IMR-90 human fibroblasts (ATCC) were cultured in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and penicillin/streptomycin. Unless otherwise stated, all non-senescent cells were quiescent for this study. Quiescence was induced by reducing FBS to 0.2% for at least 48 hours, and senescent cells were similarly treated prior to analysis. Senescence was induced by 10 Gy ionizing radiation [SEN(IR)], lentiviral shRNA-mediated depletion of sirtuin 3 [shSIRT3], lentiviral overexpression of constitutively active HRAS [RasV12] (RRID : Addgene_22262), or 10 days of continuous culture in 1 mM sodium butyrate (NaBu) or 500 nM antimycin A (MiDAS). Bleomycin-treated cultures were described previously (40). Briefly, cells were treated with either 100 ug/mL bleomycin or a matching volume of PBS stock in growth media for 3 hours at 20% oxygen. Media was then changed and cells were returned to 3% oxygen. Bleomycin-treated cells were analyzed 10 days after treatment. All cultures were considered senescent if they showed a <5% EdU labeling and > 75% senescence-associated beta-galactosidase positivity. GSE-22 (RRID : Addgene_22253) expressing cells were described previously (41). Conditioned media were generated by 24 hours of continuous culture in serum-free DMEM. All cells were confirmed mycoplasma-free.

RNA was isolated from 200,000-500,000 cells using the Isolate II RNA Extraction Kit (Bioline). RNA (250 ng/μl) was used to synthesize cDNA using a High Capacity cDNA synthesis kit (Thermo) according to the manufacturer’s instructions. Gene expression was analyzed by qPCR using universal probe library primers (Roche) previously described (36) or listed below and a LighterCycler 480 II Real Time PCR System (Roche). RNA levels were normalized to beta-actin. Primer sequences were: NAMPT forward – aagggatggaactacattcttga, Reverse – ctgtgttttccaccgtgaag, UPL Probe #6; NMNAT1 forward – gaaatccctagagccaaaaaca, reverse – ggaacagcaaaggactccaa, UPL Probe #43; NMNAT2 forward – gatcctgctgctgtgtggta, reverse - cctccatatctgcctcgttc, UPL Probe #67; NMNAT3 forward – cgtttccctctcggacct, reverse – ctgctatttgcagggctca, UPL Probe #3; NAPRT1 forward – tgctctgcctggtcagcta, reverse – tctagcagccgcttctctg, UPL Probe #64; NMRK1 forward – tcctgactattccatatgaagaatgta, reverse – tggaggctgatagacccttg, Probe #63; NADK forward – acgctgctgtacgcttcc, reverse - agctgaatggggtcagga, UPL Probe #37.

NAD was measured using a commercial kit (Biovision) according to the manufacturer’s instructions. Extractions were performed using 500,000 cells per sample homogenized in 500 μL of DNA lysis buffer, and fractionated using 10 kDa cutoff filters (Millipore) spun at 10,000 × g for 45 min. Standard curves (5–200 pg/ml) were generated for quantification.

Cells were lysed in 5% SDS in 10 mM Tris, pH 7.4, and protein content determined by BCA assay. Ten micrograms of cell lysates were loaded per well. For conditioned media, 200,000 cells were cultured in 1 mL serum-free DMEM for 24h. Media was concentrated 20-fold using 30 kDa cutoff filters spun at 10,000 x g for 20 minutes, and 100,000 cell equivalents of media were added per lane in the absence (non-reducing) or presence (reducing) of 2-mercaptoethanol. Antibodies were (Cell Signaling Technology Cat# 2531, RRID: AB_330330), (Abcam Cat# ab4074, RRID: AB_2288001) and (Abcam Cat# ab45890, No RRID). Western densitometry was quantified using ImageJ.

For human cell culture supernatants, ~200,000 cells equivalents per mL were analyzed using a human PBEF/Visfatin ELISA (BioVision) according to the manufacturer’s instructions. For mouse sera, 50 μL of serum were diluted 1:1 with Assay Buffer and analyzed by mouse/rat eNAMPT Dual ELISA (Adipogen) according the manufacturer’s instructions.

Senescence was induced in IMR90 fibroblast cells by three different stimuli: irradiation (IR;10 Gy X-ray), doxorubicin (DOXO; 250 nM, 24 hr treatment), and mitochondrial dysfunction induced senescence (MiDAS; Antimycin A; 500 nM, continuous) and separately induced quiescent as control as described previously (36, 42) SASP EV extraction was performed using size-exclusion chromatography (SEC) and ultrafiltration as detailed previously (43). Briefly, EV proteins were extracted using a SDS-based buffer, and quality check was performed by western blotting with exosome protein-specific antibodies. EV proteins were then reduced and alkylated using S-Trap mini (Protifi, Farmingdale, NY) followed by on-column digestion with trypsin (1:20 (w/w) enzyme:protein ratio). Peptides were desalted using stage tips, vacuum dried, and resuspended in aqueous 0.2% formic acid. Finally, indexed Retention Time Standards [iRT, Biognosys, Schlieren, Switzerland) (44)] were added to each sample according to the manufacturer’s instructions for mass spectrometry-based quantitative analysis.

