1. Ecology
  2. Evolutionary Biology
Download icon

Proteomics reveals synergy between biomass degrading enzymes and inorganic Fenton chemistry in leaf-cutting ant colonies

  1. Morten Schiøtt  Is a corresponding author
  2. Jacobus J Boomsma
  1. Centre for Social Evolution, Department of Biology, University of Copenhagen, Universitetsparken, Denmark
Research Article
  • Cited 0
  • Views 574
  • Annotations
Cite this article as: eLife 2021;10:e61816 doi: 10.7554/eLife.61816

Abstract

The symbiotic partnership between leaf-cutting ants and fungal cultivars processes plant biomass via ant fecal fluid mixed with chewed plant substrate before fungal degradation. Here we present a full proteome of the fecal fluid of Acromyrmex leaf-cutting ants, showing that most proteins function as biomass degrading enzymes and that ca. 85% are produced by the fungus and ingested, but not digested, by the ants. Hydrogen peroxide producing oxidoreductases were remarkably common in the proteome, inspiring us to test a scenario in which hydrogen peroxide reacts with iron to form reactive oxygen radicals after which oxidized iron is reduced by other fecal-fluid enzymes. Our biochemical assays confirmed that these so-called Fenton reactions do indeed take place in special substrate pellets, presumably to degrade plant cell wall polymers. This implies that the symbiotic partnership manages a combination of oxidative and enzymatic biomass degradation, an achievement that surpasses current human bioconversion technology.

eLife digest

Colonies of tropical leaf-cutting ants live in underground nests where a fungus grows that feeds them. The ants, in turn, provide the fungus with the freshly-cut leaf fragments it needs for nutrition. The relationship between the ants and the fungus, in which they live close together and help one another survive, is known as symbiosis. It is an ancient, extremely well integrated relationship, in which neither species can survive without the other. However, the details of how the ants and the fungus work together to break down the leaf fragments so they can be used for nutrition are not well understood.

When the ants eat the fungus, they do not digest its enzymes (the proteins that accelerate chemical reactions in a cell). Instead, the fungal enzymes travel through the ants’ gut and into their fecal liquid, which gets deposited on the fresh-cut leaves when the ants collect them. The ants then make temporary pellets out of the new leaf fragments before providing them to the fungus. To better understand how each species contributes to the breakdown of the leaf fragments, Schiøtt and Boomsma identified all the proteins present in the fecal fluid of the ants. Once they had a complete list of about 100 proteins, they determined which of them were produced by the fungus and which by the ant. Schiøtt and Boomsma observed that certain combinations of fungal and ant enzymes could trigger a Fenton reaction – a chemical reaction that efficiently begins the breakdown of the tough walls around plant cells. This reaction is so aggressive that it is rarely found in nature, but it could help explain the high efficiency of the fungus and the ants symbiotically processing leaf fragments.

But could a Fenton reaction actually proceed in the ants’ nest without hurting the ants or affecting the rest of the fungal garden? The evidence obtained suggested that the temporary pellets made by the ants serve to isolate the reaction, so the aggressive chemistry takes place away from the ants and detached from the fungal gardens.

Schiøtt and Boomsma showed that the symbiotic relationship between the ants and the fungus has led to a sustainable and efficient way of breaking down plant materials to use them for nutrition. The Fenton reaction is economically important in many industries, including bioethanol production, the detergent industry, and food production. Emulating the methods used by leaf-cutting ants, which have been fine-tuned by millions of years of natural selection, may allow humans to develop more efficient technologies for breaking down organic compounds.

Introduction

Mutualistic mergers of simpler biological entities into organizationally complex symbioses have been key steps in the evolution of advanced synergistic forms of life, but understanding the origins and secondary elaborations of such natural cooperative systems remains a major challenge for evolutionary biology (Smith and Szathmáry, 1997; Bourke, 2011). Physiological complementarity may be a main driver to maintain symbiotic associations, but new adaptations are also expected to jointly exploit the full potential of symbiosis (Herre et al., 1999; Foster and Wenseleers, 2006). The leaf-cutting ants belong to the genera Acromyrmex and Atta and form the phylogenetic crown group of the attine fungus-growing ants, which all live in mutualistic symbiosis with fungus-garden symbionts (Mueller et al., 2005). While the basal branches of the attine clade farm a considerable diversity of mostly leucocoprinaceous fungi, normally in underground chambers (Branstetter et al., 2017), the leaf-cutting ants almost exclusively farm what has been described as the highly variable cultivar species Leucocoprinus gongylophorus (Schultz and Brady, 2008). The ants cut leaf fragments from the canopy and understory and bring them back to their nest to be processed as growth substrate for the cultivar. This involves chewing forage material into smaller pieces, which are mixed with fecal fluid and deposited as new substrate on the actively growing parts of the garden (Weber, 1966).

The deep ancestors of the attine ants and their fungal cultivars were hunter-gatherers and saprotrophs (Schultz and Brady, 2008), respectively, which implies that none of them were adapted for efficiently degrading fresh plant material. Neither was there any need initially, because all phylogenetically basal attine ants provision their gardens with dead plant debris. There is a striking difference between the small colonies of these basal attine ants and the large colonies of leaf-cutting ants which are conspicuous functional herbivores with a large ecological footprint (Mehdiabadi and Schultz, 2010; Mueller, 2002; Shik et al., 2021). However, most of the leaves that the ants harvest are heavily protected against herbivory by recalcitrant cell walls and toxic substances (Chen, 2008), defenses that needed to be overcome for obtaining net nutrition. In recent decades, a substantial amount of research has elucidated the selection forces that shaped the functional herbivory niche of the leaf-cutting ants. It was shown that: 1. the ‘higher’ attine ants to which the leaf-cutting ants belong rear specialized, gongylidia-producing cultivars that lost their free-living relatives and could thus begin to co-evolve with the ants (Schultz and Brady, 2008); 2. the leaf-cutting ant cultivar Leucocoprinus gongylophorus evolved to be obligatorily multinucleate and polyploid, which may well have allowed higher crop productivity (Kooij et al., 2015a); 3. leaf-cutting ant queens evolved to be multiply inseminated, in contrast to all phylogenetically more basal attines studied; this meant that not only the garden symbiont but also the farming ant families became genetically ‘chimeric’, albeit not as cell-lines but as lines of fungal nuclei and individual patrilines of ants (Villesen et al., 2002); 4. the leaf-cutting ants evolved distinct division of labor within the worker caste, with small, medium, and large workers and specialized soldiers (the latter only in Atta) (Weber, 1966); and 5. the crown group of the attine ants underwent significant changes in their gut microbiota that likely facilitated the conversion efficiency of fungal diet substrate into ant biomass (Sapountzis et al., 2015; Sapountzis et al., 2019).

While these advances have significantly enriched our understanding of the spectacular biology of the attine ants in general and the leaf-cutting ants in particular, very few studies have addressed the molecular synergy mechanisms that were decisive for integrating the complementary innovations of the ants and their cultivars. Recent genome sequencing showed that cultivars likely changed their chitinase processing abilities in co-evolution with the ants’ labial glands (Nygaard et al., 2016), but the clearest and most detailed evidence of functional physiological complementarity has been found for the digestive enzymes of the mutualistic farming symbiosis. These insights were initiated by pioneering work by Michael M. Martin (Boyd and Martin, 1975b; Boyd and Martin, 1975a; Martin and Martin, 1971; Martin and Martin, 1971), which suggested that proteases in the fecal fluid of Atta leaf-cutting ants had actually been produced by the fungal symbiont and passed unharmed through the ant guts to mediate protein degradation of the plant forage material when it was mixed with fecal fluid. Our previous work not only confirmed this hypothesis for Acromyrmex leaf-cutting ants (Kooij et al., 2014b), but also expanded the list of cultivar-derived fecal fluid proteins to include pectinases, hemicellulases, and laccases (Rønhede et al., 2004; Schiøtt et al., 2010; De Fine Licht et al., 2013; Kooij et al., 2016). These studies also showed that all these enzymes are produced in the specialized gongylidia of the fungal cultivar that the ants ingest as their almost exclusive food source (Schiøtt et al., 2010; De Fine Licht et al., 2013; Kooij et al., 2014b; Kooij et al., 2016).

