 M NaCl, 20 mM imidazole, and 10 mM ␤-mercaptoethanol). Following sonication and centrifugation at 18,000 ϫ g for 30 min, the cleared lysate was applied to an equilibrated 1-ml HisTrap FF column (GE Healthcare, catalogue no. 17-5255-01), washed with 10 column volumes of wash buffer (10% buffer B: 20 mM HEPES (pH 7.5), 1 M NaCl, 250 mM imidazole, and 10 mM ␤-mercaptoethanol), and eluted by a linear gradient of imidazole from 10 to 70% over 40 column volumes using an ÄKTA start FPLC (GE Healthcare). The eluted fractions were analyzed by SDS-PAGE, and fractions containing caspase-3 were supplemented with 5 mM tris(2-carboxyethyl)phosphine, concentrated to 2 ml, and applied to a HiPrep TM 16/60 Sephacryl TM S-200 HR size-exclusion column in SEC buffer (buffer C: 50 mM HEPES (pH 7.5), 300 mM NaCl, 10 mM ␤-mercaptoethanol 1 mM EDTA, and 10% glycerol). Fractions containing pure caspase-3 as determined by 4 -20% SDS-PAGE were pooled, quantified, and stored at Ϫ80°C as 1-ml aliquots.

Molecular modeling and Open SPR for measurements of protein-protein interactions
The existing Protein Data Bank files containing NMR structure of p16 (2A5E) and X-ray crystallography structure of caspase-3 (5IBP) from the RCSB were transferred into ZDOCK Server (http://zdock.umassmed.edu/). 4 The top model was then used to predict the sites of association between p16 and caspase-3, as we described previously (57).
The protein-protein association between recombinant purified human p16 (Novus Biologicals, NBP2-35199) and caspase-3 was measured using an Open SPR-localized surface plasmon resonance biosensor (Nicoya Life Science, Inc., Kitchener, Canada), as described previously (60). In short, 100 l (50 g/ml) of p16 Ink4a , purchased from Novus Biologicals (NBP2-35199), was immobilized on a COOH sensor chip (Nicoya SEN-AU-100 -12-COOH) at a flow rate of 20 l/min in 1ϫ PBS buffer (pH 7.4) and 0.1% v/v Tween 20. Free activated carboxyl groups were deactivated with the addition of 100 l of blocking buffer (Nicoya). The immobilized ligand was washed with analyte running buffer (1ϫ PBS (pH 7.4); 0.1% v/v Tween 20; 0.1% w/v BSA) until a stable baseline was achieved. Buffer-matched recombinant caspase-3 (80 l; 50 -1000 nM) analytes were injected into the flow cell at a rate of 20 l/min. Following a 4-min interaction time, the dissociation was recorded for an additional 16 min. The ligand-immobilized sensor chip was regenerated between each analyte injection using 20 mM HCl at a 150 l/min flow rate. Kinetic binding analysis was performed with the TraceDrawer software package (Ridgeview Instruments, Uppsala, Sweden). Sensorgram traces were fit to a 1:1 Langmuir model to derive association (K a ), dissociation (K d ), and affinity (K D ) constants.

Statistical analyses
All data are presented as mean Ϯ S.D., and group comparisons were performed with a two-tailed Student's t test. Wilcoxon rank sum test and paired t test were used to analyze small-cell and non-small cell tumor microarray data. Tumor xenograft experiments were analyzed using two-way ANOVA with Tukey's post hoc test, and testes weights from G4 and G5 mice were analyzed using one-way ANOVA with Tukey's post hoc test. The results were analyzed using GraphPad Prism 7 software, and p Ͻ 0.05 was considered statistically significant (27, 54 -56). For the comparison of several groups, a variance analysis (ANOVA) was carried out under normal distribution assumption.