Polyamines Control eIF5A Hypusination, TFEB Translation, and Autophagy to Reverse B Cell Senescence

1Immune senescence is characterized by the failure of lymphocytes to respond adequately to infection, malignancy, and vaccination. During a regular influenza season, about 90% of the deaths occur in people older than 65 years. Immune responses to vaccines are known to be particularly ineffective in the elderly population (>65 years of age), and yet some vaccines, such as those for influenza, are given primarily to the elderly. A major correlate of protection for vaccinations is the specific antibody titer generated by long-lived plasma B cells. With a lifespan of several decades, long-lived lymphocytes are particularly prone to accumulation of intracellular waste. Autophagy recycles unwanted cytoplasmic material. Autophagy-deficient lymphocytes are unable to generate adequate responses, in particular long-lived lymphocytes, memory T and B cells, and plasma B cells. Reversing immune aging would also open opportunities to improve management of age-related morbidities. However, little is known about how immune senescence can be reversed.

2Genetic activation of autophagy extends lifespan in mice. However, only a few autophagy-inducing drugs reverse aging, one being rapamycin, an inhibitor of mTOR. Because of the unwanted effects of mTOR inhibition, there is a need to better understand mTOR-independent control of autophagy for drug development. Furthermore, to test anti-aging drugs, blood biomarkers are critical. As an endogenous polyamine metabolite that declines with age, spermidine may be key in controlling cellular aging via autophagy. Previously we found in aged mice that spermidine improves memory T cell responses, and showed cardiac function was restored with spermidine, both in an autophagy-dependent manner. Here we investigated whether spermidine is able to rejuvenate long-term B cell responses, the main correlate of protection for vaccination. We found an autophagy-dependent improvement of B cell responses by spermidine in old mice, indicating an effect across different immune cell types. We further identified the mechanism by which it regulates autophagy. Spermidine post-translationally modifies (hypusinates) eIF5A, which regulates protein synthesis and autophagy in primary B cells. Because of the rigid structure of proline, triprolines slow down translation and require hypusinated eIF5A to form peptide bonds more effectively. We find that this is specifically required for the synthesis of the autophagosomal and lysosomal master regulator TFEB, which is a short-lived protein containing one triproline motif in mouse and two in human. Moreover, this pathway is considerably downregulated in immune cells from old individuals, making its components suitable biomarkers. Importantly, spermidine improves hypusination of eIF5A, TFEB protein expression, autophagic flux, and antibody titers in old human B cells, which has direct translational relevance.

3Previously, we demonstrated that autophagic flux is decreased with age in human and murine T cells. Here, we first investigated the extent of the decrease across bone marrow progenitors and splenic mature hematopoietic cells, with a method of quantifying endogenous LC3-II, a marker for autophagosomes, using flow cytometry that we adapted for primary hematopoietic cells. To measure active autophagic flux, we used bafilomycin A1 (BafA1), which is a lysosomal V-ATPase inhibitor that prevents acidification of lysosomes and their degradative functions, thereby causing accumulation of autophagic substrates, including LC3-II (Figure 1A). Hematopoietic stem cells (HSCs) were found to have the highest basal autophagy levels compared with other hematopoietic cells. Although their autophagic flux was mildly reduced in old mice, it was significantly decreased in B and T cells from old mice as compared to young mice (Figure 1A; gating strategy in Figures S1A and S1B). Treatment of mice for 6 weeks with spermidine increased autophagic flux in most hematopoietic cell types tested (Figure 1A). We then tested if the in vivo administration of spermidine alleviates typical hallmarks of hematopoietic aging, including expansion of phenotypic HSCs, myeloproliferation, and lymphopenia. We observed a significant increase of phenotypic HSCs and Lin−Sca1+cKit+ cells (LSKs) in old mice, which spermidine administration did not affect (Figures 1B and S1C). A myeloid-biased phenotype, including increased myeloid cells and more significantly diminished mature B and T cells, was found in spleens of old mice (Figures 1C and S1D). Consistently, multiple bone marrow myeloid progenitors are expanded (Figures 1D, S1E, and S1F). Bone marrow pro-B cells (CD43+) also mildly accumulate (Figures 1E, 1F, S1G, and S1H), in line with the paradigm that pro-B cell maturation is blocked with age. Six weeks of treatment with spermidine does not affect these aging phenotypes in either spleen or bone marrow (Figures 1C–1F and S1D–S1H), although longer treatment may confer a more significant effect, as lifespan extension has previously been demonstrated with 6 month treatment.

4Consistent with the flow cytometry staining of LC3-II (Figure 2A), reduced autophagic flux was confirmed in old mature B lymphocytes by western blot (Figure 2B) and by confocal microscopy using GFP-LC3 transgenic mice (Figure 2C). Old B lymphocytes accumulated LC3-II, which did not further increase by BafA1 treatment (Figures 2A–2C). This indicates that autophagic flux is impaired in old cells at the level of the lysosome.

5We then examined whether spermidine improves B cell function. In mice older than 22 months, the antibody response to immunization with the model antigen NP-CGG was markedly reduced, as expected, and administration of spermidine in drinking water significantly improved IgG1 responses (Figure 2D). In contrast, Atg7-deleted B cells showed defective memory responses that mimic the aging phenotype, but spermidine failed to promote their responses (Figures 2E, S2A, and S2B). However, it should be noted that this does not exclude alternative rejuvenation mechanisms of spermidine in addition to autophagy induction. At the time of sacrifice, we found very small numbers of long-lived bone marrow NP-specific plasma cells in old mice, which were significantly restored with spermidine treatment (Figure 2F). Spermidine did not induce autophagy in B cells or T cells or improve the antibody responses in young mice (Figures S2C–S2F). Thus, spermidine restored autophagic flux and the responses of B lymphocytes in old mice in vivo, and its immune-boosting effect is lost in autophagy-deficient B cells.

6We next investigated how spermidine regulates autophagy. First, we confirmed 100 μM spermidine to be the optimal concentration to induce autophagy in the mammalian lymphocytic Jurkat cell line (Figures S3A and S3B;, ). Several signaling pathways have been described downstream of spermidine, including inhibition of histone acetyltransferases (HATs). Therefore, we assessed whether spermidine inhibits HAT activity and thereby induces ATG7 mRNA expression, as shown in yeast. In none of three approaches (HAT colorimetric assay, ATG7 qPCR, or acetylated H3 pull-down) did we find that spermidine had this effect (Figures S3C–S3E). The acetyltransferase p300 has been shown to directly acetylate certain autophagy proteins, such as ATG7 to inhibit autophagy, and spermidine was reported to inhibit p300 activity. However, ATG7 acetylation was not affected by two tested concentrations of spermidine within 6 h of treatment (Figures S3E and S3F). Spermidine also failed to affect the activity of AMPK, as assessed using AMPK phosphorylation (Figure S3G). However, we found that spermidine induces cellular stress (Figure S3H) at the autophagy-inducing dose of 100 μM or higher, demonstrated by increased levels of ATF4 (Figure S3I), phosphorylated eIF2α (Figure S3J), and increased expression of CHOP mRNA (Figure S3K) in Jurkat cells. Furthermore, high-dose spermidine induced cell death after 24 h (Figure S3L), presumably because spermidine levels are high in transformed cell lines, and further uptake is toxic and non-specifically induces cellular stress responses, likely also inducing autophagy. Cellular stress was also induced by spermidine in NIH 3T3 cells as assessed by increased ATF4 expression (Figure S3M), indicating that high-dose spermidine induces cellular stress across different cell types.

