Homeodomain-interacting protein kinase maintains neuronal homeostasis during normal Caenorhabditis elegans aging and systemically regulates longevity from serotonergic and GABAergic neurons

To precisely determine where HPK-1 functions in the regulation of longevity, we overexpressed hpk-1 within neuronal (rab-3p), muscle (myo-3p), intestine (ges-1p), or hypodermal (dpy-7p) tissues to identify where activity was sufficient to increase lifespan. Increased neuronal expression of hpk-1 was sufficient to increase lifespan by 17%, which is comparable to constitutive overexpression of hpk-1 throughout the soma (Figure 1A; Das et al., 2017). In contrast, transgenic animals overexpressing hpk-1 in either muscular, intestinal, or hypodermal cells had no significant change in lifespan (Figure 1B–D). These results were somewhat unexpected, as we previously found that loss of hpk-1 solely within either neuronal, intestinal, or hypodermal cells is sufficient to shorten lifespan (Das et al., 2017). hpk-1 is broadly expressed during embryogenesis and larval development, but HPK-1 protein is only expressed within the nervous system in adult animals (Berber et al., 2013; Berber et al., 2016; Das et al., 2017; Raich et al., 2003). Furthermore, hpk-1 and orthologs are known to have roles in development, differentiation, and cell fate specification (Berber et al., 2013; Blaquiere and Verheyen, 2017; Calzado et al., 2007; Hattangadi et al., 2010; Rinaldo et al., 2007; Rinaldo et al., 2008; Steinmetz et al., 2018). Therefore, we conclude that HPK-1 functions primarily within the nervous system in adult animals to regulate aging.

We performed a locomotion assay to test whether neuronal hpk-1 could rescue an age-associated decline in movement, a readout of C. elegans healthspan (Berber et al., 2016; Das et al., 2017; Herndon et al., 2002; Huang et al., 2004; Samuelson et al., 2007). As expected, hpk-1(pk1393) null mutants displayed a reduced frequency of body bends at day 2 of adulthood (Figure 1E). Increased expression of hpk-1 in neurons was sufficient to partially mitigate the locomotion defect of hpk-1 null mutants in response to active stimuli. Next, we tested whether restoring hpk-1 exclusively within the nervous system would prevent the premature age-associated decline in movement of hpk-1 null mutant animals. Indeed, restoring hpk-1 neuronal expression rescued the premature age-associated decline in movement of hpk-1 null mutant animals (Video 1). Thus, HPK-1 activity within the nervous system is sufficient to improve healthspan. Lastly, neuronal hpk-1 rescue was sufficient to restore a wild-type lifespan in hpk-1 null mutant animals (Figure 1F). It should be noted that this same transgene increases wild-type lifespan (Figure 1A), which suggests that HPK-1 retains essential functions for adult healthspan and lifespan outside of the nervous system; we posit these functions are related to development as HPK-1 is no longer expressed outside of the nervous system in adult animals (Das et al., 2017).

We next asked whether neuronal hpk-1 expression would be sufficient to rescue the ability of animals to survive acute stress. We and others previously found that in the absence of hpk-1, animals are vulnerable to thermal stress (Berber et al., 2016; Das et al., 2017). Restoring hpk-1 expression solely within the nervous system was sufficient to rescue the reduced thermotolerance of hpk-1 null mutant animals (Figure 1G). Collectively, these findings are consistent with the notion that hpk-1 primarily acts within the nervous system to regulate aging and stress resistance.

We sought to determine whether hpk-1 expression changes during normal aging. hpk-1 mRNA expression increases in wild-type animals during normal aging between day 2 and 10 of adulthood (Figure 2A). Adult C. elegans are composed of 959 somatic cells, of which 302 are neuronal; all somatic cells are post-mitotic and arise from an invariant lineage. A recent comprehensive single-cell analysis of age-associated changes in gene expression throughout C. elegans identified 211 unique ‘cell clusters’ within the aging soma: groups of cells within a lineage or tissue defined by gene expression profiles (Roux et al., 2022). The ‘C. elegans Aging Atlas’ analysis identified 124 neuronal cell clusters and obtained age-associated changes in gene expression for 122, with high concordance of overall gene expression between neuron clusters to a previous project, which mapped the complete gene expression profiles of the C. elegans nervous system in late larval animals (The CeNGEN project, Taylor et al., 2021). The C. elegans Aging Atlas provides the means to investigate age-associated changes in gene expression at an unprecedented level of resolution.

Using the C. elegans Aging Atlas dataset, in which differentially expressed genes were identified within each cell cluster in old versus young animals, we find that hpk-1 expression increases during aging within 48 cell clusters: broadly within the nervous system and in 11 non-neuronal cell clusters (Figure 2—figure supplement 1, see Materials and methods). With aging, neuronal expression of hpk-1 increased within interneurons, motor neurons, sensory neurons, and all neuronal subtypes (cholinergic, dopaminergic, GABAergic, serotonergic, and glutamatergic). We found no cell clusters in which hpk-1 is downregulated during normal aging. It is well known that many longevity-associated TFs increase lifespan when overexpressed and shorten lifespan when lost (reviewed in Denzel et al., 2019), which infers these TFs might be upregulated, or become chronically activated, during normal aging. A major finding of the C. elegans Aging Atlas study was that many longevity-associated TFs are upregulated during normal aging in more than 200 cell clusters (Roux et al., 2022). Increased expression of longevity-associated TFs is thought to reflect an age-associated adaptive response to dysregulation of homeostatic mechanisms and accumulating damage (Roux et al., 2022). We posit hpk-1 neuronal expression increases during normal aging in wild-type animals to mitigate accumulating damage and cellular dysfunction.