LC-MS/MS analyses were performed on a Dionex UltiMate 3000 system coupled online to an Orbitrap Eclipse Tribrid mass spectrometer (Thermo Fisher Scientific, San Jose, CA). Samples were acquired in data-independent acquisition (DIA) mode using a 210-min chromatographic gradient. Full MS1 spectra were collected at 120,000 resolution (AGC target: 3e6 ions, maximum injection time: 60 ms, 350-1,650 m/z), and MS2 spectra at 30,000 resolution (AGC target: 3e6 ions, maximum injection time: Auto, NCE: 27, fixed first mass 200 m/z). The DIA precursor ion isolation scheme consisted of 26 variable windows covering the 350-1,650 m/z mass range with a window overlap of 1 m/z ( Supplementary Table S1 ) (45). DIA data were processed in Spectronaut v15 (version 15.1.210713.50606; Biognosys) using directDIA. Data was searched against the Homo sapiens proteome with 42,789 protein entries (UniProtKB-TrEMBL), accessed on 12/07/2021. Protein identification was performed requiring a 1% q-value cutoff on the precursor ion and protein level (experiment), and 5% q-value cutoff on the protein level. Quantification was based on the peak areas of the best 3-6 MS2 fragment ions, with no normalization, and iRT profiling and q-value sparse data filtering applied. Differential protein expression analysis was performed using a paired t-test, and p-values were corrected for multiple testing, specifically applying group wise testing corrections using the Storey method (46). NAMPT (P43490) protein was identified and quantified with 4 unique peptides.

Experiments were conducted at Joslin Diabetes Center with approval of its Animal Care and Use Committee. Mice were kept on a 12-hour light/dark cycle with water and food ad libitum. DIO very high fat diet (VHFD) 60kcal% fat (Fisher Scientific) was administered during 4 or 8 weeks to C57Bl6/J 8-week-old male mice acquired from Jackson Labs. Mice were treated with ABT-263 (Selleck Chemicals, in ethanol:polyethylene glycol 400:Phosal 50 PG) or vehicle. ABT-263 was administered to mice by gavage at 50 mg/kg body weight per day (mg/kg/d) for 4-5 d per cycle, with a week between the cycles during 8 weeks. For the 4 weeks HFD experiment, the period between ABT-263 cycles was increased to two weeks.

Genotoxic stress is a common inducer of both senescence and NAD depletion. To evaluate the relationship between genotoxic stress, NAD, and senescence - we irradiated human IMR-90 fibroblasts with 10 Gy of ionizing radiation (IR) and measured both NAD and NAMPT RNA levels over the next 24-48 hours. IR promoted an immediate acute loss of NAD over the first 3 hours, in agreement with previous studies (47, 48), followed by a gradual recovery until reaching pre-irradiation levels by 24 hours ( Figure 1A ). Since NAMPT is the rate-limiting enzyme in NAD salvage, we hypothesized that this recovery might be due to elevation of NAMPT levels. In agreement with this, NAMPT RNA levels increased over the first 3 hours and rapidly returned to pre-IR levels by 24 hours ( Figure 1B ). Since senescence is a chronic condition, and previous reports indicate that NAMPT is elevated during senescence (37), we also measured NAMPT and NAD levels in senescent cells 10 days after IR. Senescent cells had elevated NAMPT levels ( Figure 1C ), but this was not linked to an increase in NAD ( Figure 1D ). To determine if this was a common feature of senescent cells, we also assayed senescence induced by overexpression of constitutively active RAS (RasV12) ( Figures 1E, F ) and mitochondrial dysfunction-associated senescence (MiDAS) induced by shRNA-mediated depletion of sirtuin 3 (shSIRT3) ( Figures 1G, H ). In each case, senescence was accompanied by increased NAMPT levels, but no commensurate increase in NAD levels. Indeed, RAS-induced senescence was associated with lower levels of NAD ( Figure 1F ), despite strongly elevated NAMPT levels. We also measured RNA levels of other NAD synthetic enzymes at 3 hours and 10 days after IR, but only observed small (<2 fold) increases in these, and none showed the strong increases with senescence that we observed for NAMPT ( Figure 1I ).

Loss of NAD results in AMPK activation (36, 37), and poly-ADP ribose polymerase (PARP) activity during recovery from genotoxic stress also elevates AMP levels and AMPK activation (48). We therefore sought to assess the role of AMPK in elevation of NAMPT in genotoxic stress and senescence. AMPK was rapidly phosphorylated following IR, and this phosphorylation decreased over time ( Figure 2A ). To address the role of AMPK in NAD recovery, we treated irradiated cells with an AMPK inhibitor (Compound C) (49) or vehicle (DMSO) and measured both NAD ( Figure 2B ) and NAMPT RNA levels ( Figure 2C ). Compound C treatment prevented recovery of NAD levels following irradiation ( Figure 2B ), and this was coupled to a failure to elevate NAMPT levels ( Figure 2C ). Therefore, AMPK activity is required for NAMPT elevation following genotoxic insult. Conversely, 24 hours of compound C treatment had no effect on NAMPT levels once senescence was fully established 10 days after irradiation ( Figure 2D ). Thus, distinct mechanisms elevate NAMPT during the acute DNA damage response and during chronic senescence.

