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

Aging and the age-associated decline of the proteome is determined in part through neuronal control of evolutionarily conserved transcriptional effectors, which safeguard homeostasis under fluctuating metabolic and stress conditions by regulating an expansive proteostatic network. We have discovered the Caenorhabditis elegans homeodomain-interacting protein kinase (HPK-1) acts as a key transcriptional effector to preserve neuronal integrity, function, and proteostasis during aging. Loss of hpk-1 results in drastic dysregulation in expression of neuronal genes, including genes associated with neuronal aging. During normal aging hpk-1 expression increases throughout the nervous system more broadly than any other kinase. Within the aging nervous system, hpk-1 induction overlaps with key longevity transcription factors, which suggests that hpk-1 expression mitigates natural age-associated physiological decline. Consistently, pan-neuronal overexpression of hpk-1 extends longevity, preserves proteostasis both within and outside of the nervous system, and improves stress resistance. Neuronal HPK-1 improves proteostasis through kinase activity. HPK-1 functions cell non-autonomously within serotonergic and γ-aminobutyric acid (GABA)ergic neurons to improve proteostasis in distal tissues by specifically regulating distinct components of the proteostatic network. Increased serotonergic HPK-1 enhances the heat shock response and survival to acute stress. In contrast, GABAergic HPK-1 induces basal autophagy and extends longevity, which requires mxl-2 (MLX), hlh-30 (TFEB), and daf-16 (FOXO). Our work establishes hpk-1 as a key neuronal transcriptional regulator critical for preservation of neuronal function during aging. Further, these data provide novel insight as to how the nervous system partitions acute and chronic adaptive response pathways to delay aging by maintaining organismal homeostasis.


Introduction
The gradual decline of function within the proteome (proteostasis) is a characteristic of aging, which precipitates the onset and progression of a growing number of age-associated diseases (Balch et al., 2008;Gidalevitz et al., 2006;Gidalevitz et al., 2010;Kikis et al., 2010;Morimoto, 2011;Sala et al., 2017;Taylor and Dillin, 2011): Alzheimer's disease (AD), Parkinson's disease (PD), Huntington's disease, and amyotrophic lateral sclerosis (ALS) are neurodegenerative diseases driven by genetic alterations that typically predispose a mutant protein isoform to aggregate and have toxic gain of function properties on the neuronal proteome. Outside of the nervous system, age-associated proteostatic decline also leads to disease, for example, the onset of diabetes mellitus (Jaisson and Gillery, 2014); thus, the decline in proteostasis during normal aging is not limited to the nervous system. A growing number of studies suggest that aging is not simply the result of stochastic accumulation of damage, but is determined through genetics and coordinated mechanisms across tissues and cell types (Labbadia and Morimoto, 2015;Lazaro-Pena et al., 2022;López-Otín et al., 2013). Proteostatic decline corresponds with the age-associated breakdown of a large proteostasis network (PN), which integrates stress-responsive control of protein folding, degradation, and translation in response to myriad cell intrinsic and non-autonomous signals. And yet, while major components of the proteostatic network have been identified, how multi-cellular organisms maintain systemic proteostasis within and across tissues to delay aging is still poorly understood.
The relatively simple metazoan animal Caenorhabditis elegans (C. elegans) is a premier model system to elucidate how proteostasis is coordinated across cell and tissue types in response to myriad signals. Discoveries of cell non-autonomous signaling first made in C. elegans have evolutionarily conserved components, and revealed that the nervous system acts as a key cell non-autonomous regulator of organismal proteostasis and longevity (reviewed in Miller et al., 2020). For example, a pair of thermosensory neurons systemically regulate the heat shock response (HSR) in a serotonin-dependent manner (Prahlad et al., 2008;Tatum et al., 2015). In contrast, a second regulatory component from the GABAergic and cholinergic systems normally limits muscle cell proteostasis (Garcia et al., 2007;Silva et al., 2013).
The first discovery of a putative link between the transcriptional cofactor homeodomain-interacting protein kinase (HPK-1) and longevity was from our genome-wide RNAi screen that identified 103 genes essential for decreased insulin-like signaling (ILS) to extend longevity (Samuelson et al., 2007). In C. elegans, hpk-1 preserves proteostasis, stress response, and organismal longevity (Berber et al., 2016;Das et al., 2017;Samuelson et al., 2007). Additionally, we have previously shown that HPK-1 extends longevity through distinct genetic pathways defined by HSF-1 and the target of rapamycin complex 1 (TORC1) (Das et al., 2017). Activation by either metabolic or genotoxic stressors has been observed from yeast to mammals, suggesting that this family of transcriptional cofactors arose early in evolution to couple metabolic and stress signaling (Hofmann et al., 2003;Kim et al., 2006;Lee et al., 2008;Shojima et al., 2012).
We sought to identify the tissue and cell types in which the HPK-1 transcriptional circuits act to extend longevity, increase resistance to acute proteotoxic stress, and which PN components are regulated across tissues. We have discovered that HPK-1 functions as a key regulator of the proteostatic response, originating in the nervous system of C. elegans. We find that hpk-1 is the most broadly upregulated kinase during normal C. elegans aging, predominantly within the nervous system, and acts to preserve neuronal function and proteostasis within the nervous system. Further, HPK-1 kinase activity within the nervous system initiates cell non-autonomous signals that trigger protective peripheral responses to maintain organismal homeostasis. Responses can qualitatively differ depending on the neuronal type from which they arise: serotonergic HPK-1 activity increases thermotolerance and protects the proteome from acute stress by enhancing the HSR, without increasing lifespan or altering basal autophagy. In contrast, GABAergic HPK-1 activity induces autophagy and extends longevity, without altering thermotolerance. HPK-1 overexpression is sufficient to induce autophagy gene expression, and autophagosome formation requires hlh-30 (TFEB), mxl-2 (MLX), and daf-16 (FOXO) TFs. We posit that serotonergic HPK-1 signaling responds to acute proteotoxic stress, whereas GABAergic HPK-1 responds to chronic metabolic changes mediated through TORC1 (Das et al., 2017). Yet, each of these distinct adaptive responses is sufficient to activate the PN and improve proteostasis outside of the nervous system. Our findings reveal HPK-1 as a key neuronal regulator in coordinating adaptive metabolic and stress response pathways across tissues to delay aging throughout the animal.