7Therefore, in subsequent experiments, rather than adding spermidine, we opted for the depletion of spermidine either by genetic knockdown (KD) of the key spermidine-synthesizing enzyme, ornithine decarboxylase (ODC), or with the ODC inhibitor difluoromethylornithine (DFMO) in NIH 3T3 cells (Figure 3A). Indeed, genetic KD of Odc with small interfering RNA (siRNA) (Figure S4A), or DFMO treatment, reduces spermidine levels substantially, while supplementing cells with spermidine rescues its levels partially, as measured using gas chromatography-mass spectrometry (GC-MS) in cell lysates (Figure 3B). This demonstrates, first, that cultured cells actively synthesize spermidine via ODC and, second, that exogenous spermidine is efficiently taken up to rescue its intracellular levels. Next, we examined pathways downstream of spermidine that had not been previously linked to autophagy. In eukaryotic cells, spermidine is a unique substrate for the hypusination of the translation factor eIF5A (Figure 3A; ). To date, eIF5A is the only known protein containing the unusual amino acid hypusine. Upon KD of Odc in NIH 3T3 cells, or addition of DFMO, we found reduced LC3-II levels and reduced eIF5A hypusination, while supplementation with spermidine rescued both (Figures 3C, S4B, and S4C). Similarly, reduced LC3-II levels were observed when Eif5a was knocked down (Figures 3D and S4D–S4F). We next depleted the two hypusinating enzymes, deoxyhypusine synthase (DHS) and deoxyhypusine hydroxylase (DOHH), to study their effects on autophagy. The KD of Dhs and knockout (KO) of Dohh, or GC7, a specific inhibitor of DHS, reduced eIF5A hypusination and LC3-II levels (Figures 3E, S4G, and S4H). Although exogenous spermidine restored LC3-II levels in cells depleted of spermidine, inhibition of eIF5A hypusination with siDhs or GC7 abrogated this effect (Figures 3F and S4I). Taken together the data indicate that spermidine maintains cellular basal autophagy by hypusinating eIF5A.

8In activated T cells, eIF5A is one of the 20 most abundant proteins. However, little is known about its role in lymphocytes. We measured eIF5A levels and autophagy in primary B cells during activation. LC3-II levels increased substantially over time upon activation in B cells, as measured using flow cytometry (Figure 3G) and western blot (Figures 3H and S4J). This time-dependent increase correlated with increasing levels of both total and hypusinated eIF5A (Figure 3H). In line with our data in NIH 3T3 cells, LC3-II levels decreased with increasing doses of GC7 (Figure 3I). This pathway operates independently of mTOR, whose activation inhibits autophagy. Indeed, GC7 did not activate mTOR, as demonstrated by S6 phosphorylation, a downstream substrate of mTOR (Figure S4K). Overall the data indicate that the eIF5A-autophagy pathway is induced upon activation of primary B cells, and physiological levels of hypusinated eIF5A are required for efficient autophagy.

9We next addressed how eIF5A regulates autophagy by identifying changes in expression of proteins involved in the autophagy pathway upon inhibition of eIF5A hypusination. We performed label-free quantitative mass spectrometry (MS) on nuclear and cytoplasmic fractions of activated primary B cells treated with GC7 (Figures 4A and S5A). This was followed by stable isotope labeling with amino acids in cell culture (SILAC) on activated primary B cells (Figures 4B and S5B). It has recently been reported that eIF5A regulates ATG3 protein synthesis in cell lines. However, ATG3 protein levels were not found reduced in primary B cells treated with GC7, as shown in our MS results (Figure 4A) and confirmed by western blot (Figure S5C). Of the autophagy-related proteins detected, nuclear TFEB was the only one repeatedly found decreased upon GC7 treatment in both approaches. TFEB is a key transcription factor regulating autophagosomal and lysosomal biogenesis. The effect of inhibition of eIF5A hypusination by GC7 on overall and nuclear TFEB was confirmed by western blot in activated primary B cells (Figures 4C and 4D). Accordingly, gene expression of several TFEB targets was markedly decreased upon GC7 treatment. Genetic KD of Tfeb was sufficient to reduce autophagic flux in NIH 3T3 cells (Figures 4F and S5D), as reported. KD of Eif5a and Dhs, or GC7 treatment in NIH 3T3 cells, led to a decrease in TFEB protein (Figures 4G and 4H). To test if TFEB is affected by spermidine depletion, we inhibited spermidine synthesis in NIH 3T3 cells by siOdc or DFMO. This caused a reduction in TFEB, which was rescued by exogenous spermidine administration (Figures 4I and S5E), suggesting that spermidine controls cellular TFEB levels. However, spermidine failed to maintain TFEB levels when Eif5a was knocked down or eIF5A hypusination was inhibited (Figures 4J, S5F, and S5G), indicating that the regulation of TFEB by spermidine is mediated via eIF5A hypusination.

10Being a paralog of TFEB, TFE3 has also been reported to promote autophagy and lysosomal biogenesis. Although TFE3 was not detected in either MS experiments (Figures 4A and 4B), western blots show that TFE3 was also reduced by GC7 treatment in B cells (Figure S5H). However, knocking down Tfe3 alone did not significantly compromise autophagy or further reduce autophagy when knocked down together with siTfeb in NIH 3T3 cells (Figure S5I). Thus we focused on TFEB hereafter.

11TFEB is a very short-lived protein (<4 h half-life) compared with other proteins in the cell (median 36 h half-life) and is therefore expected to require more active translation than most proteins. We first investigated if eIF5A controls overall protein synthesis in lymphocytes, as previously reported. We measured translation by flow cytometry with O-propargyl-puromycin (OPP) and found that inhibition of eIF5A hypusination by GC7 caused a 30% reduction in protein synthesis rate in activated primary B cells (Figure 5A). A similar reduction was observed using a puromycin-independent assay with L-azidohomoalanine (AHA; a methionine analog) incorporation (Figure S5J). To further assess translational changes caused by GC7, we performed ribosome profiling in activated primary B cells. Increased ribosome occupancy at the start codon was observed by GC7 treatment (Figure S5K). However, whether the accumulation of reads at the start codon is an artifact of cycloheximide treatment during sample preparation will require further validation. Stalling of ribosomes at the triproline-encoding motif (PPP) was found in activated primary B cells (Figure S5L), as observed in yeast and bacteria, indicating that polyproline is a conserved ribosome-pausing motif across kingdoms. However, GC7 treatment did not further increase the ribosome occupancy at PPP motifs (Figure S5M). One explanation is that the no-go decay mechanism of mRNA at the severely stalled regions may lead to loss of reads.

