Discovery of piperlongumine as a potential novel lead for the development of senolytic agents

Accumulating evidence indicates that senescent cells play an important role in many age-associated diseases. The pharmacological depletion of senescent cells (SCs) with a “senolytic agent”, a small molecule that selectively kills SCs, is a potential novel therapeutic approach for these diseases. Recently, we discovered ABT-263, a potent and highly selective senolytic agent, by screening a library of rationally-selected compounds. With this screening approach, we also identified a second senolytic agent called piperlongumine (PL). PL is a natural product that is reported to have many pharmacological effects, including anti-tumor activity. We show here that PL preferentially killed senescent human WI-38 fibroblasts when senescence was induced by ionizing radiation, replicative exhaustion, or ectopic expression of the oncogene Ras. PL killed SCs by inducing apoptosis, and this process did not require the induction of reactive oxygen species. In addition, we found that PL synergistically killed SCs in combination with ABT-263, and initial structural modifications to PL identified analogs with improved potency and/or selectivity in inducing SC death. Overall, our studies demonstrate that PL is a novel lead for developing senolytic agents.


INTRODUCTION
Cellular senescence, an essentially irreversible arrest of cell proliferation, can be triggered when cells experience a potential risk for malignant transformation due to the activation of oncogenes and/or DNA damage [1][2][3][4][5][6][7]. While eliminating aged or damaged cells by inducing senescence is an effective barrier to tumorigenesis, the accumulation of senescent cells (SCs) over time compromises normal tissue function and contributes to aging and the development of ageassociated diseases [6,8,9]. Often, SCs secrete a broad spectrum of pro-inflammatory cytokines, chemokines, growth factors, and extracellular matrix proteases, a feature collectively termed the senescence-associated secretory phenotype. These factors degrade the local tissue environment and induce inflammation in various tissues and organs if SCs are not effectively cleared by immune system [6,[8][9][10][11].
Studies have shown that the genetic clearance of SCs extends the lifespan of mice and delays the onset of several age-associated diseases in both progeroid and naturally-aged mice [12][13][14][15]. It has also been shown that rapamycin and metformin increase lifespan in mice and marmoset monkeys, by suppressing the induction of senescence [16][17][18][19][20]. These findings support the hypothesis that SCs play a causative role in aging and age-associated diseases [6,21,22] and, importantly, highlight the tremendous therapeutic potential of pharmacologically targeting SCs [23,24]. Consistent with these findings, we have shown that ABT-263 (navitoclax), an inhibitor of the antiapoptotic Bcl-2 family proteins, acts as a potent senolytic agent to deplete SCs in vivo and functionally rejuvenates hematopoietic stem cells in both sublethally irradiated and naturally-aged mice [25]. Complementary studies from other labs have confirmed that the Bcl-2 protein AGING (Albany NY) family is a promising molecular target for the development of senolytic drugs [26,27]. These studies further establish the concept that the pharmacological depletion of SCs is a promising, novel approach for treating age-associated diseases [28].
ABT-263 was identified by screening a small library of structurally diverse, rationally-selected small molecules that target pathways predicted to be important for SC survival [25]. By titrating their cytotoxicity against normal human WI-38 fibroblasts and ionizing radiation (IR)-induced senescent WI-38 fibroblasts, this targeted screen also identified the promising senolytic agent piperlongumine (PL, Fig. 1A); PL is a natural product isolated from a variety of species in the genus Piper [29]. Here, we report the characterization of PL as a potential novel lead for the development of senolytic agents.

Piperlongumine is a potential senolytic agent
Because we identified PL as a potential senolytic agent by screening a library of rationally-selected compounds with IR-induced senescent WI-38 fibroblasts, we tested its ability to selectively kill senescent human WI-38 fibroblasts induced by different means. PL exhibited moderate selectivity in reducing the viability of IRinduced WI-38 SCs (IR-SCs) compared to nonsenescent WI-38 cells (NCs) (Fig. 1B and Table 1), and PL induced cell death in a time-dependent manner (Fig.  1C). We also assessed the survival of WI-38 cells in which senescence was induced by replicative exhaustion or by ectopic expression of the oncogene Ras (Fig.  1B). Replicative WI-38 SCs, which were previously

AGING (Albany NY)
shown to be more resistant to ABT-263 [25], were slightly more sensitive to PL (Fig. 1B) than IR-and Ras-induced SCs. The mechanisms underlying the difference of SCs induced by different stimuli have yet to be elucidated.
Structurally, PL contains two electrophiles, the C2-C3 and C7-C8 α,β-unsaturated imides, both of which are important for the toxicity of PL in cancer cells [30]. Thus, we investigated whether the integrity of the twoelectrophile system was also important for the ability of PL to kill SCs. Consistent with the findings in cancer cells, 2,3-dihydro-PL and 7,8-dihydro-PL ( Figure 1A), in which the C2-C3 olefin or the C7-C8 olefin was saturated, respectively, showed little or no senolytic activity toward IR-SCs (Fig. 1D).