HPK-1 acts from neuronal tissue to promote longevity and healthspan
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 eLife digest Proteins are essential for nearly every cellular process to sustain a healthy organism.
A complex network of pathways and signalling molecules regulates the proteins so that they work correctly in a process known as proteostasis.
As the body ages, this network can become damaged, which leads to the production of faulty proteins. Many proteins end up being misfolded -in other words, they are misshapen on the molecular level, which can be toxic for the cell. A build-up of such misfolded proteins is implicated in several neurological conditions, including Alzheimer's, Parkinson's and Huntington's disease.
Cells have various ways to detect and respond to internal stressors, such as tissue or organ damage. For example, specific proteins in the nervous system can raise a 'central' alert when damage is detected, which then primes and coordinates the body's systems to respond in the peripheral cells and tissues. But exactly how this happens is still unclear.
To find out more about the central coordination of stress responses, Lazaro-Pena et al. studied one such sensor protein, called HPK-1, in the roundworm C. elegans. They first overexpressed the protein in various tissues. This revealed that only when HPK-1 was overactive in nerve tissue, it protected proteins and prolonged the lifespan of the worms. An increased amount of HPK-1 improved the health span of the worms and older worms also moved better. However, genetically manipulated worms lacking HPK-1 in their nerve cells showed a faster decline in nervous system health as they aged, which could be reversed once HPK-1 was activated again.
Lazaro-Pena et al. then measured the amount of HPK-1 in worms at different stages of their life. This showed that as the worms aged, the amount of HPK-1 increased in the nerve cells. The nerve cells in which HPK-1 levels increased overlapped with an increased expression of proteins associated with longevity. Moreover, when HPK-1 was overexpressed, it stimulated the release of other cell signals, which then triggered protective responses to prevent the misfolding and aggregation of proteins and to help degrade damaged proteins.
This study shows for the first time that HPK-1 appears to play a protective role during normal ageing and that it may act as a key switch to stimulate other protective mechanisms. These findings may give rise to new insights into how the nervous system can coordinate many different stress responses, and ultimately delay ageing throughout the whole body.
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.

hpk-1 expression increases broadly during normal aging
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. Video 1. Homeodomain-interacting protein kinase (HPK-1) functions within the nervous system to regulate healthspan. Video of wild-type, hpk-1(pk1393) null animals, and hpk-1 null animals with neuronal expression of hpk-1 (rab-3p::HPK-1). Videos were generated from consecutive images taken approximately every 15 min over the course of the animals lifespan on 6 cm plates using the Lifespan Machine method (Stroustrup et al., 2013 Figure 2. During normal aging hpk-1 is the most broadly upregulated kinase and overlaps with key longevity-associated transcription factors (TFs) within the nervous system. (A) Relative expression of hpk-1 mRNA during normal aging, as measured by RT-qPCR. (B) Venn diagram depicting TFs with significant age-associated differential expression alongside hpk-1 within 3 or more of the 36 neuronal and/or 11 non-neuronal cell clusters (yellow and blue, respectively), and the intersection with TFs that have previously been implicated in C. elegans longevity (red). (C, D) Heat map of neuronal cell clusters in which hpk-1 expression increases during normal aging along with TFs that have the most broadly co-occurring differential expression with age (≥9 cell-clusters). Positive (red) and negative (blue) fold-change with aging of a given TF are shown in (C) and (D), respectively, grouped by average fold-change across the indicated clusters. Neuronal types, subtypes, cell cluster, or individual neuron pair, as well as whether a TF has previously been implicated in aging is indicated. Primary data for this analysis was generated in the C. elegans Aging Atlas (Roux et al., 2022) and significant differences in expression with aging were determined by filtering differential expression results from a comparison of old vs young animals, based on adjusted p-values <0.05 and a log 2 fold-change magnitude of at least 0.5 (between days 8,11,15 and days 1,3,5). (E) Relative expression of hpk-1, xbp-1, and gei-3 mRNA of day 11 adult animals, compared to day 1 as measured by RT-qPCR. (F) Representative images of hpk-1 expression within the nervous system of day 1 and 11 adult animals (hpk-1p::mCherry; unc-47p::GFP). See Figure 7F and (I) hpk-1 is differentially expressed with age in more cell clusters than any other kinase, and mostly in neurons. Heat map shows log 2 fold-change in old vs young animals for kinases with differential expression in ≥10 cell clusters. Primary data for this analysis was generated in the C. elegans Aging Atlas (Roux et al., 2022) and filtered for significant age-associated expression changes as described in (C, D). On panels (A and E), bars represent ± SEM of 3 technical replicates with more than 500 animals per condition. t-Test analysis with *p<0.05, **p<0.01, ***p<0.001,and ****p<0.0001. See Supplementary file 4 for additional RT-qPCR details, and Supplementary files 5 and 6 for dataset.
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).
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 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.