We next assessed whether any TFs showed patterns of spatial upregulation similar to the age-associated increase in expression of hpk-1, and then segregated those cell types into neuronal and non-neuronal subgroups. Seventy-one TFs were differentially upregulated with hpk-1 during aging within 3 or more neuron clusters, 30 within non-neuronal clusters, with an overlap of 21 TFs between neurons and non-neurons (Figure 2—figure supplement 2 and Supplementary file 5). Overall, 26 of these TFs have previously been implicated in longevity. There are 10 TFs broadly upregulated with aging in the nervous system that overlap with hpk-1 (significant differential expression in 9 or more neuronal clusters), 6 of which have previously been linked to longevity: daf-16 (FOXO, overlaps in 19/37 neuronal clusters), hlh-30 (TFEB, 18/37), dpy-22 (mediator complex subunit 12L, 15/37), gei-3 (10/37), skn-1 (NRF2, 10/37), and hif-1 (hypoxia inducible factor 1, 10/37) (Figure 2C; An and Blackwell, 2003; Chen et al., 2009; Fawcett et al., 2015; Kenyon et al., 1993; Lapierre et al., 2013; Lin et al., 1997; Ogg et al., 1997; Roux et al., 2022; Suriyalaksh et al., 2022; Tacutu et al., 2012; Zhang et al., 2009). Conversely, 12 TFs exhibited an inverse correspondence, and were largely downregulated with aging in the same neurons where hpk-1 is normally upregulated (significant differential expression in nine or more neuronal clusters) (Figure 2D); only two of which have previously been implicated in longevity: xbp-1 (X-box binding protein 1) a key regulator of the endoplasmic reticulum unfolded protein response (ER-UPR), and mdl-1 (MAX dimerization protein) (Johnson et al., 2014; Nakamura et al., 2016; Riesen et al., 2014). Of note, mdl-1 is a Myc-family member with opposing roles in longevity from mxl-2 (Mlx) (Johnson et al., 2014), and mxl-2 is required for hpk-1 to extend longevity (Das et al., 2017).

We tested whether we could verify changes in age-associated gene expression for several candidate longevity-associated TFs identified in the Aging Atlas study. Age-synchronized wild-type animals at either day 1 or 11 of adulthood were harvested and mRNA levels of hpk-1, xbp-1, and gei-3 were assessed. As expected, both hpk-1 and gei-1 were significantly induced in older animals, and conversely xbp-1 expression was decreased (Figure 2E). We conclude that we can recapitulate results obtained from the Aging Atlas study and confirm that longevity-associated TFs are significantly differentially expressed during normal aging. Next we determined whether hpk-1 expression is significantly induced within the C. elegans nervous system during normal aging; we assessed age-associated changes in neuronal hpk-1 mRNA levels in vivo (hpk-1p::mCherry). As expected, we find that hpk-1 is significantly upregulated within isolated C. elegans neurons during normal aging (Figure 2F).

In non-neuronal clusters we found a much smaller subset of TFs with corresponding positive age-associated fold-changes. Only six of these TFs have been implicated in longevity: fkh-9 (FOXG), sea-2, zfh-2 (ZFHX3) (Walter et al., 2011), xbp-1, mdl-1, and hlh-6 (achaete-scute family bHLH TF 3) (Figure 2—figure supplement 1).

Within neuronal cell clusters, hpk-1 and the aforementioned longevity-associated TFs are broadly expressed throughout the nervous system by late larval development, with the exception of the nematode-specific gene gei-3, for which overlap is restricted to a subset of neurons (Figure 2—figure supplement 3). TFs with age-associated increases in expression in clusters overlapping with hpk-1 are associated with positive regulation of adult lifespan, heat and oxidative stress response, bacterial innate immunity, chromatin remodeling, development, mitophagy, axon regeneration, neddylation, and ubiquitination (Figure 2C). TFs with decreased expression in aging in hpk-1-upregulated neuronal cell clusters have been implicated in the ER unfolded protein response and the innate immune response (Figure 2D). Interestingly, when assessed at the level of individual time points rather than ‘old’ and ‘young’ groups, we find that the age-associated upregulation of hpk-1 and expression changes of longevity-associated TFs does not necessarily occur in a tightly linked manner suggestive of a single upstream regulatory process (see Appendix 1). Rather, we posit that hpk-1 is induced within the nervous system along with key longevity TFs during normal aging in response to accumulating stressors and collapsing homeostatic mechanisms to mitigate the physiological effects of aging.

We were surprised to find that hpk-1 mRNA expression increased during aging, as kinase regulation at the level of gene expression seemed atypical. Therefore, using the C. elegans Aging Atlas dataset we assessed whether any of the 438 C. elegans kinases (Zaru et al., 2017) are differentially expressed during aging: 115 kinases are upregulated and conversely 120 are downregulated in at least one cell cluster, while 254 kinases do not change expression (Figure 2—figure supplement 4, Supplementary file 6). hpk-1 is upregulated in more cell clusters than any other kinase, while hpk-1 is upregulated in 48, the next most broadly upregulated kinase is gcy-28 (natriuretic peptide receptor 1, 38 cell clusters) (Li and van der Kooy, 2018; Shinkai et al., 2011; Tsunozaki et al., 2008). Only 14 kinases are downregulated in 10 or more cell clusters. Conversely, 20 kinases are upregulated in 10 or more cell clusters (Figure 2—figure supplement 4). In addition to hpk-1 and gcy-28, five kinases previously linked to longevity are upregulated: nipi-3 (a Tribbles pseudokinase that regulates SKN-1; Wu et al., 2021), aak-1 (AMP-activated kinase catalytic subunit alpha-2) (Lee et al., 2009), scd-2 (ALK receptor tyrosine kinase) (Shen et al., 2007), rsks-1 (ribosomal protein S6 kinase) (Hansen et al., 2007; Pan et al., 2007), and smg-1 (SMG1) (Masse et al., 2008).

