Allosteric regulation of C. elegans AMP-activated protein kinase.

(A) AMP-binding sites within AAKG-2, AAKG-4, and AAKG-5 modelled using PyMOL. AAKG-1-3 show high conservation of predicted key AMP-binding residues with human PRKAG1, and AAKG-2 is shown as an example. AAKG-4 and AAKG-5 do not show conservation of AMP-binding residues with PRKAG1. For each isoform, the predicted binding residues are shown as sticks: residues in green are conserved at the functional level with PRKAG1, residues in red are not. AMP molecules are shown in magenta. A multiple sequence alignment is shown below: deeper colouring indicates higher conservation between sequences, and key predicted AMP binding residues are outlined in red. (B) C. elegans has five AMPKγ isoforms which can be grouped by the presence or absence of conserved residues predicted to influence its ability to bind AMP (as in (A) ). Mutant alleles of each γ isoform were combined genetically to create strains that we predict act AMP-dependently or AMP-independently. (C) Phosphorylation of the AAK-2 protein (α subunit) at Thr172. Western blot analysis with and without 4.5mM phenformin. Graph shows quantification from three biological replicates, representative blot shown below the graph. Data shown is normalised to untreated WT/N2 phosphorylation. (D) Effect of aakg mutation on WT lifespan. Representative plot of three biological replicates shown. n =>100 for each group. (E) Effect of aakg mutation on fecundity. Data pooled from three biological replicates. n =>27 for each group. (F) Effect of aakg mutation on the lifespan of animals with reduced insulin signalling ( daf-2 mutants). Representative plot of four biological replicates shown. n =>100 for each group. In C and E , *= p <0.05, **= p <0.01, ***= p <0.001, using Student’s two-tailed t -test. Error bars indicate SEM. For D and F, log-rank test was used – see text for p values.

Description

AMP-activated protein kinase (AMPK) is well-conserved throughout eukaryotes, and critically important for controlling energy balance in a wide variety of organisms (Hardie 2015; Carling 2017; Viollet and Foretz 2016). C. elegans has been especially useful for understanding the role of this kinase in a whole organism context, including its importance for lifespan and metabolism (Zhang et al. 2020; Viollet and Foretz 2016). AMPK is a heterotrimeric protein consisting of an α catalytic, β linker, and γ regulatory subunit. Active AMPK is characterised by phosphorylation of Thr172 on the α subunit, and in C. elegans research has focused on this subunit, using a mutant allele of the gene ( aak-2 ) that removes it entirely. AMPK is regulated allosterically by the nucleotide AMP binding within the γ subunit, initiating conformational change across the kinase which both promotes phosphorylation and inhibits dephosphorylation of the α catalytic subunit (Lee et al. 2008; Hardie 2015; Carling 2017). C. elegans encodes five γ subunits (denoted aakg-1-5) which are 30-47% similar to mammalian PRKAG1 at the DNA sequence level.

The primary amino acid sequence of C. elegans AAKG revealed isoform-specific differences at residues required for nucleotide binding (Tullet et al. 2014). AMP-sensing relies on interaction between positively-charged amine groups and negatively-charged phosphate groups on AMP molecules (Figure 1A). Histidine and arginine residues involved in AMP binding in AAKG-1-3 are highly conserved in human PRKAG1 whereas in AAKG-4 and AAKG-5 these residues are not conserved (Figure 1A). This suggested to us that C. elegans encoded two isoform families of AAKG, with different sensitivities to AMP (Figure 1B).

To determine the role of these two putative AAKG isoform families in the regulation and function of AMPK, we obtained and generated a library of individual aakg mutants, using classical genetics to create aakg-1; aakg-2; aakg-3 triple mutants and aakg-4; aakg-5 double mutants. All strains were homozygous viable and appeared healthy. To investigate the impact of combining these aakg mutations on AMPK activation, we compared phospho-activation of the α subunit in WT C. elegans vs . aakg-1; aakg-2; aakg-3 and aakg-4; aakg-5 mutants . As expected, WT animals showed moderate levels of AAK-2 Thr172 phosphorylation which was moderately increased in response to the known AMPK activator phenformin (Figure 1C) (Cabreiro et al. 2013; Weir et al. 2017). However, Thr172 phosphorylation was almost undetectable in aakg-1; aakg-2; aakg-3 triple mutants, indicating that this family of aakg isoforms is required to induce phosphorylation of the α subunit. In contrast, the α subunit can be phosphorylated in aakg-4; aakg-5 double mutants. In summary, aakg-1, aakg-2, and aakg-3 are required for C. elegans AMPK to respond effectively to AMP.