HPK-1 preserves integrity of the C. elegans 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.

Loss of hpk-1 results in neuronal dysregulation of neuronal gene expression
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 log 2 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). of day 2 adult hpk-1 null mutant animals (n=3) compared to N2 wild-type animals (n=3) identified 2201 genes with significantly altered expression with loss of hpk-1 (adjusted p-value <0.05 and |log 2 fold-change|≥1), 84% of which were upregulated in the mutants. The volcano plot illustrates the criteria applied for selecting genes with significant and substantial expression changes, shown as blue dots: vertical dashed lines are the fold-change threshold and the horizontal dashed line is the p-value threshold. (B, C) Enrichment for functions and pathways associated with the 348 downregulated genes (B) or the 1853 upregulated genes (C) in hpk-1 animals yields innate immune response as the most significantly over-represented term for downregulated genes, but reveals a large number of significant associations to pathways and processed for the upregulated genes -many of which are associated with neurons and neuronal signaling. Each bar represents a functional term or pathway from the Gene Ontology (GO), KEGG, or Reactome databases, as indicated. Bar colors show the -log 10 transformed enrichment p-value. The number in each bar is the size of the overlap between the set of differentially expressed genes and the genes in the term or pathway. The x-axis indicates the fraction of the total genes in the term or pathway that were differentially expressed. All results shown have significant enrichment after adjustment for multiple testing. (D) Functional enrichment for 283 genes upregulated with hpk-1 loss that are specific to, or enriched in, C. elegans neurons shows broad dysregulation of functions associated with neurons including neurotransmitter transport and release, cell-cell junctions, axons, neuropeptides, and sensory processes, among others; 880 genes uniquely expressed or significantly enriched in neurons were derived from bulk tissue-specific RNA-Seq profiling results in Kaletsky et al., 2018, and 32% of these are significantly upregulated in hpk-1 null animals. See Supplementary file 10 and Supplementary file 11 for dataset and additional analysis.
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 ageassociated 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).

Neuronal HPK-1 regulates proteostasis cell intrinsically and cell nonautonomously via neurotransmission
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 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     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 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 nonautonomous signaling occurs through neurotransmission.

Neuronal HPK-1 induces molecular chaperones and autophagy in distal tissues
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.
GABAergic and serotonergic HPK-1 differentially regulate longevity and acute stress signaling 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.
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 Mycfamily 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.