We previously demonstrated that HPK-1 is an essential component of TORC1-mediated longevity and autophagy (Das et al., 2017). Interestingly, we find that only 2 of 11 TOR-associated genes, rsks-1 and aak-1, are upregulated during aging in three or more cell clusters (Blackwell et al., 2019; Hindupur et al., 2015; Pan et al., 2007). We find rsks-1 is exclusively upregulated outside of the nervous system and has poor overlap with hpk-1 (2/13). aak-1 is primarily but not exclusively upregulated outside of the nervous system during aging, with some overlap with hpk-1. In future studies it will be interesting to dissect whether HPK-1 intersects with these components of TORC1 signaling within and/or outside of the nervous system.

Based on the neuronal HPK-1 capacity to promote longevity, we sought to determine whether HPK-1 preserves the anatomy and physiology of the nervous system during aging. To assess age-associated changes in neuroanatomy of ventral and dorsal cord motor neurons (VD and DD), we used the unc-25p::GFP cell-specific reporter (Figure 3A; Cinar et al., 2005; Kraemer et al., 2003). Aged populations of hpk-1 mutant animals displayed more than a threefold difference in VD axonal breaks and more than a twofold difference in DD axonal breaks compared to wild-type animals (Figure 3A–B and D). These age-associated breaks could be rescued by neuronal expression of hpk-1 (Figure 3C). Importantly, young wild-type and hpk-1 mutant animals did not show neuronal breaks (Figure 3B). Thus, loss of hpk-1 does not disrupt axon formation during development, rather HPK-1 preserves axonal integrity in aging motor neurons.

In C. elegans, locomotion is controlled by acetylcholine-releasing excitatory motor neurons and GABA-releasing inhibitory motor neurons (Jorgensen and Nonet, 1995; Richmond and Jorgensen, 1999). Thus, locomotion is a useful readout to assess neurotransmission. To examine the role of hpk-1 in synaptic transmission, we performed a paralysis assay after acetylcholinesterase inhibition with aldicarb, where mutations that impair cholinergic signaling cause a reduction in rate of paralysis (Mahoney et al., 2006; Oh and Kim, 2017). Therefore, mutants with lower synaptic transmission undergo reduced paralysis (i.e., animals are resistant), while mutants with higher synaptic transmission exhibit greater sensitivity. Interestingly, hpk-1 mutants showed increased aldicarb resistance by L4/YA (Figure 3E), suggesting that loss of hpk-1 reduces cholinergic synaptic transmission and that HPK-1 normally preserves neuronal function.

C. elegans serotonin (5-HT) signaling in conjunction with dopamine (DA) modulates locomotion; animals treated with exogenous 5-HT exhibit paralysis over time (Hardaker et al., 2001; Omura et al., 2012). In a paralysis assay with exogenous 5-HT, we observed that late L4/YA hpk-1 mutant animals paralyzed sooner: at 2 min of 5-HT treatment, 63% of mutants were immobilized compared to 25% of wild-type animals immobilized (Figure 3F). Based on the aforementioned findings, we conclude that HPK-1 prevents an age-associated decline in axonal and synaptic integrity.

We sought to gain deeper insight into how loss of hpk-1 negatively impacted the C. elegans nervous system. To identify early transcriptional dysregulation events that occur due to the absence of hpk-1, we conducted bulk RNA-sequencing (RNA-Seq) of day 2 adult wild-type and hpk-1(pk1393) null mutant animals, an age at which axonal breaks have yet to occur (Figure 3B and D), but after neuronal function has begun to decline (Figure 3E and F). Differential expression analysis identified 2201 total genes with significant expression changes with loss of hpk-1, 1853 upregulated and 348 downregulated (adjusted p-value <0.05 and log₂ fold-change magnitude ≥1) (Figure 4A and Supplementary file 10). Functional enrichment with over-representation analysis was performed for the up- and downregulated genes separately (GOSeq, adjusted p-value <0.01). Among downregulated genes, we find 25 genes implicated in the innate immune response, as well as TGF-beta and WNT signaling (five ubiquitin ligase complex component genes), glycosphingolipid metabolism, and surfactant metabolism (Figure 4B). By contrast, more than 50 Gene Ontology (GO) terms and pathways were enriched for the upregulated genes. The most significantly enriched single term reveals a substantial number of cuticle and collagen-associated genes as dysregulated (Figure 4C). Upregulation of some collagens and ECM-remodeling processes have been associated with longevity in C. elegans (Ewald et al., 2015; Rahimi et al., 2022). Alternatively, aberrant upregulation could lead to stiffening of the cuticle, a hallmark of aged C. elegans and some progeric mutants (Ewald et al., 2015; Rahimi et al., 2022).

A striking number of other functions significantly enriched in the hpk-1 upregulated genes from across GO, KEGG, and Reactome are related to neuronal function and signaling, including neuropeptide and neurotransmitter pathways, mechanosensation, chemotaxis, and axon guidance, among others (Figure 4C). To explore further, we looked at the hpk-1 null upregulated genes in the context of a curated compendium of genes with functions that are important for the nervous system (Hobert, 2013), as well as genes identified as uniquely expressed in neurons or significantly enriched in neurons from tissue-specific bulk RNA-Seq (Kaletsky et al., 2018).

Among 1496 genes known, or expected through homology, to function in the nervous system, 264 are upregulated with loss of hpk-1, representing a much larger overlap than expected by chance (hypergeometric test p-value <0.0001) (Supplementary file 11). Genes with higher fold-change of expression in hpk-1 mutants were associated with ciliated sensory neurons (motor proteins, sensory cilia transport), calcium binding, and neuropeptides. We find that 55 of these neuron-associated genes are normally upregulated with age in wild-type animals in at least one cell cluster in the C. elegans Aging Atlas, 87 are downregulated with age, with the remaining having mixed or no significant age-associated changes. This suggests that while loss of hpk-1 results in expression of some genes associated with aging, the bulk of differentially expressed genes at day 2 of adulthood are consistent with dysregulation of neuronal gene expression. Refined genetic perturbation within the nervous system coupled with single-cell RNA-Seq over time would be required to fully reveal the impact of HPK-1 function on age-associated changes in gene expression.