While these recent studies generalized Martin’s findings across the two leaf-cutting ant genera and for multiple groups of enzymes, they left a number of deeper questions regarding the symbiotic decomposition process unanswered. First, the degradation of leaf fragments in Acromyrmex fungus gardens proceeds very efficiently, as ant-produced pellets of chewed leaf material deposited on top of the garden turn black within a few hours to then be spread out over the garden by the ants. Second, the fungal cultivar has lost a major ligninase domain (Nygaard et al., 2016), raising the question how these recalcitrant (difficult-to-degrade) cell wall components are breached to allow fungal hyphae access to the nutritious interior of live plant cells. To address what processes we might have missed, the present study focused on obtaining a full proteome of the fecal fluid of the leaf-cutting ant Acromyrmex echinatior, extending the earlier partial proteome from 33 to 101 identified proteins (Schiøtt et al., 2010). This revealed a surprisingly high proportion of oxidoreductases, which made us conjecture that the mutualistic ant and fungal partners might jointly realize aggressive biomass conversion via the inorganic Fenton reaction, which produces hydroxyl radicals that can break down lignocellulose. We thus set up a series of experiments to investigate whether Fenton chemistry indeed takes place when fecal fluid interacts with chewed leaf fragments concentrated in the ca. 3 mm diameter pellets that the ants construct. We confirmed our basic hypothesis and also found other enzymatic and iron components of the fecal fluid participating in the recycling of Fenton chemistry reactants. The synergistic process of enzymatic and inorganic chemistry appears to sustain a continuous production of hydroxyl radicals until the ants terminate the process. They do that by dismantling their Fenton bioconversion pellets to offer the pre-digested plant substrate to the fungal hyphae as small fragments and seemingly without any collateral damage to ant or fungal tissues due to uncontrolled hydroxyl radicals.

Results

Characterization of the fecal fluid proteome

Our proteomic analysis of worker fecal fluid from four different A. echinatior colonies identified 174 proteins (Supplementary file 1), which showed substantial overlap among the pooled colony-specific samples (Figure 1A). To avoid including spurious proteins, we restricted our analyses to a shortlist of 101 proteins present in at least three of the four colony samples, and we were able to annotate 91 (90%) of these proteins (Table 1). This core proteome consisted mainly of degradation enzymes belonging to five major functional categories: oxidoreductases, proteolytic enzymes, carbohydrate active enzymes (CAZymes), phosphate-liberating enzymes, and lipid degrading enzymes (lipidolytic enzymes) (Figure 1B). Of the 10 remaining proteins, only three could not be functionally annotated. The vast majority (86) of the core fecal proteome originated from the fungal symbiont, with only 15 being encoded by the ant genome. Based on the number of spectra obtained for each identified protein, MaxQuant produced an approximate measure of relative abundance of each protein called label-free quantification intensity (LFQ intensity) (Table 1). This revealed that lipidolytic enzymes constitute only a very small fraction and that the proteins in the ‘miscellaneous’ category collectively had a moderate representation although some of these proteins had high abundance. The four remaining categories had summed LFQ intensities of the same order of magnitude (Figure 1B) and two specific proteins stood out as being exceptionally abundant (Supplementary file 1), the Alkaline phosphatase (Protein ID 75 in Table 1) and the previously described Laccase LgLcc1 (Protein ID 97 in Table 1; De Fine Licht et al., 2013), which jointly made up almost a quarter of the total fecal protein mass. Three other proteins also showed remarkably high LFQ-abundance levels: a 1,3-beta-glucanase (Protein ID 88 in Table 1), a Metallopeptidase M35 (Protein ID 206 in Table 1), and a Serine protease S53 (Protein ID 64 in Table 1), which implies that the five most abundant proteins contributed more than 40% of the total fecal protein mass. These results confirmed our previous findings that a large proportion of the fecal fluid proteins consists of proteases (Kooij et al., 2014b), pectinases (Schiøtt et al., 2010), hemicellulases (Kooij et al., 2016), and laccases (De Fine Licht et al., 2013), but we now also identified several phosphate-liberating and lipidolytic enzymes.

Figure 1
Download asset Open asset
Statistics of the fecal fluid proteome of Acromyrmex echinatior.

(A) Venn diagrams, constructed using the web application Venny 2.1 (http://bioinfogp.cnb.csic.es/tools/venny/index.html) showing the overlap of protein profiles identified in the four fecal fluid samples obtained from colonies Ae263, Ae322, Ae356, and Ae372. (B) Pie chart showing the abundances of proteins across colonies assigned to six functional categories based on the label-free quantification (LFQ) values provided by MaxQuant.

Table 1
Fecal fluid proteins that were found in three or four of the examined colony samples (Figure 1), and thus likely to belong to the core fecal fluid proteome.

For each protein the predicted function, its fungal or ant origin, the number of samples containing the protein, the relative abundance of the protein, and an identifier number are listed. Proteins were assigned to one of six functional categories (Figure 1). Although a number of oxidoreductases are listed in the CAZy database in the subcategory of auxiliary activities we have for the present study kept them separate. This is because it remains ambiguous whether they are true CAZymes if they are defined as redox enzymes that ‘act in conjunction with CAZymes’. This separation also resolved the problem that some of the oxidoreductases listed in Table 1 are not covered by the CAZy database. Question marks indicate that annotations were inconclusive.