Discussion
It has been known that a core group of TFs integrates various types of metabolic and stress signals; moreover, this group of TFs can extend longevity when overexpressed, and conversely hasten aging when lost (reviewed in Denzel et al., 2019), but whether the activity or expression of these longevityassociated transcriptional regulators are activated by signals of increasing damage or dysfunction during normal aging was unknown. Recently, single-cell RNA-Seq has revealed that these longevityassociated TFs increase in expression during aging in wild-type animals across multiple somatic tissues (Roux et al., 2022). This comprehensive empirical dataset provides foundational support for a key theme in aging research: during aging in wild-type animals, increasing damage and dysfunction is mitigated by the activation of stress response programs. Our findings that: (1) hpk-1 expression is upregulated during normal aging in more neurons and non-neuronal cells than any other kinase, and (2) that hpk-1 is induced in cell clusters that overlap with longevity-associated TFs, including daf-16, hlh-30, skn-1, and hif-1, implies that HPK-1 is a key transcriptional regulator that mitigates aging. We posit this response is activated in the nervous system during normal aging in wild-type animals to mitigate increasing stress and the breakdown of homeostatic mechanisms within the nervous system. Consistently, the presence of human neurodegenerative disorders, a severe form of neuronal proteotoxic stress, correlates with activation of human hpk-1 expression: for example, human HIPK2 is activated in frontal cortex samples of patients with AD, in motor neurons of ALS patients, and in mouse models of ALS . Thus, our results suggest that overexpression of hpk-1 in the C. elegans nervous system may extend longevity by preemptively delaying the accumulation of age-associated damage and dysfunction to the proteome ( Figure 9A). However, hpk-1 homologs in more complex metazoans can also promote cell death, a key homeostatic component of tissue regeneration, indicating that chronic activation of HIPKs is not always beneficial (Calzado et al., 2007;Lee et al., 2016;Wang et al., 2020;Zhang et al., 2020). Whether activated HIPKs will fortify proteostasis and extend longevity, or promote cell death is likely to be dependent on many factors, including the specific homolog, levels of expression, physiological context, genetic background, and cell type. Figure 9. Differential regulation of the proteostatic network through homeodomain-interacting protein kinase (HPK-1) activity in serotonergic and γ-aminobutyric acid (GABA)ergic neurons. (A) During normal aging hpk-1 expression increases broadly throughout the nervous system in response to accumulating damage and dysfunction. Overexpression of hpk-1 within the nervous system fortifies proteostasis by priming and preemptively activating mechanisms that delay the progression of aging. In the absence of hpk-1 the proteostatic network is compromised, resulting in neuronal dysfunction and increased expression compensatory mechanisms to maintain homeostasis (see text for details). (B) HPK-1 activity in serotonergic and GABAergic neurons initiates distinct adaptive responses, either of which improve proteostasis in a cell non-autonomous manner. Serotonergic HPK-1 protects the proteome from acute heat stress, while GABAergic HPK-1 fortifies the proteome by regulating autophagy activity in response to metabolic stress, such as changes in target of rapamycin complex 1 (TORC1) activity. The MXL-2, HLH-30 (TFEB), and DAF-16 (FOXO) transcription factors are all required for hypodermal induction of autophagy in response to increased HPK-1 activity in GABAergic neurons. Gray indicates predicted interactions based on our prior genetic analysis (Das et al., 2017). MML-1 and MXL-2 heterodimerize and encode the orthologs of the Myc-family of transcription factors Mondo A/ChREBP and Mlx, respectively.
We find that hpk-1 prevents age-associated decline of axonal and synaptic integrity, which together preserve neuronal function during the aging process and thus also promote longer healthspan. In support of this concept, adult hipk2-/-mice display normal neurogenesis but manifest delayedonset severe psychomotor and behavioral deficiencies, as well as degeneration of dopaminergic neurons, reminiscent of PD (Zhang et al., 2007). Loss of hpk-1 results in dysregulation of neuronal gene expression by early adulthood; some changes are consistent with neuronal aging but in other cases we observe an inverse correlation. For example, one-third of all neuropeptides are induced in the absence of hpk-1; yet increased neuropeptide signaling has been associated with extended C. elegans longevity and is considered protective in AD Hodge et al., 2022;Petrella et al., 2019). One possibility is in the absence of hpk-1, attenuated inducibility of the proteostatic network accelerates age-associated cellular dysfunction ( Figure 9A), and upregulation of neuropeptide signaling compensates in the nervous system for the loss of proteostatic response. This concept is consistent with our finding that increasing neuronal HPK-1 activity does not require neuropeptides to improve proteostasis in distal tissues. An alternative possibility for the connection between protective effects of HPK-1 and neuropeptides in neurons is that HPK-1 may reduce insulin signaling by suppressing the processing of ins peptides, thereby decreasing ILS to activate DAF-16 (FOXO) transcriptional quality control programs. For example, loss of egl-3, a neuropeptide processing gene, extends longevity by inactivating processing of insulins (Hamilton et al., 2005), and egl-3 is upregulated in hpk-1 loss of function.