aak-2 is required for normal lifespan (Miller et al. 2020). Individual aakg isoforms have also been implicated in lifespan, indicating that both catalytic and allosteric regulation of AMPK are important for longevity (Tullet et al. 2014). To test the role of AMP-binding in longevity we measured the lifespan of our aakg-1; aakg-2; aakg-3 and aakg-4; aakg-5 mutants compared to WT. We found that both aakg-1; aakg-2; aakg-3 and aakg-4; aakg-5 mutants had shorter lifespans than WT (Figure 1D, p <0.001 and p <0.01, respectively), but that the effect was most pronounced in the aakg-1; aakg-2; aakg-3 mutant. Notably, the lifespan of aakg-1; aakg-2; aakg-3 triple mutants is similar to that reported for C. elegans lacking aak-2 (Figure 1D) , implying that disruption of aakg-1, -2, and- 3 (and thereby AMP-sensitive regulation) effectively mimics the removal of the α subunit, at least in terms of WT lifespan (Apfeld et al. 2004; McQuary et al. 2016). As our aakg mutants were shortening lifespan, we were concerned that these mutations were making the animals sick. To assess their general health, we measured their progeny production compared to WT. However, despite its dramatic impact on lifespan, combined mutation of aakg-1, aakg-2, and aakg-3 did not alter C. elegans brood size. In contrast, aakg-4; aakg-5 double mutants displayed slightly reduced WT fecundity (Figure 1E, p <0.05). This suggests that the longevity effects are not correlated to overall health.

aak-2 is also required for the long life incurred by genetic downregulation of the insulin signalling pathway, which is achieved in C. elegans through mutation of the insulin receptor daf-2 ; and the aak- 2 mutation completely supresses daf-2 longevity (Apfeld et al. 2004; Kenyon 2010; Viollet and Foretz 2016). To test the impact of AMP-binding on daf-2 longevity we generated daf-2; aakg-1; aakg-2; aakg-3 and daf-2; aakg-4; aakg-5 mutants and measured their lifespan compared to daf-2 mutation alone. Removing either family of aakg isoforms significantly suppressed daf-2 longevity (Figure 1F, p <0.001 for both comparisons), but similarly to the situation in WT animals, the effect was most pronounced when aakg-1, aakg-2, and aakg-3 were mutated. This suggests that the conserved AMP-binding region of these isoforms is important for AMPK’s contribution to daf-2 longevity as well as that of WT. However, in this case, the suppression of daf-2 longevity is not as complete as it is with aak-2 mutation (Apfeld et al. 2004), indicating that both families of aakg isoforms have a role to play in this context. Again, we examined the health of these strains by quantifying progeny production. In contrast to the situation in WT animals, daf-2; aakg-1; aakg-2; aakg-3 mutants exhibited slightly lower fecundity compared to daf-2 controls (Figure 1E, p <0.01), whereas brood size was not altered in daf-2; aakg-4; aakg-5 mutants (Figure 1E). This suggests that longevity and reproductive capability are not correlated in this context.