CategoryProtein IDPredicted functionSourceSamples with proteinRelative abundance
Oxidoreductases
(including CAZymes with auxilliary activity)		15LaccaseFungus4106.0
97LaccaseFungus4431.8
180LaccaseFungus414.7
184LaccaseFungus421.4
354-carboxymuconolactone decarboxylase, alpha-beta hydrolaseFungus44.3
81Copper radical oxidase, glyoxal oxidase, galactose oxidaseFungus48.9
186Copper radical oxidase, glyoxal oxidase, galactose oxidaseFungus48.3
98FAD/FMN-containing isoamyl alcohol oxidaseFungus441.1
9GMC oxidoreductase, aryl-alcohol oxidaseFungus45.9
86GMC oxidoreductase, aryl-alcohol oxidaseFungus410.3
178GMC oxidoreductase, aryl-alcohol oxidaseFungus46.9
191GMC oxidoreductase, aryl-alcohol oxidaseFungus418.3
23GMC oxidoreductase, aryl-alcohol oxidaseFungus31.9
145GMC oxidoreductase, glucose dehydrogenaseAnt482.2
149Succinate dehydrogenase (ubiquinone) flavoprotein subunitAnt34.8
Proteolytic enzymes6Aspartic peptidase A1A, polyporopepsinFungus412.0
111Aspartic peptidase A1A, saccharopepsinFungus474.3
125Metallopeptidase M1, ERAP2 aminopeptidaseAnt411.3
112Metallopeptidase M14A, zinc carboxypeptidaseAnt427.4
144Metallopeptidase M14A, zinc carboxypeptidaseAnt47.3
141Metallopeptidase M28D, carboxypeptidase QAnt415.9
188Metallopeptidase M35, deuterolysinFungus423.9
206Metallopeptidase M35, peptidyl-lys metallopeptidaseFungus4218.6
207Metallopeptidase M35, peptidyl-lys metallopeptidaseFungus411.4
56Metallopeptidase M36Fungus44.3
133Serine protease S1A, chymotrypsinAnt456.0
134Serine protease S1A, chymotrypsinAnt43.8
135Serine peptidase S1A, chymotrypsin-2Ant48.8
36Serine peptidase S8A, cuticle degrading peptidaseFungus47.2
103Serine peptidase S8A, cerevisinFungus451.6
16Serine peptidase S10, carboxypeptidaseFungus32.0
14Serine peptidase S10, carboxypeptidaseFungus41.3
24Serine peptidase S10, carboxypeptidase OcpAFungus431.5
95Serine peptidase S10, serine-type carboxypeptidase FFungus411.7
199Serine peptidase S28, carboxypeptidaseFungus46.3
33Serine peptidase S53, grifolisinFungus420.5
64Serine protease S53, grifolisinFungus4192.4
Phosphate-liberating enzymes74Acid phosphatase, 3-phytase AFungus37.2
0Acid phosphatase, nucleotidaseFungus416.9
61Acid phosphatase, nucleotidaseFungus424.8
75Alkaline phosphataseFungus4432.8
100Alkaline phosphataseFungus414.8
109Phytase esterase-likeFungus4105.1
59Phytase esterase-likeFungus319.2
193Phytase, histidine acid phosphatase domainFungus46.5
187Phytase, phosphoglycerate mutase, histidine acid phosphataseFungus425.2
72PLC-like phosphodiesteraseFungus33.5
37Ribonuclease T1Fungus45.6
34Ribonuclease T2Fungus411.5
73GH3, beta-glucosidaseFungus452.0
99GH3, beta-xylosidaseFungus476.0
55GH5, endocellulaseFungus33.7
179GH5, exo-1,3-beta-glucanaseFungus49.8
183GH5, mannan endo-1,4-beta-mannosidase FFungus47.6
13GH5, mannan endo-1,4-beta-mannosidaseFungus311.7
38GH10, endo-1,4-beta-xylanaseFungus33.9
Carbohydrate active enzymes121GH12, xyloglucanaseFungus34.8
132GH13, alpha-glucosidase, maltaseAnt32.9
123GH15, glucoamylaseFungus466.5
88GH17, 1,3-beta-glucanaseFungus4290.8
8GH18, chitinaseFungus316.8
66GH20, betahexosaminidaseFungus31.3
214GH25?, lysozymeFungus4102.4
25GH27, alpha-galactosidaseFungus427.9
83GH28, endo-polygalacturonaseFungus483.3
80GH29, alpha-L-fucosidaseFungus31.8
104GH31, alpha-glucosidaseFungus439.1
157GH31, alpha-glucosidaseAnt47.9
108GH35, beta-galactosidaseFungus413.6
154GH37, trehalaseAnt419.8
90GH43, arabinan-endo-1,5-alpha-L-arabinosidaseFungus421.8
11GH51, arabinofuranosidaseFungus485.7
46GH53, arabinogalactanaseFungus436.2
77GH55, exo-1,3-beta-glucanaseFungus434.5
7GH78, α-L-rhamnosidaseFungus414.0
63GH79, betaglucoronidaseFungus435.0
62GH88, glucuronyl hydrolaseFungus43.0
76GH92, exo-alpha-mannosidase?Fungus416.6
68GH92, exo-alpha-mannosidase?Fungus33.8
57CE8, pectinesteraseFungus469.0
40CE12, rhamnogalacturonan acetylesteraseFungus45.8
67CE12, Rhamnogalacturonan acetylesteraseFungus427.6
208PL1, pectate lyaseFungus434.6
122PL4, rhamnogalacturonan lyaseFungus49.6
Lipidolytic enzymes124Lipase, triacylglycerol lipaseFungus42.5
65Neutral/alkaline nonlysosomal ceramidaseFungus46.2
138Pancreatic lipase-related proteinAnt46.5
44Phosphatidylglycerol/phosphatidylinositol transfer proteinFungus32.8
32Alpha/beta hydrolase, cephalosporin esteraseFungus410.1
18Alpha/beta hydrolase, triacylglycerol lipase, carotenoid ester lipaseFungus45.5
27Carboxylesterase; alpha-beta hydrolase; lipaseFungus41.5
Miscellaneous proteins212Fruit-body specific protein DFungus444.0
31Plant expansin, papain inhibitorFungus31.7
129RegucalcinAnt45.8
113SnodProt1, cerato plataninFungus4114.7
10UbiquitinFungus41.7
136Epididymal secretory protein E1Ant46.4
158CEN-like protein 2, OV-16 antigenFungus44.8
106Hypothetical proteinFungus439.2
195Hypothetical proteinFungus34.5
119Hypothetical protein, symbiosis related protein, MAPK?Fungus452.8

The complementary components needed for Fenton chemistry

Five of the identified oxidoreductases were annotated as fungal glucose-methanol-choline (GMC) oxidoreductases. A subgroup of these enzymes are the aryl alcohol oxidases that are believed to play an important role in lignin degradation by brown-rot fungi (Guillén et al., 2000). This category of wood degrading fungi is distinct from the white-rot fungi, which degrade lignin using ligninolytic peroxidases (Arantes et al., 2012; Arantes and Goodell, 2014). The latter mechanism has most likely evolved only once in the Agaricomycotina lineage (to which Leucocoprinus belongs), but has subsequently been lost many times giving a complex phylogenetic pattern (Floudas et al., 2012). We showed previously (Nygaard et al., 2016) that the leaf-cutting ant cultivar has lost its ligninolytic peroxidases, so their decomposition spectrum would functionally be more reminiscent of brown-rot fungi than white-rot fungi although they are not wood degrading fungi and are thus not formally covered by this terminology. Assuming that convergent adaptation could have taken place in the Leucocoprinus symbiont, we thus pursued the pathways that could enable the Fenton reaction in more detail.

In order to break down lignocellulose, brown-rot fungi produce aromatic alcohols, which are subsequently oxidized by aryl alcohol oxidases to produce hydrogen peroxide (H2O2). The hydrogen peroxide is subsequently used in a Fenton reaction in which reduced iron (Fe2+) reacts with hydrogen peroxide to produce hydroxyl radicals that can diffuse into the lignocellulose substrate and break the chemical bonds responsible for the rigid structure of lignocellulose. This in turn will loosen the cell walls and allow access for more bulky degradation enzymes to target specific components of the plant cell walls. Unless the oxidized iron (Fe3+) is recycled by reduction to Fe2+, the production of reactive oxygen species will end when all Fe2+ has been used. It has therefore been suggested that fungi secrete secondary metabolites such as hydroquinones to reduce Fe3+ to Fe2+ (Suzuki et al., 2006). In spite of the decomposition potential of the Fenton reaction, it remains a major challenge for any living tissues to be exposed to concentrated doses of free radicals. Numerous adaptations have in fact evolved to remove low concentrations of free radicals from living tissues (Matés et al., 1999), so we conjectured that attine ants should have found ways to avoid such damage in case they would use Fenton chemistry.

Against this background knowledge from brown-rot fungi, we drew up a hypothetical scenario (Figure 2) that might allow the leaf-cutting ant symbiotic partnership to use Fenton chemistry, to sustain and control that reaction according to need, and to have it take place in a spatial setting where collateral damage to living tissues would be avoided. As it appears, the spatial issue appears to have been resolved because Acromyrmex colonies concentrate their chewed leaf fragments in pellets of ca. 3 mm diameter at the top of their fungus gardens for some hours before the material is dispersed across the actively growing ridges of the fungal mycelium mass. However, if these pellets would be the optimal location for a sustained Fenton reaction, the symbiotic partnership would have been required to also evolve a fecal fluid enzyme that reduces Fe3+ while decomposing degradation products from the plant substrate. We hypothesized that glucose dehydrogenase (Protein ID 145 in Table 1), the most abundant ant-encoded fecal fluid protein (Table 1, Supplementary file 1), could have this function using glucose as electron donor to reduce Fe3+, either directly or indirectly via an intermediate redox compound (Figure 2). We therefore set up a number of biochemical assays to test the proposed scenario and report the results below.