Impairment of the proteostatic network after loss of hpk-1 results in broad upregulation of gene expression, which may provide additional levels of compensatory mechanisms to mitigate neuronal dysfunction. For example, loss of hpk-1 induces expression of intraflagellar transport (IFT) components in sensory cilia. Mutations within sensory neurons can alter longevity (Alcedo and Kenyon, 2004;Apfeld and Kenyon, 1999), and decreased IFT during C. elegans aging impairs sensory perception and metabolism (Zhang et al., 2021). Deficiencies in primary cilia impairs neurogenesis in AD transgenic mice, resulting in accumulation of Aβ 42 and tau protein and disease progression (Armato et al., 2013;Chakravarthy et al., 2012;Ma et al., 2022;Morelli et al., 2017). However, our bulk RNA-Seq analysis cannot decipher changes in gene expression at the level of individual neurons after loss of hpk-1. 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 cannot preclude the possibility that HPK-1 may function to broadly limit neuronal gene expression during the course of normal development and differentiation, thereby specifying neuronal transcriptional programs. Mis-differentiation of the nervous system in the absence of hpk-1 could result in both intracellular transcriptional noise and cell non-autonomous noise across cell types and tissues.
We show that HPK-1 expression in serotonergic neurons activates a cell non-autonomous signal to prime induction of molecular chaperones in muscle cells, whereas HPK-1 expression in GABAergic neurons signals induces autophagy in hypodermal tissue ( Figure 9B). Our findings place HPK-1 within a selective cohort of transcriptional regulators within the C. elegans nervous system that are activated in response to metabolic, environmental, and intrinsic cues; thereby initiating distinct cell nonautonomous signals to regulate components of the PN in peripheral tissues (Douglas et al., 2015;Durieux et al., 2011;Frakes et al., 2020;Imanikia et al., 2019;Prahlad et al., 2008;Prahlad and Morimoto, 2011;Silva et al., 2013;Tatum et al., 2015;Taylor and Dillin, 2013;Morimoto, 2020). Loss of serotonin production in tph-1 mutant animals has a minimal impact on lifespan at basal temperature (Murakami and Murakami, 2007). However, simultaneous loss of both serotonin and DA biosynthesis shortens lifespan (Murakami and Murakami, 2007). Serotonin biosynthesis is also required for several conditions that extend longevity: animals treated with an atypical antidepressant serotonin antagonist (mianserin) (Petrascheck et al., 2007), changes in food perception (Entchev et al., 2015), and neuronal cell non-autonomous induction of the mitochondrial unfolded protein response (Berendzen et al., 2016;Zhang et al., 2018). In contrast, loss of either of the serotonin receptors ser-1 or ser-4 have opposing effects on lifespan, suggesting antagonistic activity between serotonin receptors (Murakami and Murakami, 2007).
Surprisingly, we find that increased HPK-1 activity in serotonergic neurons does not increase lifespan, but is sufficient to improve muscle proteostasis and amplify the expression of molecular chaperones that maintain proteostasis within the cytosol and nucleus. Interestingly, neuronal HSF-1 regulation of the HSR through thermosensory circuits is also separable from lifespan (Douglas et al., 2015). Serotonin signaling and the serotonin receptor ser-1 are essential for the induction of the HSR (Prahlad et al., 2008;Tatum et al., 2015). ADF serotoninergic neurons receive upstream stress signals from the thermosensory circuit, consisting of the amphid sensory neurons (AFD) and the AIY interneurons, which regulate HSF-1 activity in the germline in a serotonin-dependent manner (Tatum et al., 2015). However, increased serotonergic HPK-1 activity is not equivalent to altered thermosensory signals: hpk-1 improves proteostasis within body wall muscles but not intestinal cells through neurotransmitters, and is independent of dense core vesicles. In contrast, loss of the guanylyl cyclase (gcy-8) in AFD thermosensory neurons or the LIM-homeobox TF (ttx-3) in AIY interneurons improves intestinal proteostasis through dense core vesicle neurosecretion (i.e., unc-31). hpk-1 is expressed in AFD, but only at very low levels in AIY neurons (data not shown and Taylor et al., 2021), and hpk-1 is essential for the induction of the HSR within intestinal cells (Das et al., 2017). Whether hpk-1 activity within thermosensory neurons functions within the canonical thermosensory-serotonin circuit to regulate the intestinal HSR or proteostasis is not mutually exclusive from our findings and will be interesting to pursue in future studies. Our results suggest that serotonergic HPK-1 is an important regulator of the HSR, potentially functioning in parallel to the thermosensory-serotonin circuit to restore proteostasis in response to acute thermal stress.