12We next investigated whether TFEB synthesis was affected by GC7 treatment by detecting nascent (recently made) proteins. By pulling down nascent proteins after 4 h of AHA labeling, a significant reduction of nascent TFEB was observed (Figures 5B and S5N). Interestingly, nascent eIF5A was also mildly reduced, suggesting an auto-regulatory loop of eIF5A at the translational level (Figures 5B and S5N). This is not reflected in overall eIF5A levels, probably because of the short treatment and the long half-life of eIF5A. Both human and mouse TFEB have at least one triproline motif (Figure 5C). Indeed, sequences around this triproline are also ribosome-pausing motifs (SPP and PPV), suggesting that this triproline-containing motif may confer TFEB with the specific requirement to rely on eIF5A for smooth translation. Therefore, we examined if the triproline motif in murine and human TFEB is sufficient to affect translation rate as regulated by hypusinated eIF5A. We generated three different constructs, using labile mCherry expression to report on the synthesis of either the TFEB triproline-containing motif, 13 consecutive prolines (13Pro), or a random sequence (Figure 5C). After transfection into NIH 3T3 cells, we measured translation using the ratio of the reporter mCherry to the transfection control GFP. As expected, the mCherry/GFP ratio was mildly reduced by GC7 for the random sequence and reduced by nearly half for the 13Pro motif, while the synthesis of TFEB triproline motif was also significantly inhibited (Figure 5D), indicating that the TFEB triproline motif requires hypusinated eIF5A for efficient synthesis. Moreover, mutating the triproline motif (PPP to AAA) partially rescued the expression of TFEB in the absence of hypusinated eIF5A (Figure 5E), indicating that the triproline motif is one, but not the only mechanism accounting for GC7-induced TFEB reduction. Overexpression of the triproline-mutated TFEB induced autophagy to a similar extent as wild-type TFEB (Figure 5F). However, when the nuclear localization sequence (NLS) was mutated in addition to triprolines, autophagy induction was lost. Taken together this indicates that the polyproline mutation does not compromise the function of TFEB as a transcription factor. Moreover, overexpressing mutant TFEB was sufficient to rescue autophagy in GC7-treated cells (Figure 5G), indicating that TFEB is critical for eIF5A-regulated autophagy. These data suggest that hypusinated eIF5A directly facilitates the synthesis of TFEB partially via the triproline-containing motif. Future studies are required to reveal other mechanisms of eIF5A-regulated TFEB expression.

13eIF5A and its hypusination are essential for cellular growth in cultured cells and early embryonic development in mice, but its tissue-specific function is still unclear, although inducible whole-body KO of Dhs in mice leads to a fatal wasting syndrome. To investigate the function of hypusinated eIF5A during B cell development and activation in vivo, we generated competitive mixed bone marrow chimeras with inducible deletion of the hypusinating enzyme DHS. Dhs deletion was induced after long-term engraftment and B cell lineage reconstitution of CD45.2+ cells was examined on days 8 and 30 after deletion (Figure S6A). Percentages of CD45.2+ of both transitional and mature circulating B cells in peripheral blood were significantly affected by Dhs deletion on day 30 (Figures S6B and S6C). Upon sacrifice of the mice on day 34, we investigated if this was due to a loss of progenitors in the bone marrow. Unexpectedly, all multipotent hematopoietic progenitors investigated (HSC, MPP, LMPP, and LSK) were severely affected by DHS depletion (Figure S6D). This is in line with a bone marrow hypocellularity reported previously in whole-body Dhs-KO mice. Consistent with depleted HSCs, splenic CD45.2+ myeloid cells, which mostly lack self-renewal capacity and rely on replenishment from progenitor cells, were also severely depleted after Dhs deletion (Figure S6E). Similarly, early B cell progenitors (pre-pro B cells), newly formed B cells and mature B cells from bone marrow as well as splenic transitional, marginal zone and follicular B cells were found significantly diminished (Figures S6F–S6H). The reduced B cell progenitors may result from either loss of HSCs or intrinsic survival defects. We checked if the remaining CD45.2+ cells were indeed deleted for Dhs in blood (Figure S6I), bone marrow, and spleen (Figure S6J). This analysis revealed that although on day 8 after deletion with tamoxifen, blood cells were adequately deleted in the floxed Dhs mice, in cells that survived to day 34, Dhs was not deleted in most mice (Figure S6J). The overall cellularity in bone marrow and spleen was not affected despite the reduced CD45.2+ cell numbers (Figure S6K), indicating that the remaining CD45.1+ and non-deleted CD45.2+ cells expanded and filled up the niche. These data together demonstrate a profound requirement of the eIF5A pathway in hematopoiesis and long-term survival of mature B cells.

14To circumvent the defects of hematopoiesis and assess the role of eIF5A hypusination in B cell activation, we further attempted to delete Dhs after the first immunization with NP-CGG in Dhs-KO bone marrow chimeric mice supported with Rag1−/− bone marrow cells (to provide wild-type hematopoietic cells for the survival of the mice, without T and B cells). As before, after 3 weeks of deletion, all remaining lymphocytes expressed wild-type Dhs only. Levels of NP-specific IgG1 post-boost were not significantly changed in KO mice (data not shown), which can be attributed to contamination from non-deleted memory B cells. Overall the data imply that complete inhibition of eIF5A hypusination kills hematopoietic cells, and its effect on generation of antibodies cannot easily be investigated in vivo, because of (1) the rapid elimination of cells knocked out for Dhs and (2) the limiting deletion efficiency in mature B cells with inducible deletion models.

15To investigate if eIF5A has an effect on B cell activation, we induced B cell activation ex vivo with LPS while inhibiting hypusination with GC7. The upregulation of early activation markers, including CD86, CD69, and MHC-II (Figure S6L) was not changed upon GC7 treatment. However, activation-induced cell proliferation was severely impaired upon inhibition of hypusinated eIF5A (Figure S6M) in a time-dependent fashion, in line with the known function of eIF5A in supporting cellular growth. Taken together, these results indicate that hypusinated eIF5A is required for signaling events leading to specific cell fates such as proliferation and persistent antibody production for which autophagy might be required via its role in maintenance of organelle homeostasis.