Figure 2
Download asset Open asset
Hypothesized reaction scheme for the generation of hydroxyl radicals when ant fecal fluid interacts with chewed leaf fragments in temporary substrate pellets, based on the presence of enzymes in the total fecal fluid proteome (Supplementary file 1).

Fungal enzymes produced in gongylidia of the symbiotic garden-cultivar that the ants ingest pass unharmed through the gut to end up in the fecal fluid (Boyd and Martin, 1975b; Kooij et al., 2014b; Schiøtt et al., 2010; De Fine Licht et al., 2013; Kooij et al., 2016). After droplets of fecal fluid are deposited and become exposed to oxygen, the fungal oxidoreductases produce hydrogen peroxide while aryl alcohols are converted to aryl aldehydes. The hydrogen peroxide then reacts with reduced iron (Fe2+) to produce hydroxyl radicals (OH•) in a Fenton reaction, which aggressively breaks down cell walls of the plant substrate. Oxidized iron (Fe3+) can then be reduced again by ant-encoded glucose dehydrogenase, using glucose released via plant cell wall decomposition as electron donor. The leaf substrate is initially concentrated in green pellets of ca. 3 mm diameter distributed across the top of fungus gardens, which turn black in a few hours when subjected to Fenton-mediated degradation (inset image). Compounds ultimately derived from the fungal symbiont are in red text and compounds directly produced by the ants in blue.

Hydrogen peroxide production by fungal GMC oxidoreductases in ant fecal fluid

A phylogenetic tree of all five fungal GMC oxidoreductases isolated from fecal fluid, together with representative GMC oxidoreductases from various basidiomycete fungi, showed that these five enzymes clustered within the aryl alcohol oxidases (Figure 3), while being distinct from the four other known functional groups (glucose oxidases, methanol oxidases, pyranose 2-oxidases, and cellulose dehydrogenases) (Ferreira et al., 2015). We subsequently measured the amount of hydrogen peroxide that these GMC oxidoreductases produce with and without substrates, which showed that fecal fluid contained a substantial amount of hydrogen peroxide. Addition of aromatic veratryl alcohol further increased this amount, whereas glucose and methanol did not have any effect (Figure 4A), confirming the presence of aryl alcohol oxidases in the fecal fluid. As an extra control assay to check that the increase in absorbance was directly caused by hydrogen peroxide, we added the enzyme catalase, which is a very effective and specific hydrogen peroxide degrading enzyme. As expected, catalase completely removed the signal, confirming that hydrogen peroxide is indeed naturally produced in fecal fluid (Figure 4A). Two other fecal fluid oxidoreductases (Protein IDs 81 and 186 in Table 1) were annotated as copper radical oxidases with close similarity to glyoxal oxidases. These enzymes may also produce hydrogen peroxide, using the small organic compound glyoxal and other aldehydes as substrate (Daou and Faulds, 2017). Experimental addition of glyoxal to fecal fluid of Acromyrmex worker did indeed increase the amount of hydrogen peroxide, confirming that glyoxal oxidases are also present in the fecal fluid (Figure 4A).

Figure 3
Download asset Open asset
Gene tree of the five fungal glucose-methanol-choline (GMC) oxidoreductases identified in the ant fecal fluid (present study) and in representative other basidomycete fungi, all assigned to functional groups based on a previous study (Ferreira et al., 2015).

All fungal GMC oxidoreductases from the fecal fluid (Protein IDs 9, 23, 86, 178, and 191 in Table 1; red text) clustered among the known aryl-alcohol oxidases. Other closely related functional groups are glucose oxidases, methanol oxidases, pyranose-2 oxidases, and cellulose dehydrogenases, which were retrieved in Phanerochaete chrysosporium (PHACH), Rhodonia placenta (RHOPL), Trametes versicolor (TRAVE), and Phlebia brevispora (PHLBR), but not in fecal fluid of the leaf-cutting ant A. echinatior. Note that the ant-encoded glucose dehydrogenase is also a GMC oxidoreductase, but would sequence-wise not fit into this phylogeny of fungal proteins. Numbers are aLRT SH-like support values for nodes. The scale bar represents 0.5 substitutions per site.

Figure 3—source data 1

Amino acid sequences used for the gene tree shown in Figure 3.

https://cdn.elifesciences.org/articles/61816/elife-61816-fig3-data1-v2.xlsx
Figure 4
Download asset Open asset
Bioassays to demonstrate that inorganic.

Fenton chemistry must be taking place when ant fecal fluid is exposed to oxygen while being deposited on leaf pulp pellets chewed by the ants, testing the key interactions hypothesized in Figure 2. (A) Bar plot showing the concentrations (means ± SE across six colonies) of hydrogen peroxide in fecal fluid after adding potential substrates (glucose, methanol, glyoxal, or veratryl alcohol) of glucose-methanol-choline (GMC) oxidoreductases with or without the hydrogen peroxide degrading enzyme catalase. One-way ANOVA showed a highly significant overall effect of treatments: F6,36 = 14.597, p=2.863e-08, and pairwise post hoc t-tests on matching samples from the same ant colonies (corrected for multiple testing with the Holm–Bonferroni method) confirmed both the enhancing effects of veratryl alcohol and glyoxal and the inhibiting effect of catalase (p-values in plot unless non-significant) (NS). (B) Deoxyribose assays showing that ant fecal fluid has the capacity to produce hydroxyl radicals (means ± SE across six colonies). Phenanthroline is known to work as an iron chelator and significantly reduced degradation of 2-deoxy-D-ribose while the solvent (methanol) of phenanthroline did not. Paired t-tests followed the same protocol as in the A-panel except that the overall ANOVA was omitted because there were only three means to compare. (C) A Ferrozine assay (means ± SE across six colonies) showing the capacity of ant fecal fluid to reduce Fe3+ to Fe2+, confirming that addition of glucose increases the rate of iron reduction. Statistics as in the B-panel.

Figure 4—source data 1

Data used for Figure 4A, B, and C.

https://cdn.elifesciences.org/articles/61816/elife-61816-fig4-data1-v2.xlsx

Does ant fecal fluid contain sufficient iron to maintain Fenton chemistry?

For the Fenton reaction to take place, the produced hydrogen peroxide must react with Fe2+ atoms. Brown-rot fungi are believed to be able to actively increase the concentration of iron atoms in their wood substrate for this purpose (Kirker et al., 2017), but such a mechanism would not be applicable to the fecal fluid Fenton reaction since it is separated from the fungal hyphae. To be viable, this reaction must therefore rely on iron present in the leaf material collected by the ant workers and in the fecal fluid itself. Making fungal iron available would be adaptive for the ant farming symbiosis as a whole because direct use of iron from live plant tissues might well be constrained because plants also produce iron chelator compounds to inhibit the growth of plant pathogens (Dellagi et al., 2005). We measured the iron levels in fecal fluid and found an average (± SD) level of 370 ± 110 µM iron across six ant colonies. This concentration of iron should be sufficient to allow the Fenton reaction to occur (Hegnar et al., 2019; Elias and Waterhouse, 2010). It corresponds to the levels relevant for natural wood degrading systems (Hegnar et al., 2019) although it is one to two orders of magnitude lower than the levels typically used for pretreatment of plant biomass in vitro (Jung et al., 2015). However, in such human bioconversion experiments, oxidized iron is not recycled into reduced iron, which means that the pool of useable iron is continuously decreasing and therefore needs a high starting concentration. In living systems there are limits to how much iron can be accumulated without posing a risk for unintended oxidation of organic compounds. Natural selection producing a system for iron recycling could thus be highly adaptive.