We posit that increased serotonergic HPK-1 activity further activates HSF-1 in response to heat ( Figure 9B). This would be consistent with known hyperphosphorylation of HSF-1 and amplification of the HSR in response to increasingly severe heat stress (reviewed in Lazaro-Pena et al., 2022). Interestingly, HPK-1 limits SUMOylation of HSF-1, SUMO limits the HSR in both C. elegans and mammals, and severe heat shock results in HSF1 deSUMOylation (Das et al., 2017;Hietakangas et al., 2003;Hietakangas et al., 2006). To our surprise, while increased neuronal hsf-1 expression increases longevity (Douglas et al., 2015;Morley and Morimoto, 2004), we could find no study that has identified specific neuronal cell types in which increased HSF-1 activity is sufficient to extend longevity or enhance the HSR. Whether or not HPK-1 regulates HSF-1 within the nervous system and/ or in peripheral muscle cells through cell non-autonomous signals will be interesting to determine in future studies. HPK-1 co-localizes with HSF-1 in neurons, HPK-1 protein is stabilized after heat shock, hpk-1 is essential for induction of the HSR, and hsf-1 and hpk-1 are mutually interdependent to extend longevity (Das et al., 2017; Figure 9B).
Increased HPK-1 activity in GABAergic neurons induces autophagy in peripheral cells, improves proteostasis and longevity. Whether HPK-1 functions in GABAergic neurons through alteration of GABA or another neurotransmitter is unknown, but proper differentiation of GABAergic neurons is essential for neuronal HPK-1 activation of autophagy in distal tissue. We posit that HPK-1 cell nonautonomous signaling occurs via antagonism of GABA signaling. Loss of GABA increases C. elegans lifespan (Chun et al., 2015;Yuan et al., 2019). However, not all neurotransmitters regulate lifespan, as only mutants for the GABA biosynthetic enzyme, unc-25, increase lifespan. This GABAergic-mediated control in lifespan seems to be mediated by the gbb-1 metabotropic GABA B receptor (Chun et al., 2015). Interestingly, GABA signaling inhibits autophagy and causes oxidative stress in mice, which is mitigated via Tor1 inhibition (Lakhani et al., 2014). In fact, we previously found that neuronal expression of HPK-1 is inhibited by TORC1 and hpk-1 is required for TORC1 inactivation to extend longevity and the induction of autophagy gene expression (Das et al., 2017). Both TORC1 inactivation and hpk-1 overexpression requires the Myc-family member mxl-2 to extend longevity (Das et al., 2017;Johnson et al., 2014;Nakamura et al., 2016; Figure 8-figure supplement 2). Consistently, TORC1 also acts within neurons to regulate aging (Zhang et al., 2019); our results suggest that GABAergic neurons might be the specific neuronal cell type through which alterations in TORC1 activity regulate longevity. Collectively, our results position HPK-1 activity in GABAergic neurons as the likely integration point for cell non-autonomous regulation of autophagy via TORC1 ( Figure 9B).
A growing number of longevity-associated TFs have been linked to autophagy, including pha-4 (FOXA), mxl-2 (MLX), hlh-30 (TFEB), daf-16 (FOXO), and hsf-1 (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). We find that hlh-30, daf-16, and hlh-30 are all required for the induction of autophagosomes when hpk-1 is overexpressed in GABAergic neurons. We posit these TFs respond within the distal hypodermal seam cells in response to a neuronal signal initiated by HPK-1, as neurons are largely refractory to RNAi. However, this does not preclude cell intrinsic neuronal interactions; future refined spatial genetic analysis will be required to definitively decipher the precise tissues of action. DAF-16 and HLH-30 have previously been shown to function in combination to promote stress resistance and longevity (Lin et al., 2018). Analogously, the MML-1::MXL-2 complex and HLH-30 activate each other and are downstream effectors of longevity after TORC1 inhibition (mml-1 encodes the MondoA/ ChREBP ortholog, an obligate interactor with MXL-2) (Nakamura et al., 2016). Based on our discovery that all three of these longevity-associated TFs are essential for GABAergic HPK-1 activity to induce autophagy, it is tempting to speculate HPK-1 acts as a key mediator of pro-longevity signaling after TORC1 inhibition through coordinated activation of MML-1::MXL-2, HLH-30, and DAF-16 ( Figure 9B). This work reveals a new level of specificity in the capacity of the nervous system to initiate adaptive responses by triggering distinct components of the proteostatic network across cell types. Results elucidating how neuronal cell types generate heterotypic signals to coordinate the maintenance of homeostasis are just beginning to emerge. For example, a previous study discovered that activating non-autonomous signaling of the ER-UPR via the XBP-1 TF can be initiated from serotonergic neurons and dopaminergic neurons; these signals are unique and converge to activate distinct branches of the ER-UPR within the same distal tissues (intestinal cells) (Higuchi-Sanabria et al., 2020). In contrast, our work demonstrates differential, yet specific, activation of cytosolic components of the PN in separate peripheral tissues through neurotransmitters. Our work suggests that the nervous system partitions neuronal cell types to coordinate the activation of complementary proteostatic mechanisms across tissues, despite utilizing the same transcriptional regulator within each neuronal cell type. One possibility is that HPK-1 regulates specific adaptive responses based on the unique epigenetic states between differentiated neuronal subtypes, which would be consistent with recent findings: during development and differentiation, transcriptional rewiring of cells sets chaperoning capacities and alters usage of mechanisms that maintain protein quality control (Nisaa and Ben-Zvi, 2022;Shemesh et al., 2021). Our work reveals HPK-1 as a novel transcriptional regulator within the nervous system of metazoans, which integrates diverse stimuli to coordinate specific adaptive responses that promote healthy aging. 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/). All plasmids generated in this study are available upon request from the Samuelson laboratory.