16To assess the role of TFEB in murine B cells, we induced the KO of Tfeb in cultured murine B cells ex vivo (Figures 6A and S6N). Consistent with the reported functions of TFEB in inducing autophagy in other cell types (Figures 4F, 5F, S5D, and S5I; ), deletion of Tfeb led to reduced autophagic flux in B cells (Figures 6A and 6B) without reducing cell viability (Figure S6O). Also, the expression of activation markers CD86 and MHC II was not compromised (Figures 6C and S6P). Functionally, antibody production was significantly reduced by Tfeb KO as assessed by secreted IgM in culture supernatants (Figure 6D). Therefore TFEB regulates antibody production of B cells, but the exact mechanisms require further investigation.

17Spermidine levels have been reported to reduce with age in multiple organs, including mouse spleens. To assess whether the identified eIF5A-TFEB pathway is also affected during physiological aging and whether these changes can be rescued by exogenous spermidine, 2-year-old mice were treated with spermidine for 6 weeks. Interestingly, not only eIF5A hypusination but also overall eIF5A was significantly reduced in B cells from old mice (Figure 6E). The diminished overall eIF5A may partially result from a chronic decline in hypusination and eIF5A protein synthesis. Notably, a striking reduction of overall and especially nuclear TFEB was observed (Figures 6E and 6F), which might be a key cause of the reduced autophagic flux shown earlier (Figures 2A–2C). Importantly, spermidine supplementation restored the hypusination of eIF5A, overall eIF5A expression, and both overall and nuclear TFEB levels (Figures 6E and 6F). Thus, we observed an age-related decline of the spermidine-eIF5A-TFEB-autophagy pathway in B cells, which can be improved by spermidine supplementation. The fact that TFEB expression was only partially rescued by spermidine (Figures 6E and 6F) suggests the existence of alternative mechanisms regulating reduced TFEB expression during aging, which requires future investigation.

18To investigate if this pathway regulates aging in human B cells, we first measured the levels of hypusinated eIF5A and TFEB in peripheral blood mononuclear cells (PBMCs) from healthy donors of different ages. Consistent with our findings in mice, PBMCs from donors aged 65 years and older showed a distinct reduction of TFEB protein. In some aged donors, TFEB was even undetectable by western blot (Figure 7A). In contrast, the mRNA of TFEB was not reduced in PBMCs from old donors (Figure 7B), further indicating that the diminished TFEB protein expression in the elderly is due to post-transcriptional changes. The overall eIF5A was significantly diminished in human PBMCs from donors ≥65 years of age, but the remaining eIF5A was well hypusinated (Figure 7A), suggesting that cells may coordinate the expression of overall eIF5A with their hypusination levels during human aging. Indeed, TFEB protein levels correlate well with the expression of eIF5A (Figure S7A). Using mass spectrometry, we show that endogenous spermidine in human PBMCs declines with age (Figure S7B), as reported. Next, we investigated whether TFEB is regulated by spermidine and eIF5A in human primary B cells. Human B cells from young donors stimulated with anti-IgM and CD40L and treated with DFMO ex vivo showed reduced eIF5A hypusination and TFEB, which were rescued by exogenous spermidine supplementation (Figure 7C).

19Last and most important, we tested if spermidine improves old human B cell responses. In contrast to non-stimulated naive PBMCs, ex vivo activated old human B cells showed accumulated non-hypusinated eIF5A, suggesting a defect in hypusination (Figure 7D). The levels of both eIF5A hypusination and TFEB (Figures 7D and 7E), as well as autophagic flux (Figure 7F) were significantly restored by spermidine treatment in B cells from aged donors in which spermidine levels are naturally low. IgG production by B cells from aged donors was also improved (Figure 7G). GC7 treatment abrogated these effects of spermidine (Figures 7D–7G), suggesting that spermidine improves autophagy and antibody production of old B cells via hypusinating eIF5A. In contrast, neither the eIF5A-TFEB-autophagy pathway nor IgG production was increased by spermidine in young human B cells (Figures S7C–S7E). Thus, replenishment of spermidine in B cells from aged donors rejuvenated their function via eIF5A hypusination.

20This study demonstrates that autophagic flux is regulated translationally. Although acute autophagy induction may not require gene expression and protein synthesis, with components being readily available, long-term autophagic flux does depend on continued gene expression. We identified one specific mechanism of this regulation, whereby the polyamine metabolite spermidine is a substrate for the hypusination of the translation factor eIF5A, which in turn controls the synthesis of TFEB protein. This mechanism operates normally in the young, in whom spermidine is abundant. In B cells from old individuals, however, replenishing spermidine levels restored this pathway, thereby boosting responses in vivo in mice and ex vivo in humans. Depleted spermidine levels and subsequent low TFEB expression may be an important cause of the reduction in autophagy in the aging adaptive immune system, as well as in other tissues. However, other potential direct or indirect mechanisms may play a role in the anti-aging effects of polyamines, such as epigenetic modifications or changes in metabolism.

21It was reported that GC7 induces autophagy in the human fibrosarcoma 2fTGH cell line. The discrepancy between this and our findings may be a result of the high dose of GC7 (200 μM) used as opposed to 10 μM for primary B cells and 100 μM for NIH 3T3 cells used here. Treatment with a high drug dose may induce stress responses and autophagy non-specifically, as demonstrated here for spermidine (Figures S3H–S3M). In line, the study by shows that genetic ablation of eIF5A or of its hypusination enzymes consistently tends to reduce (not induce) LC3-II expression in 2fTGH cells.

22Translation in lymphocytes is little studied in T cells and B cells. Naive B cells are in a metabolically quiescent state, with low protein synthesis (low eIF5A expression) and autophagy, both of which are quickly upregulated after activation. Autophagy and the translation machinery may then form a regulatory loop to support the metabolic requirements of proliferation and immunoglobulin production. In this loop, autophagy provides substrates and energy for increased translation, while inducing the synthesis of certain autophagic proteins such as TFEB. Specifically, a particular requirement for rapid provision of amino acids and unfolded protein degradation by autophagy may exist in long-lived plasma cells, which secrete antibodies at the rate of about 2,000 molecules per second.

23The anti-aging function of TFEB seems to be highly conserved across species. Earlier studies found it to be an important transcription factor for lifespan extension in C. elegans. Moreover, it is likely that the control of TFEB via translation is a broad, maybe universal mechanism, operating across many tissue types (here in NIH 3T3 cells and in mouse and human primary B cells). TFEB is known to be post-translationally controlled via (1) mTORC1-dependent phosphorylation for cytoplasmic retention, and (2) calcium-dependent dephosphorylation to signal for nuclear translocation. We now add an mTOR-independent step of regulation at the level of translation.