Piperlongumine induces apoptosis in SCs
Next, we investigated the mechanism by which PL selectively kills SCs. Because PL induces apoptosis in cancer cells [31][32][33][34][35][36][37][38][39][40][41], we hypothesized that the same is true for SCs. We used Annexin V and propidium iodide staining and subsequent fluorescence-activated cell sorting to detect apoptosis, respectively, in senescent WI-38 cells. PL treatment increased the number of Annexin-V-positive cells in SCs by 5.5-fold when compared to the vehicle group ( Fig. 2A). To further confirm that PL killed cells by apoptosis, we treated IRinduced WI-38 SCs with the pan-caspase inhibitor Q-VD-OPh (QVD) [42] to inhibit apoptosis. Ten µM QVD, in the presence of PL, significantly reduced apoptosis and partially rescued SCs from PL-induced death ( Fig. 2A, B). In addition, western blot analysis showed elevated levels of activated caspase-3 and degradation of poly(ADP-ribose) polymerase (PARP) in PL-treated IR-SCs (Fig. 2C), confirming the apoptotic cell-death mechanism. Furthermore, PL had no effect on the levels of receptor-interacting protein kinase 1 and 3 (RIP1 and RIP3), indicating that PL did not induce necroptosis in IR-SCs (Fig. 2D) [43].

Piperlongumine kills senescent cells through an ROS-independent mechanism
Initially, PL has been proposed to selectively induce cancer cell death by increasing reactive oxygen species (ROS) production, based on the observation that PL elevates cellular ROS levels in various cancer cells, but not in normal cells [31]. However, structural modifications to PL have revealed that there is no correlation between a PL analog's ability to increase ROS and its toxicity toward cancer cells, leading to the conclusion that ROS-independent mechanisms are also involved in cancer cell death [30]. We hypothesized that the same scenario is true for the PL-induced killing of SCs. We used the non-fluorescent ROS indicator dihydrorhodamine 123 (DHR 123), which can passively diffuse across membranes where it is oxidized to green fluorescent rhodamine 123 in the presence of ROS, and flow cytometry to determine if PL increased ROS in IR-SCs. Treatment with 10 μM PL for 6 or 24 h significantly elevated ROS levels in IR-SCs compared to vehicle-treated IR-SCs or non-senescent WI-38 cells with PL treatment, whereas IR-SCs have a higher baseline level of ROS (Fig. 3A). In addition, similar to the results obtained in cancer cells [31, 33-35, 41, 44-47], co-treatment with 2 mM N-acetyl-L-cysteine (NAC), an antioxidant, fully reversed PL-induced ROS elevation and cell death (Fig. 3B), suggesting that the selective induction of ROS in SCs may be the basis for the senolytic activity of PL. However, a number of small molecules, including hydrogen peroxide, parthenolide, arsenic trioxide, phenethyl isothiocyanate, auranofin [a thioredoxin reductase inhibitor], buthionine sulfoximine [a γ-glutamylcysteine synthetase inhibitor], and decyl-triphenylphosphonium, that were previously shown to kill cancer cells by inducing oxidative stress were not able to selectively kill IR-SCs [25]. This finding suggests that ROS elevation alone is insufficient to selectively kill SCs. Interestingly, it has been reported that PL can chemically react with the sulfhydryl group of methyl thioglycolate to form the product of conjugate addition at C3 [30]. Based on this, we hypothesized that NAC, rather than acting as an ROS scavenger, inactivates PL through a similar reaction in cell culture media. In support of this hypothesis, we observed that PL (10 µM) disappeared within 2 h after co-incubation with NAC (2 mM) in cell culture media under conditions mimicking a cell viability assay (Fig. 3C), forming the corresponding hetero-conjugated product, NAC-PL (Fig. 3C). NAC-PL was isolated from this reaction, and it exhibited diminished toxicity toward IR-SCs (Fig. 3C). To further investigate the role of ROS in PL-induced SC death, we treated IR-SCs with PL in the presence of a different, potent antioxidant, γ-tocotrienol (GT3, 5 µM) [48]. GT3 did not decrease PL-induced cell death in IR-SCs, although GT3 reversed the PL-induced increase in ROS (Fig. 3D). These data suggest that ROS were not involved in the SC death induced by PL.