Are fecal fluid hydroxyl radicals produced by Fenton chemistry?

The hydroxyl radicals that Fenton chemistry should produce can be measured via their ability to break down 2-deoxy-D-ribose into thiobarbituric acid-reactive substances (TBARS) (Halliwell et al., 1988). To quantify this effect, we adapted the standard 2-deoxy-D-ribose assay to accommodate the small volumes of fecal fluid that could practically be collected directly from the ants and found pronounced amounts of hydroxyl radicals to be present in the ant fecal fluid (Figure 4B). Adding a strong metal chelator 1,10-phenanthroline caused a significant reduction in absorbance. This is consistent with the hydroxyl radicals being produced by Fenton chemistry because the Fenton reaction is known to be inhibited by 1,10-phenanthroline’s strong iron sequestering activity (Winterbourn, 1995). This effect was not observed when only the solvent of the 1,10-phenanthroline chemical (methanol) was added.

Is iron reduction controlled by the ant encoded glucose hydrogenase?

The Fenton reaction can only proceed as long as there are Fe2+ atoms present to be converted to Fe3+ atoms, and brown-rot fungi are believed to maintain their Fenton reactions via enzyme systems that convert Fe3+ back to Fe2+ (Guillén et al., 2000; Arantes et al., 2012). A similar mechanism would not work for the initial stages of leaf degradation in the fungus garden, as the mycelium is separated from the Fenton pellets. We therefore conjectured that the ant-encoded glucose dehydrogenase identified in the fecal fluid proteome could mediate the reduction of Fe3+ to Fe2+ using glucose as electron donor. We indeed found a measurable capacity for Fe3+ reduction in fecal fluid, which was further increased by the addition of glucose. This was consistent with Fe3+ being actively recycled into Fe2+ via fecal fluid compounds and with glucose dehydrogenase most likely being responsible for at least part of this activity (Figure 4C). An alternative pathway for Fe3+ reduction is mediated by laccase (Guillén et al., 2000), which was also found in high quantities in the fecal fluid. Preliminary experiments using laccase substrates in these assays did, however, not change the rate of Fe3+ reduction. This makes it less likely that laccase plays a role in this process, although more elaborate experimentation will be needed in order to rule out this possibility.

Discussion

Acromyrmex fecal fluid has a highly specialized proteome

The presence of fungal enzymes in the fecal fluid of leaf-cutting ants was suggested already in the 1970s based on the biochemical similarity of proteases in fecal fluid and the fungal cultivar (Boyd and Martin, 1975b; Boyd and Martin, 1975a). We achieved increasing confirmation of this idea in a series of previous studies (Schiøtt et al., 2010; Kooij et al., 2014b; De Fine Licht et al., 2013; Kooij et al., 2016), but our present study presents the first overall proteome that is complete enough to allow inferences of hitherto unknown chemical processes. The comprehensive list of A. echinatior fecal fluid proteins that we obtained (Supplementary file 1) was now also sufficiently replicated to confirm that the large majority of the identified proteins are fungal in origin and serve the farming symbiosis after being deposited on new plant substrate via the ant fecal fluid. When run on SDS-Page gels (Schiøtt et al., 2010), the fecal fluid proteins always gave the same banding pattern, rejecting the null hypothesis that they are a random collection of proteins that happen to escape digestion. Overall, 101 of the 174 identified proteins were found in at least three fecal fluid samples from different ant colonies, and for many of them we were able to measure the corresponding enzymatic activity using biochemical assays (this study and Schiøtt et al., 2010; De Fine Licht et al., 2013; Kooij et al., 2014b; Kooij et al., 2016). These results strongly suggest that most if not all of the fecal proteins have condition-dependent adaptive significance, have been selected to avoid degradation in the digestive system of the ants, and serve the efficiency of the entire symbiosis between farming ants and fungal cultivars.

While many of the enzyme activities found in the ant fecal fluid have been discussed in previous publications, the roles of phosphate-liberating enzymes and lipidolytic enzymes have not. Just like nitrogen and carbon, phosphorus is a major macronutrient required for many biological processes. Enzymatic degradation of organic matter to liberate C, N, and P is known to take place in a stoichiometrically balanced way (Sinsabaugh et al., 2009), so it is not surprising that the fecal fluid also contains a range of enzymes able to release phosphorus from various substrates. That alkaline phosphatase (Protein ID 75 in Table 1) had the highest abundance of all fecal proteins underlines the importance of phosphorus mineralization. Several enzymes able to release phosphorus from phytin were also present. Phytin is a phosphorus-storage compound in plant seeds, but also occurs in pollen and vegetative tissues (Brinch-Pedersen et al., 2002). Laboratory colonies of A. echinatior were occasionally fed with dry rice grains, but in nature leaf-cutting ants are not known to forage on seeds to any significant degree (Kooij et al., 2014a). However, they do harvest flowers, whose pollen may be a source of phytin. Finally, the fecal fluid contained ribonucleases that will degrade RNA, another major source of phosphorus. The fecal lipidolytic enzymes had low abundances overall. Apart from digesting lipids into assimilatory products, these enzymes may also hplay a function in degrading cell membranes to liberate cytosolic compounds as has been shown for biotrophic fungi (Ghannoum, 2000). They may also interfere with signal transduction pathways involved in plant resistance toward pathogens (Wu and Li, 2015), but all these putative functions require that the plant substrate is alive for a while after having been fragmented and placed in the fungus garden, which may well be the case.

Multiple lines of evidence for fecal fluid Fenton chemistry after mixing with fresh leaf substrate

The key compounds hypothesized to be involved in inorganic Fenton chemistry (GMC oxidoreductases, hydrogen peroxide, high concentrations of iron, and glucose dehydrogenase) were always found in all four colonies for which we investigated fecal fluid samples. Compared to white-rot fungi, brown-rot fungi are believed to have a reduced array of enzymes targeting lignocellulose, and to specifically lack lignin peroxidases, which is also the case for L. gongylophorus (Nygaard et al., 2016). Brown-rot fungi compensate this lack of enzymatic potential by using Fenton chemistry to degrade lignocellulose while producing extracellular hydroxyl radicals (Arantes and Goodell, 2014). In the Fenton reaction, hydrogen peroxide reacts with ferrous iron (Fe2+) and hydrogen ions to produce ferric iron (Fe3+), water, and hydroxyl radicals (Guillén et al., 2000; Arantes et al., 2012). The very reactive hydroxyl radicals produced in these free-living fungi are known to damage a wide range of biomolecules (Halliwell et al., 1988). Their putative functions are: 1. to release soluble carbohydrate fragments from plant cell walls (Schweikert et al., 2002) and 2. to reduce the viscosity of solutions containing plant cell wall components such as xyloglucan, pectin, and cellulose (Fry, 1998), as expected when degradation of these polysaccharides has taken place. The small size of hydroxyl radical molecules allows them to initiate penetration and cleavage of lignin, cellulose, and hemicellulose polymers in preparation for further degradation into assimilatory monomers by specific enzymes (Arantes et al., 2012). Brown-rot fungi are believed to use GMC oxidoreductases or copper radical oxidases to aerobically oxidize fungal alcohols or aldehydes in order to produce hydrogen peroxide. In this process they then obtain the required iron from the wood substrate (Arantes et al., 2012) or via actively mycelial transport into the degrading wood (Kirker et al., 2017). It has been suggested that fungal laccases are then oxidizing hydroquinones into the semiquinones that react with ferric iron (Fe3+) to produce quinones and ferrous iron (Fe2+), which would produce the recycling process needed to maintain the Fenton reaction over time (Guillén et al., 2000).