24ATG3 protein synthesis was recently reported to be controlled by hypusinated eIF5A in cell lines. However, we could not find a reduction of ATG3 with GC7 treatment in primary B cells, indicating that it may be cell type specific (Figures 4A and S5C). MHC II also contains polyproline motifs but did not reduce with GC7 treatment either (Figure S6L). These observations suggest that the presence of a polyproline motif alone is insufficient to cause reduced protein expression by inhibition of hypusinated eIF5A. Indeed, research in bacteria showed that polyproline-induced stalls may lead to reduced protein levels only if they limit translation more than initiation. Moreover, an auto-regulatory loop exists for the regulation of Tfeb transcription. Therefore, defective TFEB synthesis may initiate a vicious cycle involving reduced Tfeb transcription (Figure 4E) and further reduced protein synthesis, eventually leading to overall diminished TFEB protein. Additionally, we hypothesize that a stronger stalling motif (such as SPPPVP in TFEB), and TFEB’s short half-life may contribute to reduced protein expression in the absence of hypusinated eIF5A, and predict that other such proteins are regulated by eIF5A.

25We have shown eIF5A and TFEB levels are correlated with age in immune cells. Blood aging biomarkers are urgently needed to gauge the efficacy of new drugs that prolong health span. Autophagy is one of the few general mechanisms underpinning many age-related diseases and is therefore a good target for anti-aging drugs. Last, it has been accepted for decades that hypusination occurs during or shortly after the synthesis of eIF5A, and the majority of cellular eIF5A is in its hypusinated form unless artificial treatments are being used to block the hypusination pathway. Here we show that under certain physiological conditions, such as aging, a significant pool of non-hypusinated eIF5A accumulates, probably because of the lack of spermidine (Figures 6E, 7D, and 7E). This indicates that hypusination of eIF5A can potentially be harnessed therapeutically for aging or age-related diseases.

26Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Anna Katharina Simon (katja.simon@kennedy.ox.ac.uk).

27Young (6-12 weeks) and old (22-24 months) C57BL/6J wild-type mice (RRID:IMSR_JAX:000664) were purchased from Charles River UK. CD45.1+ B6.SJL mice (RRID:IMSR_JAX:100012) were purchased from Charles River UK and bred in BSU at the Kennedy Institute of Rheumatology. Mb1-cre mice were a kind gift from M Reth, Atg7flox mice from M Komatsu (1 year old mice were used in this study). GFP-LC3 mice from N Mizushima and Stefan Balabanov made the bone marrow from inducible Dhsflox mice available. Tcfebflox mice from A Ballabio. CAG-cre/Esr1 mice from the Jackson lab. 5 mM Spermidine (Cayman Chemical) was administered to mice in the drinking water, changed three times per week. All mice were held at Biomedical Services, University of Oxford. Animal experiments were approved by the local ethical review committee and performed under UK project licenses PPL 30/2809 and then PPL 30/3388.

28Cells (Jurkat, NIH 3T3, HEK293T) were cultured in RPMI-1640 medium supplemented with 10% FBS, Penicillin-Streptomycin (P/S), and L-Glutamine (all Sigma) in 5% CO2 at 37°C. Stable transduction of the lentivirus to NIH 3T3 cells was performed as previously described. The IPTG-inducible MISSION shRNA lentiviral vector pLKO-puro-IPTG-3xLacO contains either the shRNA against the 3′-UTR of murine eIF5A mRNA (custom-made from #SHCLND-NM181582-TRCN0000125229, Sigma) or a corresponding non-target shRNA control (#SHC332-1EA, Sigma). 100 μM IPTG (Thermo Fisher) was used to induce the expression of shRNA. Immortalized Dohhflox/flox and Dohh+/+ 3T3 cells were established from mouse embryonic fibroblasts. 100 nM 4-OHT (H6278, Sigma) freshly prepared in ethanol was used to induce the knockout of Dohh and Tfeb.

2925 mL of blood was obtained from non-fasting healthy donors (old, ≥ 65 years of age; young, ≤ 65 years of age) in heparinized tubes. All subjects gave written informed consent. The study was approved by the Ethics Committee of the University of Oxford.

30Bone marrow cells were collected from the femur and tibia of the donor mice (Dhs+/+ or Dhs−/− CD45.2+, with young B6.SJL CD45.1+ mice as competitors) and frozen in cryopreservation medium for long-term storage. Carefully thawed cells were counted and mixed at the indicated ratio for intravenous (iv) injection. The recipient B6.SJL mice were lethally irradiated (550 cGy twice, 4 hours apart) and rested for 1 hour prior to injection. A total of 1.5 million bone marrow cells were injected intravenously into recipient mice. After at least 16 weeks of long-term reconstitution, tamoxifen (Sigma) dissolved in corn oil was administered via oral gavage (5 mg/mouse/day for 5 consecutive days) and peripheral blood and organs were collected for analysis.

31Mice were injected intraperitoneally (ip) with 50 μg NP-CGG (N-5055D-5, Biosearch Technologies) precipitated in Imject Alum adjuvant (Thermo Fisher) on day 0 followed by secondary immunization on or after day 42 (50 μg NP-CGG in PBS ip). Spermidine was given 2 days before the first immunization and then continuously in drinking water. Peripheral blood samples were collected from tail vein on indicated days.

32Murine splenic B cells were purified using the Pan B Cell Isolation Kit II (130-104-443, Miltenyi). Cells were cultured at a density of 1 million/ml supplemented with 10 μg/ml LPS (Santa Cruz) for stimulation. Medium was replaced on day 3 of culture by replacing half volume with fresh culture medium containing LPS followed by analysis on day 4 and day 5. For 4-OHT-inducible Tfeb knockout B cells, cells were stimulated with 10 ng/mL murine recombinant IL-4 (214-14, PeproTech) and 5 μg/mL anti-mouse CD40 antibody (Thermo Fisher) for 4 days.

33GC7 (Merck Millipore) was added for 24 hours to B cells (10 μM) on day 2 after B cell stimulation (unless otherwise indicated) or to Jurkat/NIH 3T3 cells (100 μM). 10 nM bafilomycin A1 (Cayman Chemical), 10 μg/ml cycloheximide (Sigma), or 100 nM Torin 1 (Cayman Chemical) were added to cells for 2 hours. 10 μM etoposide (Cayman Chemical) or 1 μM thapsigargin (Cayman Chemical) was added to cell culture for 6 hours, or 1 mM difluoromethylornithine (DFMO, Enzo Life Sciences) for 24 hours.

34For one-dimensional western blot, cells were lysed using NP-40 lysis buffer containing proteinase inhibitors (Sigma) and phosphatase inhibitors (Sigma) on ice. After spinning down the debris, protein concentration in the supernatant was quantified by BCA Assay (23227, Thermo Fisher). Reducing Laemmli Sample Buffer was then added to the supernatant and heated at 100°C for 5 minutes. 5-20 μg proteins were loaded for SDS-PAGE analysis. NuPAGE Novex 4%–12% Bis-Tris gradient gel (Thermo Fisher) with MES running buffer (Thermo Fisher) was used. To improve separation of LC3-I and LC3-II, 15% Tris-HCl gel and SDS running buffer was used. Proteins were transferred to a PVDF membrane (IPFL00010, Merck Millipore) and blocked with 5% skimmed milk-TBST. Membranes were incubated with primary antibodies dissolved in 1% milk overnight and secondary antibodies dissolved in 1% milk with 0.01% SDS for imaging using the Odyssey CLx Imaging System. Two-dimensional western blot was carried out as described before. Data were analyzed using Image Studio Lite or Fiji.