AMPK is a highly-sensitive kinase which exhibits multiple levels of regulation (Hawley et al. 1996; Oakhill et al. 2010; Xiao et al. 2013). C. elegans differs from mammals in that it appears to encode forms of the kinase with potential for nucleotide-independent regulation. Our data shows that deletion of three AMPKγ subunits predicted to bind AMP ( aakg-1, aakg-2, and aakg-3 ) prevents phosphorylation of Thr172, an important event for AMPK activation in C. elegans , and this reduces WT lifespan similarly to deletion of the α subunit. We also show that removal of aakg-1-3 does not completely suppress daf-2 longevity; mutation of aakg-4 and aakg-5 also contributes here and is sufficient to alter lifespan and WT progeny production, indicating that aakg-4 and aakg-5 are not simply gene duplications/redundant. Mammalian AMPKγ isoforms exhibit tissue-specific expression and function (Salt et al. 1998; Thornton, Snowden, and Carling 1998; Cheung et al. 2000). C. elegans AMPK isoforms also exhibit tissue-specific expression ((Tullet et al. 2014), Wormbase, Version WS282). We do not know the combinations of the subunits that come together in different tissues to determine C. elegans lifespan and physiology, but it is conceivable that the five aakg isoforms could mediate AMP-independent function in specific tissues. Identifying the AMPK trimers that exist in different tissues would be informative, and provides a lucrative avenue for further study.

Methods

Animals grown on standard Nematode Growth Media (NGM). E. coli strains OP50 or OP50-1 were used as a food source (see individual protocols below). Strains were maintained at 20°C, apart from the temperature-sensitive daf-2 ( m577 ) strains, maintained at 15°C. All aakg mutations were confirmed using multiple worm PCR.

PyMOL modelling

Protein sequence files for each AAKG isoform were obtained from Wormbase (Version WS282). FASTA sequences were run through the Phyre2 modelling engine, and PDB files viewed in PyMOL (Version 1.7.4.5). Locations of key predicted nucleotide binding residues from Tullet et al . (2014) were identified.

Multiple Sequence Alignment

C. elegans protein sequences for each AAKG isoform were obtained from Wormbase (Version WS282). The PRKAG1 sequence was obtained from UniProt (UniProt Consortium 2021). Sequences were aligned using Clustal Omega (Version 1.2.2). The alignment was viewed in Jalview (Version 2.11.1.7) and coloured by percentage identity, above a threshold of 62%.

Protein extraction and Western blotting

Animals were synchronised by bleach preparation and grown on NGM plates seeded with OP50-1. L4 stage animals were washed off plates in M9 buffer, washed twice to reduce bacterial contamination, and pipetted onto either control plates or plates containing 4.5mM phenformin for 48 hours at 20°C. Western blotting was carried out as previously described (Tullet et al. 2014). In brief: Samples were collected using a Diagenode Bioruptor. Proteins were probed using rabbit phospho-AMPKα (Thr172) antibody (Cell Signaling) and mouse β-actin antibody (Santa Cruz) as a loading control. Blots were developed using Bio-Rad Clarity™ Western ECL substrate and GE Healthcare Amersham Superfilm, then imaged using an Optimax 2010 X-ray film processor. Film images were captured using 16MP smartphone camera, then analysed in LI-COR Image Studio Lite (Ver 5.0).

Lifespan assay

C. elegans were aged on standard NGM plates, seeded with OP50-1 bacteria treated with 10μM 5-fluro-deoxyuridine. All experiments carried out at 25 o C. Animals were scored every other day and assumed dead when there was no response after lightly touching the body with a platinum pick. Animals that died from causes other than ageing were censored.

Brood size assay

10 L4 stage animals were picked individually onto NGM plates seeded with OP50-1 bacteria and maintained at 25°C for the duration of the experiment. Each day, parent animals were moved to new plates, and the plate containing progeny incubated at 25°C for one day before counting.

Strain name	Genotype	Source/Reference
N2	Hermaphrodites isolated from the CGC Male strain.	CGC
DR1567	daf-2(m577)	CGC
PHX409	aakg-1(syb409)	Sunybiotech (CRISPR)
PHX410	aakg-2(syb410)	Sunybiotech (CRISPR)
PHX411	aakg-3(syb411)	Sunybiotech (CRISPR)
JMT61	aakg-1(syb409); aakg-2(syb410); aakg-3(syb411)	This Study
JMT48	aakg-4(tm5539); aakg-5(tm4052)	This Study
JMT74	daf-2(m577); aakg-1(syb409); aakg-2(syb410); aakg-3(syb411)	This Study
JMT43	daf-2(m577); aakg-4(tm5539); aakg-5(tm4052)	This Study