It is intriguing that the details of Fenton chemistry in brown-rot fungi have remained rather little known in spite of the economic importance of these fungi (Arantes and Goodell, 2014), and that the leaf-cutting ants offered an unexpected window for understanding this elusive process. The laccases, GMC oxidoreductases, and copper radical oxidases were similarly found in high quantities in the fecal fluid of Acromyrmex leaf-cutting ants, consistent with Fenton chemistry being a standard procedure in this fungus-farming symbiosis to degrade plant cell walls and/or toxic secondary plant polymers meant to deter herbivores (Figure 2). We could verify that the identified GMC oxidoreductases were aryl alcohol oxidases (Figure 3), which are known to produce hydrogen peroxide from aromatic alcohols. We also verified that fecal fluid contains concentrated hydrogen peroxide and that the concentration increased further after adding veratryl alcohol or glyoxal, which are substrates for GMC oxidoreductases and copper radical oxidases, respectively (Guillén et al., 1992; Daou and Faulds, 2017). We further documented that the iron content of the fecal fluid is maintained at levels that are amply permissive for Fenton chemistry to occur. Also the acid pH of ca. 4 in the fecal fluid should be ideal for Fenton chemistry (Bishop et al., 1968; Erthal et al., 2004) although we cannot be sure this is the ambient pH because the plant material might also influence the final pH in the substrate pellet. Finally, a separate assay showed that fecal fluid components degrade deoxyribose into thiobarbituric acid-active products, which can only be explained by the presence of hydroxyl radicals (Halliwell et al., 1988). To obtain more direct evidence for the produced hydroxyl radicals contributing to plant substrate breakdown, future research should focus on using Fourier transform infrared spectroscopy to analyze plant cell wall preparations subjected to fecal fluid degradation, a type of analytical chemistry that was beyond the scope of the present study. Without such validation, we cannot rule out other possible functions of these radicals such as general sanitation of the plant substrate. However, such alternative function has never been conjectured or shown for the ant fecal fluid, whereas brown-rot fungi are widely believed to use hydroxyl radicals for substrate degradation (Arantes and Goodell, 2014). We also note that specialized functions of the ant fecal fluid aimed at plant substrate degradation are uncontroversial (Boyd and Martin, 1975b; Boyd and Martin, 1975a; Martin and Martin, 1971; Kooij et al., 2014b; Rønhede et al., 2004; Schiøtt et al., 2010; De Fine Licht et al., 2013; Kooij et al., 2016). Our Fenton chemistry scenario is therefore supported by consistent inferential evidence, but in need of further experimental validation.

The high concentrations of laccase LgLcc1 (Protein ID 97 in Table 1) in the ant fecal fluid suggest a similar mechanism for maintaining reliable conversion of Fe3+ back to Fe2+ (using hydroquinone as an electron donor) as the Fenton chemistry known from brown-rot fungi (Guillén et al., 2000). However, the glucose dehydrogenase (Protein ID 145 in Table 1) encoded by the ant genome was also abundant in the fecal proteome and could convert Fe3+ back to Fe2+ via simultaneous oxidization of glucose. We verified this putative function as likely with a Ferrozine-based assay (Goodell et al., 2006), including its dependence on glucose availability. Interestingly, however, preliminary experiments indicated that adding laccase substrates did not influence the velocity of this reaction, which suggests that glucose liberated from the plant polysaccharides is primarily used to keep the Fenton reaction going. This substrate dependence may provide an additional feedback loop for the farming symbiotic partnership to control the inorganic Fenton chemistry process. The ant-derived glucose dehydrogenase (Protein ID 145 in Table 1) alternative might allow the ants to also use the fungal laccases specifically for the decomposition of secondary plant compounds as found in one of our previous studies (De Fine Licht et al., 2013). That study also suggested that LgLcc1 (Protein ID 97 in Table 1) came under positive selection in the common ancestor of the leaf-cutting ants, which would be consistent with laccase becoming redundant for iron reduction and therefore undergoing directional selection for other functions.

Symbiotic complementarity is necessary for making fast Fenton chemistry work

The assembly of complementary components from different sources implies that the fungal cultivar and the farming ants synergistically cooperate to accomplish the continued production of hydroxyl radicals, but only as long as Fenton chemistry is needed. The ants do not appear to keep adding fecal fluid to existing Fenton pellets and they take them apart when they have turned black after some hours to distribute the Fenton exposed leaf fragments over the actively growing ridges of the fungus garden where fungal decomposition takes over. Both the processing efficiency and the ultimate behavioral control by the ants strongly suggest refined co-adaptation at the molecular level to enable colonies to operate as integrated functional herbivores. An earlier study found a high expression of an NADH-quinone oxidoreductase in fungus gardens of A. echinatior (Grell et al., 2013), an enzyme converting quinones to hydroquinones that can react with Fe3+ to form Fe2+ after conversion to semiquinones by laccase. This is consistent with the ant encoded glucose dehydrogenase being relevant only for the degradation process in the Fenton pellets, and with NADH-quinone oxidoreductases taking over after the leaf fragments have been dispersed over the garden and the fungal hyphae decompose what remains of the substrate.

Aryl alcohol oxidases use aromatic alcohols as a substrate to produce hydrogen peroxide. White-rot fungi produce aromatic veratryl alcohol as a secondary metabolite, which plays several important roles for ligninolysis (Jensen et al., 1994). Veratryl alcohol is derived from phenylalanine via the phenylalanine/tyrosine pathway that has previously been shown to be upregulated in the gongylidia of the A. echinatior cultivar L. gongylophorus (De Fine Licht et al., 2014). This might suggest that veratryl alcohol (or a derivative thereof) is produced in the gongylidia of the fungal cultivar to be ingested but not digested by the ants so it can function in the fecal fluid to facilitate the initial stages of substrate degradation. The hydrogen peroxide needed in the Fenton reaction may also be derived from the activity of the glyoxal oxidases that we identified, which use various aldehydes as substrate for hydrogen peroxide production (Daou and Faulds, 2017). Such aldehydes could then either originate from the oxidized aryl alcohols produced by the aryl alcohol oxidases, or could be direct degradation products from plant cell wall components (Daou and Faulds, 2017). These multiple parallel pathways suggest that the co-evolved interactions between inorganic Fenton chemistry and organic enzyme functions are likely to have robust operational functionality under a variety of environmental (i.e. substrate) conditions.

Finally, as we explained in the introduction, it is important to note that hydroxyl radicals are highly toxic also for the organisms producing them. This probably constrains the use of naturally produced hydroxyl radicals for biomass degradation unless the Fenton chemistry producing these radicals remains distal to hyphal growth and is terminated by the time hyphae grow into the substrate patches pre-treated in this manner. This complication will slow down the rate of decomposition in free-living fungi, but the leaf-cutting ant symbiotic partnership appears to have resolved this constraint by compartmentalizing the complementary elements of initial aggressive break down of plant substrate. Although the necessary compounds for the Fenton reaction are all present in the guts of the ants, the low oxygen pressure there would presumably prevent the production of hydrogen peroxide inside the ants and thereby any collateral damage to ant tissues. Defecation onto concentrated pellets of chewed leaf fragments thus ensures that the reaction does not start until atmospheric oxygen is available. Consequently, the Fenton pellets appear to function as small bioconversion reactors that are spatially isolated from both the fungal and the ant tissues so that potentially harmful side-effects are avoided. Also in this respect robustness against potential malfunction appears to have evolved. Our recent genome sequencing study of Acromyrmex ants and their cultivars revealed that A. echinatior has more glutathione S-transferase genes than other (non-attine) ants, despite having fewer detoxification genes in general (Nygaard et al., 2011). Glutathione S-transferase is known to be involved in the breakdown of oxidized lipids, which would likely be produced if reactive oxygen species would be formed in the gut and react with the lipid cell membranes of the gut tissues. This suggests that physiological mechanisms to ameliorate inadvertent damage caused by hydroxyl radicals may be in place, should they appear in the ant gut before the excretion of fecal fluid.