35After treatments an equal number of cells were collected, washed with PBS and lysed using NP-40 lysis buffer with proteinase and phosphatase inhibitors on ice for 20 minutes. 950 μL lysis buffer was used for 25 million Jurkat cells. Cell debris was centrifuged down. 50 μL supernatant was collected as whole cell lysate (input control), and the protein concentration was measured by BCA assay. The remaining supernatant was mixed with 15 μL Protein A Agarose beads (Thermo Fisher) in PBS, then rotated at 4°C for 15 minutes to pre-clear the lysate. The beads were removed by centrifugation at 6000 rpm for 1 minute. Immunoprecipitation (IP) antibody (AcK, ATG7, or rabbit IgG) and 20 μL Protein A agarose beads were added to the lysate and then rotated at 4°C overnight. The beads were washed twice with NP-40 lysis buffer containing proteinase and phosphatase inhibitors and the supernatant discarded. 35 μL 2X Reducing Laemmli Sample Buffer was added to the beads and heated at 100°C for 5 minutes, and supernatant used for western blot analysis.

36For NP-IgG1 ELISA, ELISA plates (675061, Greiner Bio-One) were coated with 5 μg/ml NP-BSA (N-5050H-10, Biosearch Technologies) in bicarbonate/carbonate buffer at 4°C overnight. After three washes with PBS, plates were blocked with 5% skimmed milk in PBS at 37°C for 1 hour, followed by 3 x PBS washes. Serum samples diluted in 1% milk were added and incubated at 37°C for 1 hour. For relative quantification, a standard serum sample was made by pooling samples on day 7 post secondary immunization from a prior experiment, which was used for all subsequent experiments. Serum samples of various days were serially diluted first to determine the proper dilution, and 1:1000-1:5000 was chosen for NP-IgG1 ELISA. After serum sample incubation, plates were washed 6 times with PBS-0.05% Tween 20 followed with detection antibody incubation at 37°C for 1 hour. Alkaline phosphatase- (AP-) conjugated goat-anti-mouse IgG1 (Southern Biotech) detection antibody was diluted in 1% milk-PBS at a ratio of 1:2000. After 5 washes with PBS-0.05% Tween 20 and once with PBS, AP-substrate (S0942, Sigma) dissolved in pNPP buffer was added for 15-20 min and absorbance was measured at 405 nm by ELISA plate reader (FLUOstar Omega, BMG Labtech). To detect IgM in culture supernatants, samples were diluted at the ratio of 1:100. The IgM Mouse Uncoated ELISA Kit (88-50470, Thermo Fisher) was used according to manufacturer’s protocol.

37MultiScreenHTS-HA filter plates (MSHAS4510) were first rinsed with 35% ethanol for 30 s and washed 3 times with PBS. Plates were coated with 20 μg/ml NP-BSA in PBS at 4°C over night. After three washes with PBS, plates were blocked with RPMI-1640 medium supplemented with 10% FBS at 37°C for 30 minutes. Then bone marrow cells were added in duplicates in culture medium at the density of 3x105 /100 μL/ well and cultured at 37°C over night. Plates were then washed 3 times with PBS and 3 times with PBS-0.05% Tween 20. AP-conjugated anti-mouse IgG1 detection antibody diluted in 1% FBS was added to plates for 1 hour incubation at 37°C. After 5 washes with PBS-0.05% Tween 20 and once with PBS, AP substrate (170-6432, Bio-Rad) was added for spot development. Plates with clear spots and clean background were washed with water to stop development, dried, and counted with the AID ELISpot Reader System (ELR078IFL, AID GmbH).

38The TFEB polyproline motif-mCherry-IRES, 13P-mCherry-IRES, Random-mCherry-IRES, murine wild-type/mutant (PPP to AAA) Tfeb-HA were cloned into pMSCV-UBC:GFP. A flexible GSGSG linker was inserted between the targeted sequence and mCherry to assist folding of the fusion protein. The sequence coding C-terminal 37-amino acid degron of murine ODC was inserted to the C-terminal of mCherry. The nuclear localization sequence RRRR (AGAAGACGCAGG) of TFEB was mutated to AAAA (GCTGCAGCCGCG) by PCR.

39NIH 3T3 cells were transfected with siRNA according to Thermo Fisher protocol (13778075, Lipofectamine® RNAiMAX Reagent). Cells were collected on day 3 post-transfection for western blot analysis. For plasmid transfection assay, NIH 3T3 cells were transfected with the indicated plasmids using Lipofectamine 3000 according to the manufacturer’s protocol. Protein expression was quantified by flow cytometry (geometric mean fluorescence intensity) or western blot 24 hours post transfection.

40Cells were stained with fixable Zombie Aqua Live/Dead staining (423102, Biolegend), FcR block, and surface marker antibodies and analyzed with four-laser LSR Fortessa X-20. Acquired data were analyzed using FlowJo 10.2. For LC3-II flow cytometry staining, the FlowCellect™ Autophagy LC3 Antibody-based Detection Kit (FCCH100171, Merck Millipore) and for CellTrace staining, CellTrace Violet (C34557, Thermo Fisher) were used according to the manufacturer’s protocol.

41Cells were stained with CytoID (ENZ-51031-K200, Enzo Life Sciences) according to manufacturer’s protocol. Briefly, cells were re-suspended in CytoID staining solution (1:4000 diluted in RPMI without phenol red (R1780, Sigma) supplemented with 5% FBS) and incubated at 37°C for 30 minutes in the dark. Then cells were washed once, followed by surface marker staining and flow cytometry analysis without fixation.

42RNA was extracted using the RNeasy Plus Mini Kit (74134, QIAGEN). The concentration of RNA was measured using NanoDrop1000 (Thermo Fisher), followed by reverse transcription using the High Capacity RNA-to-cDNA Kit (4387406, Thermo Fisher). Taqman probes (Thermo Fisher), TaqMan Gene Expression Master Mix (4369016, Thermo Fisher), and the ViiA 7 Real-Time PCR System (Thermo Fisher) were used for quantitative PCR. ΔΔCt method was used for the quantification of target mRNAs expression using Gapdh as the reference gene.

43B cells from GFP-LC3 mice were purified by magnetic-activated cell sorting (MACS) and fixed with 4% paraformaldehyde at room temperature for 10 min. After nuclear staining with DAPI (Sigma), cells were transferred to slides using Cytospin 3 cytocentrifuge (Shandon) and imaged with the Olympus FV1200 Laser Scanning Microscope. CellProfiler software was used for automatic autophagosome quantification. Nuclei were defined in DAPI channel (diameter of 45-100 pixel units), autophagosomes were defined in GFP channel (diameter of 3-15 pixel units,1 pixel unit = 100 nm). An artificial but consistent threshold of 1-10 cellular GFP spots in GFP-LC3 confocal imaging was used for statistical analysis. 10%–45% B cells are GFP-negative in GFP-LC3 transgenic mice as assessed by flow cytometry. Therefore cells with 0 GFP spot (either due to no GFP expression or very low autophagy) were excluded from all samples. Cells with more than 10 GFP spots were also excluded from all samples as outliers.