We next investigated how many of these genes are predominantly or exclusively expressed in neurons, based on a set of 880 neuron-enriched and neuron-specific genes (Kaletsky et al., 2018). We found that 15% of all hpk-1 upregulated genes are enriched or specific to neurons in wild-type animals (283 genes, hypergeometric test p-value <0.0001) whereas only five hpk-1 downregulated genes normally exhibit such neuron-restricted expression. To determine what functions are over-represented among the upregulated neuronally expressed genes, we ran GO enrichment analysis. Representing these results as a network, and clustering terms which have overlapping gene associations, we identified some major themes: neurotransmitter and ion transport, synaptic components and signaling, axon and neuron projection components, feeding and locomotion behavior, mechanosensation and touch-responsive behavior are all upregulated in the absence of hpk-1 (Figure 4D). Among all genes upregulated in the hpk-1 null mutants, the largest group correspond to genes related to neuropeptide encoding, processing, and receptors. Neuropeptides are short sequences of amino acids important for neuromodulation of animal behavior, broadly expressed in sensory, motor, and interneurons. Of the 120 neuropeptide-encoding genes in the C. elegans genome, 40 (33%) are upregulated in the hpk-1 null mutants. Nine of these are insulin-related (ins), which suggest increased hpk-1 expression during normal aging or hpk-1 overexpression may positively affect longevity through decreased ins neuropeptide expression. We surmise that many of the neuronal genes that are induced in the absence of hpk-1 might be compensatory responses to maintain neuronal function. We conclude that loss of hpk-1 upregulates expression of a number of neuronal genes including those involved in neuropeptide signaling, neurotransmission, release of synaptic vesicles, calcium homeostasis, and function of sensory neurons (Appendix 2).

We hypothesized that HPK-1 maintains neuronal function by fortifying the neural proteome. We measured polyglutamine toxic aggregate formation in neuronal cells (rgef-1p::Q40::YFP) of animals overexpressing hpk-1 pan-neuronally (rab-3p). These animals showed a decreased number of fluorescent foci within the nerve ring (Figure 5Ai–iii, B). To assess whether this was associated with improved neuronal function, we assessed alterations in locomotion capacity (body bends) after placement in liquid. Increased neuronal expression of hpk-1 significantly increased the average number of body bends (17.7±1 vs 12.5±1.2 bends/30 s), suggesting that neuronal expression of hpk-1 improves proteostasis in the nervous system and mitigates locomotion defects associated with the decline in proteostasis (Figure 5C).

We next tested whether HPK-1 was necessary to preserve proteostasis within neurons. We crossed the rgef-1::Q40::YFP reporter with neuronal enhanced RNAi background (sid-1(pk3321);unc-119p::sid-1) and assessed whether neuronal inactivation of hpk-1 hastened age-associated proteostatic decline. Animals with neuronal hpk-1 inactivation had a significant increase in foci number within neurons and a decreased number of body bends in liquid (19.7±0.8 vs 25.5±1.1 bends/30 s) (Figure 5Aiv, D, E). Thus, HPK-1 functions cell intrinsically to maintain neuronal proteostasis.

We next assessed whether neuronal HPK-1 activates a cell non-autonomous signal to regulate proteostasis in distal tissues. We used a proteostasis reporter expressed exclusively in muscle tissue (unc-54p::Q35::YFP); animals display an age-associated accumulation of fluorescent foci, which are proteotoxic aggregates, and become paralyzed as muscle proteome function collapses (Lazaro-Pena et al., 2021; Morley et al., 2002). Animals with increased neuronal hpk-1 expression had significantly decreased fluorescent foci in muscle tissue (Figure 5Fi–ii). For instance, at day 3 of adulthood, animals overexpressing neuronal HPK-1 had an average of 21±0.9 polyQ aggregates, while control animals had an average of 26±1.1 polyQ aggregates (Figure 5G). In C. elegans, the aging-related progressive proteotoxic stress of muscle polyQ expression is pathological and results in paralysis. Of note, increased hpk-1 expression in neurons was sufficient to reduce paralysis; at day 13 of adulthood only 61%±0.10 of neuronal hpk-1 overexpressing animals were paralyzed compared to 81%±0.08 of paralyzed control animals (Figure 5H). Thus, increased neuronal HPK-1 activity is sufficient to improve proteostasis cell non-autonomously.

We confirmed that the kinase activity is essential for HPK-1 to improve proteostasis. We find the kinase domain of HPK-1 is highly conserved from yeast to mammals, both at the amino acid level and predicted tertiary structure, based on multiple sequence alignments, AlphaFold structure prediction, and visualization with PyMol (Figure 5—figure supplement 1; DeLano, 2002; Jumper et al., 2021; Varadi et al., 2022). We mutated two highly conserved residues within the kinase domain, K176A and D272N, which are known to be essential for the kinase activity of HPK-1 orthologs (He et al., 2010), and tested whether these kinase-dead mutant versions could improve proteostasis when neuronally overexpressed. In contrast to wild-type HPK-1 (Figure 5G and H), neither mutant had any effect on either polyglutamine foci accumulation (Figure 5—figure supplement 1C, D) or paralysis (Figure 5—figure supplement 1E, F) in muscle cells (unc-54p::Q35::YFP); thus, increased hpk-1 expression improves proteostasis through kinase activity.

We tested whether neuronal HPK-1 is required for maintaining proteostasis in distal muscle cells. Neuronal inactivation of hpk-1 (sid-1(pk3321); unc-119p::sid-1; unc-54p::Q35::YFP; hpk-1(RNAi)) was sufficient to hasten the collapse of proteostasis, significantly increasing the number of polyQ aggregates (45±1 vs 32±1.1 at day 4) and paralysis rate (84%±0.04 vs 14%±0.04 at day 9) (Figure 5Fiii, I, J). Thus, HPK-1 function within the nervous system impacts proteostasis in distal muscle tissue, which implies that HPK-1 coordinates organismal health through cell non-autonomous mechanisms.