Implications for our general understanding of the leaf-cutting ant symbiosis and its bioconversion efficiency

The extent to which cellulose is degraded in the fungus garden has been widely disputed (see Grell et al., 2013 and references therein). It is noteworthy that the fecal fluid proteome only contained a single 1,4-beta-glucanase (Protein ID 55 in Table 1) to target the cellulose backbone, and that this protein was found in low quantities and in only three of the four examined colonies. We also found a single 1,4-beta-glucosidase (Protein ID 73 in Table 1) that releases glucose moieties from cellobiose and oligosaccharides, but this enzyme is also targeting hemicellulose in addition to cellulose. Finally, effective degradation of cellulose normally requires the enzyme cellobiohydrolase, which was only found in one colony in very low quantities (Supplementary file 1). These findings are consistent with our earlier inference (Grell et al., 2013; Moller et al., 2011) that cellulose degradation is not a primary function of the fecal fluid, supporting earlier findings that recalcitrant plant cell wall cellulose is not degraded to a significant degree in Acromyrmex fungus gardens, but discarded from the colony as waste (Moller et al., 2011). Our finding that Fenton chemistry may be used for breaking chemical bonds in cell wall lignocellulose and is likely sufficient to give hyphal degradation enzymes access to their target substrates in the interior of plant cells, appears to corroborate this notion. More complete cellulose degradation would release an excess of assimilatory sugars for the fungal symbiont but without simultaneously providing nitrogen or phosphorus, which are the limiting factors for growth in any functional herbivore (Elser et al., 2007). Such sugars could well be a burden for the symbiosis if they would allow inadvertent microorganisms to thrive in fungus gardens (Grell et al., 2013).

While the documentation of Fenton chemistry substantially increases our understanding of the way in which Acromyrmex leaf-cutting ants could become crop pests by robust functional herbivory, it is intriguing that we have never observed Fenton pellets in any of the Atta colonies that we have maintained in the lab for ca. 25 years. This suggests that the colonies of this other genus of leaf-cutting ants, which are approximately two orders of magnitude larger in size and considerably more damaging as agricultural pests have evolved alternative mechanisms to boost the efficiency and robustness of their functional herbivory. Given that also Atta colonies produce enormous amounts of cellulose rich waste, we do not expect that the explanation for Atta’s success as functional herbivores will be fundamentally different. This is because both genera rear very closely related lineages of the L. gongylophorus cultivars in sympatry at our Panamanian sampling site (Kooij et al., 2015b) and across their Latin American distributions (Silva-Pinhati et al., 2004). We suspect therefore that obtaining a fecal fluid proteome of Atta might shed interesting light on whether these ants can also use forms of inorganic chemistry, or whether they rely on an enzymatic innovation not available in Acromyrmex. Another point of interest is that human in vitro experiments with Fenton chemistry have produced up to fivefold increases in the production of sugars from feedstock, although the efficiency remains highly dependent on the plant material used (Kato et al., 2014; Jung et al., 2015; Hegnar et al., 2019). A major challenge in these human bioconversion applications has been that the Fenton generated reactive oxygen species degrade the very enzymes that are instrumental in maintaining sustainable conversion efficiency. An interesting venue for further research is therefore to elucidate how Acromyrmex fecal fluid degradation enzymes avoid being harmed by the hydroxyl radicals that accompany them.

Finally, we are used to think of our own crops as being in continuous need of manure to boost nutrient availability and growth. The term manure now seems inappropriate for attine ant fecal fluid even though deposition of fecal droplets in the fungus garden seems to suggests this analogy. In retrospect, abandoning this term seems logical, because the ants rear a heterotrophic crop and it seems hard to imagine how fresh leaf substrate can be further enriched with nutrients. Our present results suggest that it is also ambiguous to use the term ‘fecal’ for the droplets that leave the hindguts of leaf-cutting ants. In a structural sense this is fecal material, but in a functional sense this fluid now seems to be analogous to a joint symbiotic circulation system tinkered together by natural selection over millions of years. The primary function of this higher-order circulation system appears to be optimizing the distribution of gongylidia-produced decomposition compounds to maintain the combined superorganismal body of ants, ant brood, and fungus garden. In this perspective, it also becomes ambiguous which partner is the farmer or the crop, because the relationship is symmetrical in its mutual dependence and the cultivar has remarkable agency even though it neither has brains to coordinate or legs to move around. This perspective is worked out in more detail elsewhere (Boomsma, 2022) and suggests that we need to be careful in applying anthropomorphic language to naturally evolved symbioses outside the human domain.

Materials and methods

Biological material

Request a detailed protocol

Colonies of Acromyrmex echinatior (numbers Ae150, Ae160, Ae263, Ae322, Ae356, Ae372, Ae480, and Ae490) were collected in Gamboa, Panama between 2001 and 2010 and maintained in the laboratory at ca. 25°C and ca. 70% relative humidity (Bot and Boomsma, 1996) where they were supplied with a diet of bramble leaves, occasional dry rice, and pieces of apple. Fecal droplets were collected by squeezing large worker ants with forceps on the head and abdomen until they deposited a drop of fecal fluid. For mass spectrometry, 0.5 µl of double distilled and autoclaved water was added to the droplet before it was collected with a micro pipette and transferred to an Eppendorf tube with Run Blue loading buffer (Expedeon Inc). For the iron assay, Fenton reaction assay, hydrogen peroxide assay, and iron reduction assay, the fecal droplets were deposited on parafilm and collected undiluted with a 5 µl glass capillary.

Mass spectrometry

Request a detailed protocol

Ten fecal droplets from each of the four colonies were run on a 12% Run Blue polyacrylamide gel (Expedeon Inc) in a Mini-Protean II electrophoresis system (Biorad) for about 10 min at 90 mA. The gel was stained in Instant Blue Coomassie (Expedeon Inc) overnight at room temperature. The stained part of each lane was then cut out with a scalpel and transferred to Eppendorf tubes, after which the samples were sent to Alphalyse (Denmark) for analysis.

The protein samples were reduced and alkylated with iodoacetamide (i.e. carbamidomethylated) and subsequently digested with trypsin. The resulting peptides were concentrated by Speed Vac lyophilization and redissolved for injection on a Dionex nano-LC system and MS/MS analysis on a Bruker Maxis Impact QTOF instrument. The obtained MS/MS spectra were analyzed using the MaxQuant software package version 1.5.4.1 (Cox and Mann, 2008) integrated with the Andromeda search engine (Cox et al., 2011) using standard parameters. The spectra were matched to a reference database of the predicted Acromyrmex echinatior proteome (Nygaard et al., 2011) combined with published predicted Leucocoprinus gongylophorus proteomes originating from the Acromyrmex echinatior, Atta cephalotes, and Atta colombica fungal symbionts (Aylward et al., 2013; Nygaard et al., 2011; Nygaard et al., 2016), as well as single protein sequences predicted from manually PCR amplified and sequenced cDNA sequences (Schiøtt et al., 2010; De Fine Licht et al., 2013; Kooij et al., 2016; Kooij et al., 2014b; Schiøtt et al., 2008). As these published fungal proteomes contained many redundant protein sequences, a non-redundant reference protein database was constructed using CD-Hit (RRID:SCR_007105) (Li and Godzik, 2006) keeping only the longest sequences of the sets of 100% matching sequences. Annotation of the identified protein sequences was performed using Blast2GO (RRID:SCR_005828) (Hegnar et al., 2019) combined with manual BlastP searches against the NCBI nr database.