Continued on next page
All primers were generated at Integrated DNA Technologies (https://www.idtdna.com/pages).

C. elegans strains and details
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/).

RNAi feeding
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.

RT-qPCR analysis
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.

Lifespan analysis
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.

Thermotolerance assay
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.
Induction of hsp-16.2p::GFP after heat shock 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.
Measurement of autophagosome formation using the GFP::LGG-1/ Atg8 reporter 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.

Measurement of neurodegeneration in motor neurons
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.

Polyglutamine aggregation in neurons and locomotion analyses
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.

Polyglutamine aggregation in muscle and paralysis analyses
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.

Aldicarb assay
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.

Exogenous 5-HT immobilization assay
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.
Sample preparation and sequencing for RNA-Seq 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 DEPCtreated 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.
Analysis of hpk-1 null and wild-type control RNA-Seq dataset 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 log 2 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).
Integration with public tissue-specific and single-cell RNA-Seq datasets C. elegans Aging Atlas and CeNGEN single-cell datasets We obtained differential expression results for old (days 8, 11, 15) vs young (days 1, 3, 5) animals across 200 cell clusters from the Aging Atlas single-cell RNA-Seq project (Roux et al., 2022). Significantly differentially expressed genes within each cluster were taken as those with adjusted p-value <0.05, and log 2 fold-change magnitude of at least 0.5 (approximately 1.4 fold up-or downregulated on a linear scale). Additional neuron identity information was obtained by mapping Aging Atlas neuron clusters to clusters identified in late-larval stage animals in the CeNGEN single-cell neuronal transcriptome project (Taylor et al., 2021), assisted by correlations between clusters of the two datasets already provided with the Aging Atlas. Chemical signaling subtypes of neurons were determined by expression of marker genes commonly used in the field for fluorescent labeling of neurons: tph-1 for serotonergic, unc-47 for GABAergic, eat-4 for glutamatergic, cat-2 for dopaminergic, cat-2 for cholinergic. Where a high correlation had been established between a cluster in the Aging Atlas and CeNGEN, a signaling type was assigned if the cluster expressed the associated marker gene at TPM greater than 75 based on the provided pre-normalized expression matrix available from the CeNGEN website. Ten Aging Atlas neuron clusters did not have an obvious corresponding cluster in the CeNGEN dataset; for these clusters neuron signaling class was based on marker genes having average cluster counts >0.5 across all cells in a cluster and, where possible, a consensus of the six Aging Atlas time points.
Normalized single-cell aging time course gene expression was obtained by processing raw counts in the H5AD file from the Aging Atlas resource website (http://c.elegans.aging.atlas.research.calicolabs.com/data). Briefly, the H5AD data from SCANPy was read in and converted into a Seurat object with MuDataSeurat (https://github.com/PMBio/MuDataSeurat) (Hao et al., 2021). SCTransform, as implemented in Seurat, was used to normalize raw counts for library size differences, yielding logscale normalized expression values for each cell that were used for further analysis.
Genes uniquely expressed in neurons or significantly enriched compared to other tissues were obtained from tissue-specific bulk RNA-Seq reported by Kaletsky et al., 2018, due to the additional dynamic range provided over current single-cell sequencing technology.
C. elegans kinome, TFs, neuronal genome, and longevity-associated genes A collection of C. elegans kinases was assembled based on Zaru et al., 2017, and cross-referenced against current gene annotation in WormBase version WS284 (Harris et al., 2020) to keep only protein-coding genes and remove pseudogenes or 'dead' genes. This produced a list of 438 kinases in the C. elegans kinome (Figure 2-figure supplement 4, Supplementary file 5). C. elegans TFs were identified from a previously published compendium (Nye et al., 1989). Genes with known or predicted functions important to neurons, and their associated roles, were assembled from tables curated in Hobert, 2013. Genes with lifespan phenotypes primarily from perturbation studies in C. elegans utilizing RNAi or mutant strains were assembled from annotation in WormBase (WS282) as well as the GenAge database (de Magalhães and Toussaint, 2004).

Statistical analysis for in vivo assays
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. would like to thank Dr. Doug Portman (URMC) and his laboratory for strains and advice. Some strains were provided by the CGC, which is funded by NIH Office of Research Infrastructure Programs (P40