To identify the signaling mechanism through which HPK-1 regulates distal proteostasis, we tested whether either neuropeptide or neurotransmitter transmission are essential for cell non-autonomous regulation of proteostasis (using unc-31 and unc-13 mutant animals, respectively; Figure 5K). Neuronal expression of hpk-1 in unc-31 mutants improved proteostasis in the absence of neuropeptide transmission and dense core vesicles, as increased neuronal hpk-1 expression still delayed the age-associated accumulation of aggregates within muscle cells. In contrast, in the absence of neurotransmitter function, neuronal expression of hpk-1 in unc-13 mutants failed to improve proteostasis in muscle tissue (Figure 5—figure supplement 2). We conclude that HPK-1-mediated longevity-promoting cell non-autonomous signaling occurs through neurotransmission.

We sought to identify the components of the PN that are cell non-autonomously regulated via HPK-1 activity in the nervous system. We tested whether increasing levels of neuronal hpk-1 alters the expression of molecular chaperones (hsp-16.2p::GFP). Increased neuronal expression of hpk-1 in wild-type animals did not alter the basal expression of molecular chaperones (Figure 6A). In contrast, after heat shock, increased neuronal hpk-1 expression resulted in hyper-induction of hsp-16.2p::GFP expression, particularly in the pharynx and head muscles, but surprisingly not within intestinal cells (Figure 6B–E). We conclude that while hpk-1 is required for the broad induction of molecular chaperones (Das et al., 2017), neuronal HPK-1 primes inducibility of the HSR within specific tissues, rather than in a systemic manner throughout the organism.

We next tested whether increased neuronal HPK-1 activity is sufficient to induce autophagy. Using two LGG-1/Atg8 fluorescent reporter strains to visualize autophagosome levels in neurons (rgef-1p::GFP::LGG-1), hypodermal seam cells, and intestine (lgg-1p::GFP::LGG-1) (Kumsta et al., 2017), we found that increased neuronal hpk-1 expression significantly induces autophagy in neurons and hypodermal seam cell (Figure 6F–I), but not in the intestine (Figure 6J and K). Sustained autophagy and transcriptional activation of autophagy genes are important components for somatic maintenance and organismal longevity (Lapierre et al., 2015). Consistently, hpk-1 is required for the induction of autophagy gene expression induced by TORC1 inactivation (Das et al., 2017). We tested whether neuronal overexpression of hpk-1 is sufficient to induce autophagy gene expression, and found that increased neuronal overexpression induces expression of atg-18 (WD repeat domain, phosphoinositide interacting 2) and bec-1 (Beclin 1) (Figure 6L). Thus, increased HPK-1 activity within the nervous system is sufficient to induce basal levels of autophagy both cell intrinsically and non-autonomously, which we posit occurs through increased autophagy gene expression to extend longevity.

hpk-1 is broadly expressed within the C. elegans adult nervous system (Das et al., 2017), but the neuronal cell types responsible for extending longevity are unknown. We overexpressed hpk-1 in different subtypes of neurons and assessed changes in muscle proteostasis (unc-54p::Q35::YFP). Increased hpk-1 expression in glutamatergic or dopaminergic neurons was not sufficient to decrease the formation of fluorescent foci or reduce paralysis (Figure 7—figure supplement 1). While it is possible that hpk-1 overexpression in these neuronal subtypes failed to reach a threshold effect, these results suggest that HPK-1 does not act within these neuronal subtypes to cell non-autonomously regulate proteostasis. In contrast, increased expression of hpk-1 in either serotonergic (tph-1p::hpk-1) or GABAergic (unc-47p::hpk-1) neurons was sufficient to improve proteostasis in muscle tissue (Figure 7A–D). Accordingly, we found that hpk-1 is expressed in serotonergic and GABAergic neurons, based on colocalization between hpk-1 and neuronal cell-type specific reporters (Figure 7E and F, Figure 7—figure supplement 2), consistent with the notion that HPK-1 initiates pro-longevity signals from these neurons.

We sought to determine whether signals initiated from serotonergic and GABAergic neurons induced similar or distinct components of the PN in distal tissues, and assessed the overall consequence on organismal health. Since pan-neuronal expression of hpk-1 was sufficient to induce autophagy, improve thermotolerance, and increase lifespan, we chose to dissect whether these three phenotypes were regulated within distinct neuronal cell types. Increased HPK-1 serotonergic signaling was sufficient to improve thermotolerance, without altering lifespan or autophagy (Figure 8A–D). In contrast, increased hpk-1 expression in GABAergic neurons did not alter thermotolerance, but induced autophagy slightly but significantly increased lifespan (Figure 8E–H).

To identify the mechanism through which increased neuronal HPK-1 activity induces autophagy in distal tissues, we examined serotonergic neurons. As expected, loss of tph-1, which encodes the C. elegans tryptophan hydroxylase 1 essential for biosynthesis of serotonin (Sze et al., 2000), failed to block the induction of autophagy by increased neuronal hpk-1 expression (Figure 8I, tph-1(mg280);rab-3p::hpk-1). unc-30 encodes the ortholog of human paired like homeodomain 2 (PITX2) and is essential for the proper differentiation of type-D inhibitory GABAergic motor neurons (Jin et al., 1994; Mclntire et al., 1993). Consistent with our previous result demonstrating that only GABAergic expression of hpk-1 promotes autophagy, loss of unc-30 was sufficient to completely abrogate the induction of autophagy in seam cells after increased pan-neuronal expression of hpk-1 (Figure 8I, unc-30(ok613);rab-3p::hpk-1). We conclude that HPK-1 activity in serotonergic neurons promotes systemic thermotolerance, whereas HPK-1 in GABAergic neurons promotes autophagy. We surmise that the two neuronal subtypes are likely to contribute synergistically to the extension of longevity.