AGING (Albany NY)
To determine if PL-based senolytic agents with increased potency/selectivity could be developed, we synthesized a series of PL analogs that have been reported as potent anti-cancer agents. No obvious correlation between ROS induction and senolytic potency was observed in these analogs. Specifically, BRD4809 [49], an abbreviated PL analog (Fig. 3E), and PL-DI [30], a PL dimer (Fig. 3F), showed unchanged or increased potency against IR-SCs, respectively, compared to PL; however, these analogs did not affect ROS levels in SCs at concentrations near their EC 50 values for SC viability (Table 2). On the other hand, PL-FPh [30], which contains an alkenyl substituent at C2 of PL, selectively induced ROS production in IR-SCs and had increased potency and selectivity in killing these SCs when compared to PL (Fig. 3G). Finally, PL-7 [30], a PL analog with an enlarged ring, inhibited ROS production, yet retained the senolytic potency of PL toward IR-SCs (Fig. 3H). Taken together, these results further confirm that PL and its analogs kill SCs in an ROS-independent manner.

The synergistic senolytic effect of piperlongumine and ABT-263
PL has been tested for its synergistic anti-tumor effect in combination with TNF-related apoptosis-inducing

AGING (Albany NY)
combinations. An additive effect for the combination of 10 µM PL and 0.08 µM ABT-263 (CDI = 0.99) was observed. The CDI values for the other combinations ranged from 0.02-0.41, indicating that PL and ABT-263 exerted a strong synergistic senolytic effect on IR-SCs. It is worth noting that, in our previous studies, increasing the concentration of ABT-263 from 1.25 µM to 5 µM did not increase cell killing in WI-38 SCs. PL appeared to eliminate the subpopulation of IR-SCs that was resistant to ABT-263.

DISCUSSION
Selective depletion of SCs is a potentially novel antiaging strategy that may prevent cancer and various human diseases associated with aging and rejuvenate the body to live a longer, healthier life. As such, several senolytic agents, including ABT-263, have been identified recently [23,[25][26][27], demonstrating the feasibility of pharmacologically targeting SCs. However, ABT-263 induces thrombocytopenia [53], and it remains to be determined whether ABT-263 can be used to safely treat age-related diseases, since individuals may require long-term treatment with a senolytic drug. Thus, it is necessary to identify a safer senolytic drug. In the present study, we evaluated PL as a novel senolytic agent. PL induced caspase-mediated apoptosis in SCs and effectively killed SCs induced by IR, replicative exhaustion, or ectopic expression of the oncogene Ras. Similar to the observations in cancer cells versus normal cells, PL elevated ROS levels in IR-SCs, but not in non-senescent WI-38 cells.
Because NAC blocks PL-induced ROS elevation in cancer cells and abolishes the anti-tumor effect of PL, it has been proposed that induction of ROS production is a key mechanism of PL-induced cancer cell apoptosis [31, 33-35, 41, 45-47]. Indeed, we showed that cotreatment of IR-SCs with PL and NAC fully reversed the senolytic effect of PL. However, we found that PL was chemically inactivated by NAC in culture media through a conjugated addition reaction between the sulfhydryl group of NAC and the C2-C3 α,βunsaturated imide group of PL; the resulting adduct, NAC-PL, was not senolytic. Based on these findings, caution is warranted when using NAC or similar compounds that contain a nucleophilic sulfhydryl group, such as dithiothreitol [54], as an ROS scavenger to study ROS inducers such as PL. In contrast, GT3, a potent ROS scavenger that does not react with PL, effectively blocked PL-induced ROS elevation but had no inhibitory effect on PL-induced SC death. In addition, through evaluation of PL analogs, we found that there is no correlation between senolytic potency and ROS-induction in IR-SCs. These results led us to conclude that the senolytic activity of PL is ROS independent.
Unlike ABT-263, the precise mechanism of action (MOA) by which PL induces SC apoptosis remains unclear. PL modulates the activity of many cell signaling and survival pathways in cancer cells [31], and a number of studies have investigated the MOA by