44Cells in a 1.5 mL Eppendorf tube were washed once with cold PBS and re-suspended in 200-500 μL HLB buffer (10 mM Tris-HCl pH 7.5, 10 mM NaCl, 2.5 mM MgCl2) with protease inhibitors and phosphatase inhibitors. Then an equal volume of HLB+2N buffer (HLB+0.4% NP-40, 1N is 0.2% NP-40) was added and carefully mixed. Samples were incubated on ice for 5 minutes and then underlayed with 200 μL HLB+NS buffer (HLB+0.2% NP-40+10% sucrose). Samples were centrifuged at 500 g for 5 minutes and the upper half of the supernatant was collected as the cytoplasmic fraction. The remaining liquid was discarded and the nuclear pellet was washed once more with HLB buffer and lysed with NP-40 lysis buffer or Laemmli sample buffer. All procedures were performed on ice.

45LPS-stimulated mouse B cells were treated with 10 μM GC7 on day 2 for 24 hours. Cells were collected and fractionated for label-free quantitative mass spectrometry (MS) analysis. For SILAC MS, B cells stained with CellTrace Violet (C34557, Thermo Fisher) were cultured in two types of medium: light and heavy. In light medium, SILAC RPMI-1640 medium (89984, Thermo Fisher) supplemented with 1.15 mM L-Arg (Sigma) and 0.22 mM L-Lys (Sigma), 10% dialyzed FBS (Sigma), and P/S, L-Glutamine, 50 μM 2-mercaptoethanol, 20 mM HEPES were used. In heavy medium, 13C615N4-L-Arg (Arg-10, Silantes) and 13C615N2-L-Lys (Lys-8, Silantes) were used instead of the light arginine and lysine respectively. 10 μM GC7 was added to either light or heavy medium in two repeats. After GC7 treatment, dead cells were first removed using the Dead Cell Removal Kit (Miltenyi). An identical number of cells that had divided at least 4 times in light or heavy medium were collected by FACS, and mixed for cell fractionation and protein MS analysis.

46Peptide samples were prepared using the Filter Aided Sample Preparation (FASP), as previously described. Briefly, Vivacon 500 filters (Sartorius, VN01H02 10 kDa/VNCT01) were pre-washed with 200 μL 0.1% trifluoroacetic acid in 50% acetone. Lysate samples were loaded to the filter and denatured with 200 μL 8 M urea in 100 mM triethylammonium bicarbonate buffer (TEAB) for 30 minutes at room temperature. Denatured proteins were reduced by 10 mM tris (2-carboxyethyl) phosphine (TCEP) for 30 minutes at room temperature and alkylated with 50 mM chloroacetamide for 30 minutes at room temperature in the dark. Subsequently, 1 μg LysC (Wako) in 150 μL 50 mM TEAB containing 6 M urea was added and incubated at 37°C for 4 hours. Then the buffer was diluted to 2 M urea by 50 mM TEAB, followed by adding 0.5 μg trypsin (Promega) overnight at 37°C. Trypsinised samples were centrifuged and the flow-through, containing peptides, was dried and resuspended in 70 μL 10% formic acid (or 5% formic acid and 5% DMSO for SILAC experiment).

47Peptides were separated on an Ultimate 3000 UHPLC system (Thermo Fisher) and electrosprayed directly into a QExactive mass spectrometer (Thermo Fisher). The peptides were trapped on a C18 PepMap100 pre-column (300 μm i.d. x 5 mm, 100 Å, Thermo Fisher) using solvent A (0.1% formic acid in water) at a pressure of 500 bar, then separated on an in-house packed analytical column (75 μm i.d. packed with ReproSil-Pur 120 C18-AQ, 1.9 μm, 120 Å, Dr. Maisch GmbH) using a linear gradient (length: 120 minutes, 15% to 35% solvent B (0.1% formic acid in acetonitrile), flow rate: 200 nl/min). Data were acquired in a data-dependent mode (DDA). Full scan MS spectra were acquired in the Orbitrap (scan range 350-1500 m/z, resolution 70000, AGC target 3e6, maximum injection time 50 ms). The 10 most intense peaks were selected for HCD fragmentation at 30% of normalized collision energy at resolution 17500, AGC target 5e4, maximum injection time 120 ms with first fixed mass at 180 m/z. Charge exclusion was selected for unassigned and 1+ ions. The dynamic exclusion was set to 20 s.

48Raw MS data were processed by MaxQuant (version 1.5.0.35i) for peak detection and quantification. MS spectra were searched against the Mus musculus UniProt Reference proteome (retrieved 12/01/17) alongside a list of common contaminants, using the Andromeda search engine with the following search parameters: full tryptic specificity, allowing two missed cleavage sites, fixed modification was set to carbamidomethyl (C) and the variable modification to acetylation (protein N terminus) and oxidation (M). The search results were filtered to a false discovery rate (FDR) of 0.01 for proteins, peptides and peptide-spectrum matches (PSM). Protein intensity distributions were log2 transformed and median-centered using Perseus (version 1.5.5.3). For the SILAC analysis, missing values were replaced by an estimated background noise value. Proteins without greater-than-background values in both replicates for at least one condition were removed. MA plots were generated using R (version 3.4.2). Reviewed autophagy proteins were searched against the Mus musculus UniProt database. Data from the independently analyzed cytoplasmic and nuclear proteins were shown as overlaid MA-plots wherein the log2 average (GC7 + Vehicle) intensity (A) is plotted on x versus the log2 (GC7:Vehicle) fold change (M) plotted on y.

49Cells were washed with PBS and the pellet resuspended in lysis buffer (80% methanol + 5% Trifluoroacetic acid) spiked with 2.5 μM 1,7-diaminoheptane (Sigma). The cell suspension, together with acid-washed glass beads (G8772, Sigma), was transferred to a bead beater tube and homogenized in a bead beater (Precellys 24, Bertin Technologies) for four cycles (6500 Hz, 45 s) with 1 minute of ice incubation between each cycle. The homogenized samples were centrifuged at 13,000 g for 20 minutes at 4°C. The supernatant was collected and dried overnight. For chemical derivatization, 200 μL trifluoroacetic anhydride was added to the dried pellet and incubated at 60°C for 1 hour, shaking at 1200 rpm. The derivatization product was dried, re-suspended in 30 μL isopropanol and transferred to glass loading-vials. The samples were analyzed using a GCxGC-MS system as described. The following parameters were used for quantification of the 1D-GC-qMS data: type: area, slope: 1000/min, width: 0.04 s, drift 0/min and T. DBL: 1000 min without any smoothing methods used. Cellular spermidine amount was normalized to total protein levels determined by BCA assay for each sample first followed by comparison between samples.