Since NAMPT was elevated during senescence, but NAD levels were not, we considered the possibility that NAMPT is secreted by senescent cells. We first analyzed our previously reported single cell gene expression dataset (40) for NAMPT expression ( Figure 3A ). Unlike cyclin-dependent kinase inhibitors and housekeeping genes, SASP factors show increased variance with senescence – with only a subset of cells showing clear elevation (40). NAMPT expression was similar to that observed with SASP factors such as IL8 or IL1A - suggesting that it might be secreted. We therefore measured extracellular NAMPT (eNAMPT) levels in conditioned media from senescent cells by ELISA. Cells induced to senesce by IR ( Figure 3B ), mitochondrial dysfunction ( Figure 3C ), and HDAC inhibition (sodium butyrate – NaBu) ( Figure 3D ) all released eNAMPT. In agreement with other inducers of senescence, NaBu–induced senescent cells also did not have elevated NAD levels ( Figure 3E ).

Release of many SASP factors is restrained by the activity of p53 (41, 50), though DAMPs such as HMGB1 are released in a p53-dependent manner during senescence (51). We therefore used a dominant negative p53 genetic suppressor element (GSE22) (52) to eliminate p53 activity and assess its role in eNAMPT release. Like many proinflammatory SASP factors, such as IL-6, elimination of p53 activity elevated eNAMPT release ( Figure 3F ). Conversely, loss of p53 activity lowered NAD levels, regardless of senescence inducer status ( Figure 3G ). Thus, eNAMPT release is regulated in a manner similar to proinflammatory SASP factors such as IL-6 or IL-1B.

eNAMPT is released in multiple forms. For example, eNAMPT is released as a catalytically active dimer in extracellular vesicles (EVs) from tissues such as visceral fat (22). This form of eNAMPT can be endocytosed by recipient cells, where NAMPT then elevates NAD levels (22). This has been shown to antagonize age-associated NAD loss, prevent diabetes, and extend lifespan in mice (22). Alternatively, eNAMPT has been reported to occur outside of EVs, where it can bind TLR4 and act as a DAMP to drive inflammation (53). The DAMP form of eNAMPT is also at least partly monomeric, as opposed to the catalytically active version which appears as a disulfide-linked NAMPT dimer (34, 35). We therefore analyzed conditioned media from non-senescent or senescent cells by western blot using reducing and non-reducing conditions, as described previously (34). Virtually all eNAMPT detected in non-reducing conditions was dimerized ( Figure 3H ), while reducing conditions showed only monomer, as expected. Under non-reducing conditions, virtually all eNAMPT was detected in conditioned media from senescent cells, and virtually none from non-senescent cells ( Figures 3H, I ).

We previously used mass spectrometry to identify new SASP factors as part of a large proteomic SASP Atlas (54). In these datasets, which also used IMR-90 fibroblasts, eNAMPT was found exclusively in the soluble secretome of conditioned media from senescent cells induced by IR or RAS (54), and no eNAMPT was found in exosomes. Since we expected at least some eNAMPT in EVs, we analyzed an additional proteomic dataset from EVs from either non-senescent cells or cells induced to senesce by IR, doxorubicin (DOXO), or mitochondrial dysfunction (MiDAS). EVs from senescent cells induced by DOXO or MiDAS showed depleted levels of eNAMPT relative to quiescent cells, although IR-induced senescence did not show significant changes ( Figure 3J ). Regardless of inducer, no EVs from senescent cells showed increases in eNAMPT, even though total eNAMPT is increased with senescence. These data indicate that eNAMPT is primarily released from senescent cells as a soluble dimer.

Diabetes is causally linked to the release of eNAMPT. Notably, EV-contained eNAMPT antagonizes diabetic phenotypes, whereas the soluble monomeric form has been shown to promote diabetes (21, 34, 35). Since senescent cells can promote diabetes, we hypothesized that they might be a source of eNAMPT during metabolic stress. To determine if pancreatic beta cells also elevate NAMPT during senescence in vivo, we analyzed our previously generated RNA-seq data from either beta-galactosidase positive or beta-galactosidase negative pancreatic beta cells from 7-8 month old mice for NAMPT expression ( Figure 4A ) ( 8). We then analyzed sera from diabetic mice that were given a diabetes-inducing high fat diet (HFD) treated with either the senolytic drug ABT-263 (ABT) (14) or a vehicle for 4 weeks ( Figure 4B ) or 8 weeks by gavage ( Figure 4C ) for eNAMPT by ELISA. Animals from the 8-week study were previously described (8). In either treatment protocol, HFD elevated eNAMPT levels, and these were lowered by ABT-263 treatment. Together, these data indicate that senescent cells are a source of eNAMPT during diabetes.