AGING (Albany NY)
which PL induces apoptosis in these cells [30, 35-40, 45, 54-69]. Data from these studies may be translatable to PL-induced SC apoptosis because SCs and cancer cells share some common pro-survival pathways [23]. In addition, mass spectrometry-based proteomic approaches using probes derived from PL could be used to identify direct molecular target(s) of PL in SCs. In this regard, novel anti-senescent protein target(s) and MOAs could be identified, making it possible to develop promising novel classes of senolytic agents.
Importantly, PL appears to be safe; the maximum tolerated dose in mice is very high, and it maintains high bioavailability after oral administration [31]. Furthermore, our initial structural modifications to PL demonstrate that we can develop PL analogs with increased potency and selectivity toward SCs (Fig. 3), supporting the use of PL as a lead for further drug discovery and development.
Another potential use of PL and its derivatives is in combination with ABT-263, or other inhibitors of Bcl-2 family proteins, for a synergistic senolytic effect. Although ABT-263 is a highly specific senolytic agent, it causes transient thrombocytopenia and neutropenia in patients [70]; this results from its inhibitory effect on Bcl-xL, which is important for platelet survival [71,72]. We showed that PL had a strong synergistic effect on the senolytic activity of ABT-263 in vitro, potentially reducing the dose of ABT-263 needed to effectively deplete SCs. We expect this therapeutic approach would significantly reduce ABT-263-induced thrombocytopenia, making senolytic treatment with ABT-263 safer.
Although clearance of SCs with a senolytic drug may be used to treat some age-related diseases, it is well recognized that cellular senescence is also functionally linked to many beneficial physiological processes, such as wound healing, tissue remodeling, and embryonic development [73]. Attempt to clear SCs in certain situations may produce some side effects. Therefore, we should proceed with caution to use senolytic drugs to treat age-related diseases before we have a better understanding of their risks. Low-passage WI-38 (< 25 passages) cells were used as controls or for the induction of senescence.

Ionizing radiation-induced senescence
To induce senescence with ionizing radiation (IR), WI-38 cells, roughly 70% confluent, were exposed to 15 Gy of IR in a J.L. Shepherd Model Mark I 137 Cesium Ƴirradiator (J.L. Shepherd, Glendale, CA) at a dose rate of 1.080 Gy/min. Three days after irradiation, cells were passaged once at a 1:3 dilution. WI-38 cells became fully senescent 10 d after irradiation.

Cell viability assays
Cell viability was measured with flow cytometry, as previously described [25].

Calculation of EC 50 values
Dose-response curves were generated for each senolytic agent, and the half-maximal effective concentrations (EC 50 values) were calculated with GraphPad Prism 6 software.

ROS assay
Control, non-senescent WI-38 cells were plated in 24well plates (

Statistical analysis
The data displayed normal variance. The data were analyzed by analysis of variance (ANOVA) with Graphpad Prism from GraphPad Software (San Diego, CA). In the event that ANOVA justified post hoc comparisons between group means, the comparisons were made with Neuman-Keuls or Tukey's multiplecomparisons test. P < 0.05 was considered to be significant. CDI was calculated as: CDI = AB/ (A × B). AB represents the percent of viable cells remaining after the treatment with the combined drugs, while A and B represent the percent of viable cells remaining after the treatment with each drug independently. CDI < 1 indicates a synergistic effect, CDI = 1 indicates an additive effect, and CDI > 1 indicates antagonism. CDI < 0.7 indicates that the drugs are significantly synergistic.

CONFLICTS OF INTEREST
Y.W., J.C., X.L., G.Z., and D.Z. filed a patent application for the use of PL and PL analogs as antiaging agents. A potential royalty stream to Y.W., J.C., X.L., G.Z., and D.Z. may occur consistent with University of Arkansas for Medical Sciences policy. G.Z. is a consultant and D.Z. is a co-founder and advisor of UNITY Biotechnology that develops senolytic drugs.

FUNDING
This work was initially supported in part by grants from the National Institutes of Health (R01CA122023 and P20GM109005) and a scholarship from the Arkansas Research Alliance. The development of various PL analogs was supported by a sponsored research agreement between UNITY Biotechnology and the University of Arkansas for Medical Sciences. This manuscript was edited in the Office of Grants and Scientific Publications at the University of Arkansas for Medical Sciences.