HPK-1 regulates autophagy gene expression and across diverse species HIPK family members directly regulate TFs (Sardina et al., 2023), therefore we sought to identify TFs essential for HPK-1 overexpression within GABAergic neurons to induce autophagy. pha-4 (FOXA), mxl-2 (MLX), hlh-30 (TFEB), daf-16 (FOXO), and hsf-1 are all required in specific contexts for the induction of autophagy and longevity (Hansen et al., 2008; Kumsta et al., 2017; Lapierre et al., 2013; Lapierre et al., 2015; Nakamura et al., 2016; O’Rourke and Ruvkun, 2013). Furthermore, we previously found that hsf-1, mxl-2, or pha-4 are all necessary for the increased lifespan conferred by hpk-1 overexpression throughout the soma (Das et al., 2017). Using feeding-based RNAi, we found that hpk-1 overexpression in GABAergic neurons still significantly induced autophagosomes in seam cells after loss of hsf-1 or pha-4 (Figure 8J); thus HPK-1 induction of autophagy is independent of hsf-1 and pha-4 and suggests the pha-4 requirement for HPK-1-mediated longevity is independent of autophagy gene regulation (at least in hypodermal seam cells). In contrast, after inactivation of mxl-2, hlh-30, or daf-16, GABAergic overexpression of hpk-1 failed to significantly increase autophagosomes (Figure 8J); thus HPK-1-mediated induction of autophagy requires mxl-2, hlh-30, and daf-16. We were puzzled to find that the average number of autophagosome in control animals (i.e., lacking hpk-1 overexpression) varied after gene inactivation (Figure 8—figure supplement 1A). To rigorously test whether hpk-1 overexpression required hlh-30 for the induction of autophagosomes, we crossed the hlh-30 null mutation into animals overexpressing hpk-1 within GABAergic neurons (unc-47p::hpk-1;hlh-30(tm1978);lgg-1p::GFP::LGG-1); we find that hlh-30 is required for the induction of autophagosomes in seam cells when hpk-1 is overexpressed in GABAergic neurons (Figure 8—figure supplement 1B).

We sought to fortify our previous findings that: (1) TORC1 negatively regulates hpk-1 under basal conditions; (2) hpk-1 is essential for decreased TORC1 signaling to increase longevity and induce autophagy; and (3) HPK-1-mediated longevity requires mxl-2, the ortholog of mammalian Mlx TF (Das et al., 2017). From these data, we predicted that TOR-mediated longevity might require mxl-2 via hpk-1. Indeed, we find that inactivation of either daf-15 (Raptor) or let-363 (TOR) increased lifespan and required mxl-2, as mxl-2(tm1516) null mutant animals failed to increase lifespan in the absence of TORC1 (Figure 8—figure supplement 2). MXL-2 functions as a heterodimer with MML-1 (Pickett et al., 2007), which have similar roles in C. elegans longevity (Johnson et al., 2014). Interestingly, one of the human homologs of MML-1, carbohydrate response element binding protein (ChREBP/MondoB/MLXIPL/WBSCR14), maps within a deleted region in Williams-Beuren syndrome patients, which among other symptoms causes metabolic dysfunction, silent diabetes, and premature aging (Pober, 2010). This further supports our model and is consistent with other studies linking the Myc-family of TFs to longevity, autophagy, and TORC1 signaling (Johnson et al., 2014; Nakamura et al., 2016; Samuelson et al., 2007). Collectively, we conclude that HPK-1 regulates serotonergic signals to improve survival in acute thermal stress, likely through HSF-1, and initiates GABAergic signals to extend longevity through increased autophagy, which is linked to decreased TORC1 signaling and regulation of the Myc TFs.

All strains were maintained at 20°C on standard NGM plates with OP50. For all experiments, animals were grown in 20× concentrated HT115 bacteria seeded on 6 cm RNAi plates. Details on the strains, mutant alleles, and transgenic animals used in this study are listed in the Key resources table. All strains generated in this study are available upon request from the Samuelson laboratory or can be found at the Caenorhabditis Genetics Center (https://cgc.umn.edu/↗).

To assemble tissue-specific constructs for increased hpk-1 expression, the hpk-1 cDNA was amplified and cloned under control of the following promoters: hypodermal dpy-7p (Gilleard et al., 1997), body wall muscle myo-3p (Okkema et al., 1993), pan-neuronal rab-3p (Nonet et al., 1997), dopaminergic neurons cat-2p (Lints and Emmons, 1999), glutamatergic neurons eat-4p (Lee et al., 1999), serotonergic neurons tph-1p (Sze et al., 2000), and GABAergic neurons unc-47p (Gendrel et al., 2016). In brief, promoter sequences were subcloned from existing plasmids into the Prab-3p::hpk-1 plasmid. To create kinase domain point mutations K176A and D272N, the Prab-3p::hpk-1 plasmid was used and mutations were performed with a QuikChange II XL Site-Directed Mutagenesis Kit (Agilent).

All the assembled plasmids were validated by sequencing prior to microinjection. These constructs were injected at 5 ng/ml together with myo-2p::mCherry at 5 ng/ml as co-injection marker and pBlueScript at 90 ng/ml as DNA carrier.

The hpk-1 RNAi clone was originated from early copies of Escherichia coli glycerol stocks of the comprehensive RNAi libraries generated in the Ahringer and Vidal laboratories. The control empty vector (L4440) and hpk-1 RNAi colonies were grown overnight in Luria broth with 50 μg/ml ampicillin and then seeded onto 6 cm RNAi agar plates containing 5 mM isopropylthiogalactoside (IPTG) to induce dsRNA expression overnight at room temperature (RT). RNAi clones used in this study were verified by DNA sequencing and subsequent BLAST analysis to confirm their identity.