Additional files
Supplementary files • Supplementary file 1. Trials, data, strains, genotypes, number of animals, and statistics for lifespan data presented in Figures 1a-d, f, 8b • Supplementary file 2. Experimental data, strains, genotypes, number of animals, and statistical analysis for data presented in Figures 1e, 5c and e.
• Supplementary file 3. Trials, data, analysis, strains, genotypes, number of animals, and statistics for thermotolerance data presented in Figures 1g, 8a and e.
• Supplementary file 4. Trials, data, analysis, strains, genotypes, RNAi conditions, and statistics for RT-qPCR expression analysis presented in Figures 2a, e and 6l. • Supplementary file 5. Transcription factors differentially expressed in old vs young animals in the same cell clusters where hpk-1 has significant expression changes with age in the Aging Atlas dataset. This includes a summary of which clusters exhibit upregulation and which exhibit downregulation for a transcription factor (TF), and if the gene has been previously shown to have aging or lifespan-associated phenotypes in RNAi or loss-of-function mutant experiments. Primary data was from the C. elegans Aging Atlas, and association to aging phenotypes was determined based on curated annotations from WormBase and GenAge. File is associated with Figure 2, • Supplementary file 6. Kinases differentially expressed in old vs young animals out of any of the 200 clusters with DE analysis results from the Aging Atlas dataset. Out of 438 kinases identified in C. elegans, 186 have significant age-associated expression changes in at least one Aging Atlas cell cluster. For these kinases, we summarized the number of age-associated up-and downregulated changes, if they have been previously identified as having aging or lifespan phenotypes in loss-offunction experiments, and the fold-changes in hpk-1 null vs WT animals for significant changes in our bulk RNA-Seq experiment. Primary data was from the C. elegans Aging Atlas, and association to aging phenotypes was determined based on curated annotations from WormBase and GenAge. File is associated with Figure 2 • Supplementary file 7. Experimental data, strains, genotypes, number of animals/cells, age, and statistical analysis for data presented in Figure 3b-d.
• Supplementary file 8. Experimental data, strains, genotypes, number of animals, hours on aldicarb, and statistical analysis for data presented in Figure 3e.
• Supplementary file 9. Experimental data, strains, genotypes, number of animals, minutes on 5-HT, and statistical analysis for data presented in Figure 3f.
• Supplementary file 10. Differential expression analysis results for RNA-Seq of hpk-1(pk1393) null mutant animals compared to wild-type, presented in this manuscript. Sheet 1: Differential expression for all 16,828 genes considered to be expressed in any sample in the experiment. Sheet 2: The subset of genes with significant expression changes between groups (adjusted p-value <0.05, log2 fold-change magnitude ≥1). Additional columns in both sheets include a summary of the raw and normalized expression for each group, and the human homologs from OrthoList, where applicable. File is associated with Figure 4.
• Supplementary file 11. Of the genes significantly differentially expressed in hpk-1(pk1393) null mutant animals vs wild-type, we found 280 overlap a set of genes with important functions in neurons (Hobert, 2013). For these genes, table includes a summary of their function, the differential expression fold-change and raw expression level in the hpk-1(pk1393) vs wild-type RNA-Seq comparison, and the clusters from the Aging Atlas dataset with significant age-associated expression changes (where applicable). File is associated with Figure 4.
• Supplementary file 13. Trials, data, analysis, strains, genotypes, number of animals, age, and statistical analysis of data for paralysis onset in animals expressing polyQ::YFP presented in Figure 5h, j, Figure 7b • Supplementary file 14. Trials, fluorescent quantification, analysis, strains, genotypes, number of animals, heat shock conditions, and statistical analysis of data presented in Figure 6c and e.

Data availability
Our primary RNA-Seq data is available for review (GEO accession GSE220744: https://www.ncbi.nlm. nih.gov/geo/query/acc.cgi?acc=GSE220744). Data analysis scripts have been deposited at: https:// github.com/samuelsonlab-urmc/hpk1_manuscript_2023 (copy archived at Cornwell, 2023). All data generated or analyzed during this study are included in the manuscript and supporting files; Source data files have been provided for all figures and figure supplements in 14 supplementary files.
The following dataset was generated: Appendix 1 While single-cell RNA-Seq is a powerful approach already, well-documented challenges remain with respect to consistency in cell captures and differentiating biological noise from technical artifacts such as 'drop-outs' from low numbers of average reads per cell (Hwang et al., 2018;Hwang et al., 2021;Ke et al., 2022;Nayak and Hasija, 2021). We attempted to identify gene-gene correlations within cell clusters across individual ages based on the C. elegans Aging Atlas dataset and found that inconsistency in the number of cells per cluster across time points limits our ability to identify many robust and significant associations -such variability is likely technical in nature, as it is not globally consistent across cell clusters for a given time point. In most cases, when aggregated into 'young' and 'old' sample groups, as was done for differential expression analysis in the original Aging Atlas analysis (Roux et al., 2022) the groups represent comparable numbers of total cells for most clusters. The trends we have observed at the level of individual time points indicate that the ageassociated upregulation of hpk-1 does not always occur at the same time point across cell clusters and is not necessarily tightly linked to the upregulation of longevity-associated TFs in the same cluster (Appendix 1-figure 1). Thus, the age-associated upregulation of pro-longevity factors in many cell clusters may not be due to a highly coordinated regulatory process, but rather a distributed response to accumulating local stress or damage. More work will be necessary to resolve if such latelife expression changes indeed contribute to maintenance of wild-type lifespan and healthspan, and the mechanism of the changes thereof (e.g., increased transcription, downregulation of miRNA targeting, epigenetic changes) and the factors that mediate these changes.  Appendix 1-figure 1. hpk-1 and longevity-associated transcription factors are upregulated in overlapping cell clusters, but timing of the age-associated changes varies between cell clusters. Per-age expression summary for cell clusters with 4 or more differentially expressed longevity-associated transcription factors, of the 48 cell clusters in which hpk-1 is upregulated in old vs young animals. Each sub-plot shows log-scale normalized scRNA-Seq expression (mean and SEM) for the indicated cluster. Only genes which were significantly differentially expressed in the cluster are shown (old vs young animals, up-or downregulated). Primary data for this analysis was generated in the C. elegans Aging Atlas (Roux et al., 2022).