50Protein synthesis rate was measured using the Click-iT Plus OPP Protein Synthesis Assay (C10456, Thermo Fisher) according to manufacturer’s protocol. Geometric mean fluorescence intensity (normalized to vehicle) was used as an indicator of the relative translation rate.

51Click-iT® Metabolic Labeling kits (C10102, C10276, Thermo Fisher) were used for AHA labeling according to manufacturer’s protocol. Briefly, cells were pre-cultured in methionine-free culture media (R7513, Sigma) supplemented with 10% dialyzed FBS, P/S, L-Cystine, and L-Glutamine for 30 minutes to deplete methionine reserves. 50 μM AHA was then added for 4 hours. Cycloheximide (10 μg/mL, Sigma) was added 30 min before adding AHA to inhibit translation. Cells were lysed in lysis buffer (50 mM Tris-HCl, pH 8.0, 1% SDS with protease/phosphatase inhibitors). The cleared cell lysate (200 μg protein/sample) was used to perform Click reaction for biotin labeling. Biotinylated proteins were directly used for western blot to determine the overall protein synthesis rate or pulled down with streptavidin-agarose beads (50 μL beads/sample, S1638, Sigma) overnight to purify nascent proteins for western blot analyses.

52The sequencing library was prepared according to the reference with the following changes. LPS-activated B cells were MACS-purified with the Dead Cell Removal Kit (Miltenyi) to remove dead cells prior to cell lysis. 100 μg/ml cycloheximide was used in the lysis buffer. The B cell lysate was spiked with the S. cerevisiae lysate at the ratio of 13 million B cells: 270 million yeast cells (which equals to 20:4 μg miRNA). Clarified B cell and yeast lysates were RNase I-digested, ribosome purified and miRNA extracted to determine the miRNA yield for spiking. Size exclusion chromatography (27-5140-01, GE Healthcare) was used to recover ribosomes. The miRNeasy Micro Kit (QIAGEN) was used to purify miRNAs. Ribosome protected fragments (RPFs) of 27-30 nt size were purified by gel extraction. The Ribo-Zero Gold rRNA Removal Kits (illumina, MRZG126, MRZY1306) were used to deplete ribosomal RNAs, in which the Ribo-Zero Removal Solutions from the mouse and yeast kits were mixed at the ratio of 8 μL: 4 μL to remove both mouse and yeast rRNAs. The following barcoding primers were used for next generation sequencing. 1. 5′-AATGATACGGCGACCACCGAGATCTACACGATCGGAAGAGCACACGTCTGAACTCCAGTCACATGCCATCCGACGATCATTGATGG. 2. 5′-AATGATACGGCGACCACCGAGATCTACACGATCGGAAGAGCACACGTCTGAACTCCAGTCACTGCATCTCCGACGATCATTGATGG. The libraries were sequenced at the Biopolymers Facility, Harvard Medical School.

53For the purposes of alignment, the reference genome assemblies for Mus musculus (GRCM38) and Saccharomyces cerevisiae (SacCer3) were concatenated to create a combined genome fasta file. Prior to alignment, adaptor sequence (ATCTCGTATGCCGTCTTCTGCTTG) was removed from demultiplexed reads and low quality reads (Phred score < 20) were discarded. To remove contaminating rDNA reads, reads were aligned to the combined rDNA genome using bowtie2 and all unaligned reads were kept for alignment to the combined genome. Non-rDNA reads were aligned to the combined genome using bowtie2. All post-alignment analysis was performed using custom scripts written in R. P-site offsets for each read-length from 26 to 35 nts in the RPF libraries were determined by generating metagenes relative to the START codon and applied to aligned reads. For all metagenes, transcripts with no reads within the window of interest were excluded from analysis before applying any other filters. Additionally, for each position in the windows of the 5′ and 3′ metagenes, the highest and lowest 0.1% of signals were removed from analysis.

54Peripheral blood mononuclear cells (PBMCs) were isolated by density gradient centrifugation using Ficoll-Paque. PBMCs were counted using trypan blue. They were either freshly used or frozen until further use. Thawed cells were placed in full medium overnight to allow selection of viable cells. PBMCs were washed with PBS and lysed using NP-40 lysis buffer containing proteinase inhibitors and phosphatase inhibitors on ice for western blot assay or using lysis buffer (80% methanol + 5% Trifluoroacetic acid) spiked with 2.5 μM 1,7-diaminoheptane (Sigma) for spermidine measurement by GC-MS. B cells were sorted using a negative selection kit (B Cell Isolation Kit II, human, Miltenyi Biotec) according to the manufacturer’s protocol. B cells (5 × 105 cells) were seeded in 24-well plates, then activated with anti-IgM (5 μg/ml, Jackson Immuno Research) and CD40L (100 ng/ml, Enzo Life science) and treated with 1 mM DFMO, 10 μM spermidine, or 10 μM GC7 for 7 days. Cells then were lysed for western blot or stained for LC3-II measurement by flow cytometry. IgG release in culture supernatants was measured by heterologous two-site sandwich ELISA, according to the manufacturer’s protocol (Invitrogen).

55Prism software (GraphPad) was used for statistical analyses. Data are represented as mean ± SEM. All data points refer to the measurement of distinct samples (biological repeats). Unpaired two-tailed Student’s t test was used for comparisons between two normally distributed datasets with equal variances unless specified. One-tailed Student’s t test was used when the null hypothesis has a clear direction according to current knowledge (e.g., spermidine is an autophagy inducer in multiple systems or Bafilomycin A1 treatment leads to accumulation of LC3-II). Welch’s unequal variances t test was used when the two normally distributed datasets have unequal variances. Holm-Sidak method was used to adjust P values of a family of multiple t tests with a single null hypothesis. Paired or unpaired one/two-way ANOVA, followed with post hoc Sidak’s test (comparing two means within each group), Dunnett’s test (comparing multiple means with the single control within each group), or Tukey’s test (comparing the mean of each sample with the mean of every other sample), was used for multiple comparisons of normally distributed datasets with one/two variables. Mann-Whitney U test was used for comparisons between non-parametric datasets. Linear regression with 95% confidence interval was used to assess the relationships between age and the expression of target proteins or spermidine levels, in which R2 was used to assess the goodness of fit and the P value calculated from F test was used to assess if the slope was significantly non-zero. P value was used to quantify the statistical significance of the null hypothesis testing. ∗p ≤ 0.05, ∗∗p ≤ 0.01, ∗∗∗p ≤ 0.001, ∗∗∗∗p ≤ 0.0001.

56Original images are available at https://doi.org/10.17632/cd4f7j2hbh.1

57Sequencing results have been deposited to GEO with the accession number GSE133934.