For measurement of gene induction, animals were synchronized and grown at 20°C, then isolated at the specified age. For time course of hpk-1 mRNA upregulation, animals were harvested at specified time points (D2, D4, D6, D8, and D10). For hpk-1 mRNA regulation at D11 of adulthood, animals were synchronized, grown at 20°C, and harvested at D1 and D11 of adulthood. For autophagy genes upregulation, animals were synchronized, grown at 20°C, and harvested at D1 of adulthood. After harvesting animals, RNA extraction was followed by using TRIzol reagent (Life Technologies) followed by RNeasy Plus Mini Kit (QIAGEN). RNA concentration was measured using a Nanodrop, and RNA preparations were reverse transcribed into cDNA using the Bio-Rad cDNA synthesis kit (#1708890) as per the manufacturer’s protocol. Quantitative real-time PCR was performed using PerfeCTa SYBR green FastMix (Quantabio) with three technical replicates for each condition. Primer sets with at least one primer spanning the exon were used to amplify the gene of interest. cdc-42 mRNA levels were used for normalization. Fold-change in mRNA levels was determined using ΔΔ Ct method (Livak and Schmittgen, 2001). Primer sequences can be found in the Key resources table.

Traditional lifespan assays were performed essentially as described in Das et al., 2017; Lee et al., 2003. Briefly, animals were synchronized by egg prep bleaching followed by hatching in M9 solution at 20°C overnight. The L1 animals were seeded onto 6 cm plates with HT115 bacteria and allowed to develop at 20°C. At L4 stage, FUdR was added to a final concentration of 400 µM. Viability was scored every day or every other day as indicated in each figure. Prism 7 was used for statistical analyses and the p-values were calculated using log-rank (Mantel-Cox) method.

For lifespan with RNAi-mediated knockdown of TOR (Figure 8—figure supplement 2), animals for the let-363 (TOR RNAi) condition were first raised on empty vector RNAi during development, and transferred to plates with let-363 RNAi bacteria at L4 to avoid developmental arrest at L3. In addition to having IPTG present in the media of the RNAi plates, an additional 200 µl of 0.2 M IPTG was added directly to the bacterial lawn of all plates and allowed to dry immediately before adding animals. The assays were otherwise performed as just described.

For lifespan assays in Figure 1F, assay plates and animal populations were prepared as just described, and then lifespan was determined from analysis of time-series image data collected for a given animal at approximately hour intervals throughout adult life on modified Epson V800 flatbed scanners with our instance of the C. elegans ‘lifespan machine’ (Oswal et al., 2021; Stroustrup et al., 2013).

In all cases, scoring of viability was blinded with respect to genotype and values obtained at previous time points, and animals that died due to rupturing, desiccation on the side of the plate or well, or clear body morphology defects were censored.

Survival assays at high temperature were conducted as previously described in Johnson et al., 2014. In brief, synchronized L1 animals were allowed to develop at 20°C and FUdR was added at the L4 stage. At day 1 adulthood, animals were moved to 35°C for a period of 6 and 8 hr. Animals were allowed to recover for 2 hr at 20°C, and viability was scored. In all cases, scoring of viability was blinded with respect to genotype and values obtained at previous time points; animals that died due to rupturing, desiccation on the side of the plate or well, or had clear body morphology defects were censored. Statistical testing between pairs of conditions for differences in the number of foci was performed using Student’s t-test analysis.

The hsp-16.2p::GFP animals were heat shocked on day 1 of adulthood for 1 hr at 35°C and imaged after 4 hr of recovery. Images were acquired using a Zeiss Axio Imager M2m microscope with AxioVision v4.8.2.0 software. The GFP a.u. values from the head and intestine of the animals were acquired using the area selection feature from AxioVision software. Two independent trials with a minimum of 20 animals per experiment were performed.

GFP::LGG-1/Atg8 foci formation was visualized as described in Kumsta et al., 2017. Briefly, L4 stage and day 1 adult animals were raised on HT115 bacteria at 20°C and imaged using a Zeiss Axio Imager M2m microscope with AxioVision v4.8.2.0 software at ×63 magnification. Two independent trials where at least 20 seam cells from 15 to 20 different animals were scored for GFP::LGG-1 puncta accumulation. In all cases scoring of puncta was blinded with respect to genotype.

Animals carrying the unc-25p::GFP reporter strain to visualize VD and DD motor neurons were synchronized and allowed to develop at 20°C and FUdR was added at L4 stage. At L4, D2, and D9 of adulthood, animals were imaged in a Zeiss Axio Imager M2m microscope and the number of animals showing axonal breaks in the VD and DD motor neurons were scored. Axonal breaks were defined as an area of discontinued fluorescence in either the ventral or dorsal nerve cords.

The visualization and quantification of the progressive decline in proteostasis in the nervous system (specifically the nerve ring) was performed as described in Brignull et al., 2006; Lazaro-Pena et al., 2021. In brief, synchronized rgef-1p::polyQ40::YFP L4 animals were treated with FUdR. Then, z-stack images from the nerve ring were acquired using a Zeiss Axio Imager M2m microscope with AxioVision v4.8.2.0 software at ×63 magnification. The fluorescent foci form compressed z-stack images from 20 animals per technical replicate were scored blind every other day from 4 to 8 days of adulthood.

To assess locomotion capacity, at day 2 of adulthood, 40 animals were transferred to a drop of 10 µl of M9 solution and the number of body bends performed in a period of 30 s from each animal was scored (as described in Lazaro-Pena et al., 2021). Statistical testing between pairs of conditions for differences in the number of foci was performed using Student’s t-test analysis.

The visualization and quantification of the progressive decline in proteostasis in muscle tissue was performed as described in Lazaro-Pena et al., 2021; Morley et al., 2002. Briefly, synchronized unc-54p::Q35::YFP L4 animals were treated with FUdR. Then, fluorescent foci from 20 animals per technical replicate were scored blind daily from days 1 to 3 of adulthood. To assess paralysis, at days 5, 7, 9, and 11 of adulthood, prodded animals that responded with head movement (and were therefore still alive) but no backward or repulsive motion were scored as paralyzed (as described in Lazaro-Pena et al., 2021). Statistical testing between pairs of conditions for differences in the number of foci was performed using Student’s t-test analysis. Two to five independent transgenics lines were tested.

The aldicarb assay was performed as described in Mahoney et al., 2006; Oh and Kim, 2017. Plates containing aldicarb 1 mM were prepared the day before the assay and stored at 4°C. Animals were synchronized by picking at L4 stage and placing them in a plate with HT115 bacteria. After 24 hr, one aldicarb plate per strain were brought out of 4°C and a small drop of OP50 was placed in the center of the plate and let dry for 30 min. Approximately 25 animals were placed on each plate, and the number of paralyzed animals was scored every hour up to 5 hr. The assay was performed at RT.

Animals were synchronized by picking at L4 stage and placing them in a plate with HT115 bacteria. After 24 hr, serotonin (5-HT) was dissolved in M9 buffer to a 20 mM concentration (Hart, 2006). Twenty worms were placed on a glass plate well with 400 µl of 20 mM serotonin for 10 min. The same number of control animals were placed in 400 µl of M9 buffer. The locomotion of worms (mobilized or immobilized) was annotated at the following time points: 2, 4, 6, 8, and 10 min. The assay was performed at RT.

10 cm RNAi plates were seeded with 1 ml of 10× concentrated HT115 EV bacteria from an overnight culture, and allowed to dry for 1–2 days. Approximately 1000 synchronized L1 animals were added to each plate (~3000 animals per condition across multiple plates). At the L4 stage, 600 µl of 4 mg/ml FUdR was added. Animals were kept at 20°C for the duration of the experiment.

Day 2 adult animals were collected in M9 buffer, washed 2× in M9, and a final wash with DEPC-treated RNase-free water. Approximately two times the volume of the animal pellet of TRIzol reagent was added to each animal preparation, followed by brief mixing and freezing overnight at –80°C. Tubes were then allowed to partially thaw, and were vortexed for 5 min to assist with disrupting the cuticle. Samples were transferred to new tubes, and 200 µl of chloroform per 1 ml of TRIzol was added, followed by 20 s vortexing; the tubes settled at RT for 10 min. Supernatants were transferred to new tubes, and an equal amount of 100% ethanol was added and mixed before proceeding to column purification with the QIAGEN RNAeasy Mini kit according to the manufacturer’s instructions. Samples were eluted with 30–50 µl of DEPC-treated water, and checked for initial concentration and quality with a Nanodrop ND-1000 spectrophotometer. Biological replicate samples were prepared from independently synchronized populations of animals.

Isolated RNA was provided to the University of Rochester Genomics Research Center for library preparation and sequencing. Prior to library preparation, RNA integrity of all samples was confirmed on an Agilent 2100 Bioanalyzer. Libraries for sequencing were prepared with the Illumina TruSeq Stranded mRNA kit, according to the manufacturer’s instructions. Quality of the resulting libraries was checked again with a Bioanalyzer prior to sequencing to ensure integrity. Sequencing was performed on an Illumina HiSeq2500 v4, yielding an average of approximately 31 million single-ended 100 bp reads per sample. Quality of the output was summarized with FastQC (Andrews, 2010) and reads were trimmed and filtered with Trimmomatic to remove adapter sequence and any low-quality content occasionally observed toward the ends of reads (Bolger et al., 2014). After filtering out low-quality reads, an average of 30 million reads per sample remained, and were used for the rest of the analysis.

RNA-Seq reads were aligned to the C. elegans genome (assembly WBcel235) with STAR 2.4.2a (Dobin et al., 2013) using gene annotation from Ensembl (version 82) (Cunningham et al., 2022). An average of 90.2% of reads per sample were uniquely and unambiguously mapped to the genome. These alignments were used as input to featureCounts (version 1.4.6-p5) for gene-level counts (Liao et al., 2014). Ambiguous or multi-mapping reads, comprising an average of approximately 9% of the reads per sample, were not included in the gene-level count results. Transcript-level quantification was also performed with RSEM (Li and Dewey, 2011) to obtain TPM (transcripts per million) expression estimates.

Further analysis was performed primarily in the R statistical software environment (version 4.0.2) (Team, 2013), utilizing custom scripts and incorporating packages from Bioconductor. Both count and TPM expression matrices were filtered to remove genes with low or no expression, keeping genes with at least 10 read counts in at least one sample. Genes were further filtered to keep only those annotated as protein coding, pseudogene, ncRNA, lincRNA, antisense, or snRNA, which could have been included in the sequenced libraries after poly-A selection; 96.3% of genes included in the final set are protein coding. Gene identifiers have been updated as necessary from newer versions of WormBase (WS284) for integration with recent resources and datasets.

Differential expression analysis was performed with DESeq2, and the optional fold-change shrinkage procedure was applied to moderate fold-changes for low-expression genes (Love et al., 2014). Genes significantly differentially expressed between hpk-1(pk1393) and N2 animals were taken as those with FDR-adjusted p-value less than 0.05 and absolute value of log₂ fold-change of at least 1 (on a linear scale, down- or upregulated by at least twofold). Functional and over-representation analysis for differentially expressed genes was performed with GOSeq (Young et al., 2010) using gene sets and pathways from the GO (Harris et al., 2004), KEGG (Kanehisa and Goto, 2000), and Reactome (Jassal et al., 2020).

Unless otherwise specified, GraphPad PRISM version 7 software was used for statistical analyses. Data were considered statistically significant when p-value was lower than 0.05. In figures, asterisks denote statistical significance as (*, p<0.05, **, p<0.001, ****, p<0.0001) as compared to appropriate controls. N number listed in figure legend indicates the number of one representative experiment among all biological trials.