Iron level measurements

Request a detailed protocol

Fecal fluid iron content was measured according to Fish, 1988. Fecal fluid from 20 workers from each of six colonies (Ae150, Ae160, Ae263, Ae356, Ae480, and Ae490) was collected with glass capillaries and weighed in 0.2 ml PCR tubes to determine the volume of each sample, assuming a density of 1 g/ml. The samples were diluted to 100 µl with 0.1 M HCl, mixed with Reagent A (0.142 M KMnO4, 0.6 M HCl), and incubated for 2 hr at 60°C. Subsequently 10 µl Reagent B (6.5 mM Ferrozine, 13.1 mM neocuproine, 2 M ascorbic acid, 5 M ammonium acetate) was added to each sample and incubated for 30 min at room temperature, after which 150 µl of this mixture was transferred to a microtiter plate and the absorbance at 562 nm measured in an Epoch microplate spectrophotometer (BioTek). Samples with known amounts of iron were processed in parallel with the unknown samples to produce a standard curve from which the iron content of the unknown samples could be determined. Three series of standards were made from ferrous ethylenediammonium sulfate tetrahydrate, ferrous sulfate heptahydrate, and ferric chloride hexahydrate. In all cases a solution of 5 mg/l iron in 0.1 M HCl was used to make a dilution series ranging from 78.1 µg/l to 5 mg/l. In addition, a blank sample without iron was included and used to correct for background absorbance.

Hydroxyl radical measurements

Request a detailed protocol

0.5 µl fecal fluid from each of the colonies Ae150, Ae160, Ae263, Ae356, Ae480, and Ae490 was diluted with 0.5 µl distilled H2O and mixed with either 1 µl H2O, 1 µl 40 mM 1,10-phenanthroline (dissolved in methanol to 200 mM and diluted 1:4 with distilled H2O) or 1 µl 20% methanol. After 15 min incubation at room temperature, 2 µl substrate (100 mM 2-deoxy-D-ribose in 0.1 M sodium acetate buffer pH 5.0) was added, and samples were incubated for 30 min at room temperature. Then, 4 µl of a solution of 1% 2-thiobarbituric acid in 50 mM NaOH was added followed by addition of 4 µl of 2.8% trichloroacetic acid. The samples were then incubated for 20 min at 99.9°C in a PCR machine, after which 2 µl was used for measuring the absorbance at 532 nm using a NanoDrop ND-1000 spectrophotometer (Halliwell et al., 1988). Control samples without addition of 2-deoxy-D-ribose were made in parallel and the absorbance value subtracted from the absorbance values of the samples. Similarly, control samples with 2-deoxy-D-ribose added but without fecal droplets were made in parallel so also their absorbance value could be subtracted from those of the samples. Control samples without both 2-deoxy-D-ribose and fecal droplets were also run in parallel and these absorbance values were added to the final values to compensate for subtracting the absorbance values of two controls from each sample value.

H2O2 measurements

Request a detailed protocol

0.5 µl fecal fluid from each of the colonies Ae150, Ae160, Ae263, Ae356, Ae480, and Ae490 was mixed with either 150 µl demineralized H2O or with 150 µl 10 mM glucose, 10 mM glyoxal, 10 mM veratryl alcohol, 10 mM methanol, or 100 U catalase. The samples were incubated for 60 min at 30°C in a PCR machine. To avoid direct oxidation of the amplex red substrate by laccases in the fecal fluid, the samples were transferred to Amicon Ultra-0.5 Centrifugal filters with 3 Kda membranes (Merck Millipore UFC500396) and centrifuged at 14,000 g for 30 min. 100 µl flow through was transferred to a microtiter plate and 100 µl amplex red reagent (100 µM Ampliflu red, 2 U/ml horse radish peroxidase, in 0.1 M phosphate buffer pH 7.4) (Zhou et al., 1997) was added, after which the absorbance at 560 nm was measured for 30 min using an Epoch microplate spectrophotometer (BioTek). This absorbance measure was then used to calculate the H2O2 concentration using a standard curve made from a twofold serial dilution of H2O2 ranging from 1.56 µM to 100 µM and including a blank sample without H2O2. The standard samples were assayed in parallel with the fecal fluid samples.

Iron reduction measurements

Request a detailed protocol

One microliter fecal fluid from each of the colonies Ae150, Ae160, Ae263, Ae356, Ae480, and Ae490 was mixed with 71 µl demineralized H2O, 10 µl 1 mM FeCl3 · 6 H2O, 115 µl 0.1 M acetate buffer pH 4.4, and either 23 µl H2O or 23 µl 100 mM glucose. Subsequently, 10 µl 1% FerroZine was added and the absorbance at 562 nM was measured during a 30 min period using an Epoch microplate spectrophotometer (BioTek) (Goodell et al., 2006). The rates of change in absorbance from 5 to 30 min were used to calculate the amount of Fe3+ reduced to Fe2+ per minute, using a standard curve made from a twofold serial dilution of FeSO4 · NH3CH2CH2NH3SO4 · 4 H2O, ranging from 5 mM to 0.078 mM and including a blank sample without iron. The standard samples were assayed in parallel with the fecal fluid samples.

Statistical data analyses

Request a detailed protocol

All analyses were done in RStudio Version 0.99.491 (RRID:SCR_000432). Shapiro–Wilk tests were used to test whether differences in response variables between treatments within the same ant colonies followed a normal distribution. For hydrogen peroxide measurements a square root transformation of response data was applied to make the data pass the Shapiro–Wilk test. Planned comparisons between treatments were performed with paired t-tests (by pairing measurements of samples coming from the same ant colonies). p-values were adjusted for multiple comparisons using the Holm–Bonferroni method.

Phylogenetic analyses

Request a detailed protocol

The five identified fecal fluid GMC oxidoreductase sequences were aligned with 15 basidiomycete GMC oxidoreductase sequences belonging to each of the five functional GMC oxidoreductase protein groups presented in Ferreira et al., 2015, using M-coffee (http://tcoffee.crg.cat/apps/tcoffee/do:mcoffee) with default parameter settings. The alignment file was then used in PhyML (RRID:SCR_014629) with Smart Model Selection (http://www.atgc-montpellier.fr/phyml-sms/) and aLRT SH-like support values. The best DNA substitution model turned out to be LG+G+I+F.

Data availability

The mass spectrometry data generated in this project as well as the amino acid sequences used for protein identification have been deposited to the ProteomeXchange Consortium via the PRIDE (Perez-Riverol et al. (2019). Nucleic Acids Research 47:D442–D450.) partner repository with the dataset identifier PXD016395 (http://www.ebi.ac.uk/pride/archive/projects/PXD016395). All other data are provided in the Supplementary file S1 and the source data files provided for Figures 3 and 4.

The following data sets were generated
    1. Schiøtt M
    2. Boomsma JJ
    (2020) PRIDE
    ID PXD016395. Leaf-cutting ant feces LC-MS/MS.
    http://www.ebi.ac.uk/pride/archive/projects/PXD016395

References

  5. Book
    1. Boomsma JJ
    (2022)
    Lifetime Commitment between Social Insect Families and Their Fungal Cultivars Complicates Comparisons with Human Farming
    In: Schultz T. R, Gawne R, Peregrine P. R, editors. The Convergent Evolution of Agriculture in Humans and Insects. Massachusetts: MIT Press. pp. 563–595.
    1. Bot ANM
    2. Boomsma JJ
    (1996)
    Variable metapleural gland Size-Allometries in Acromyrmex leafcutter ants (Hymenoptera: formicidae)
    Journal of the Kansas Entomological Society 69:375–383.
  7. Book
    1. Bourke AFG
    (2011)
    Principles of Social Evolution
    Oxford University Press.
    1. Mehdiabadi NJ
    2. Schultz TR
    (2010)
    Natural history and phylogeny of the Fungus-Farming ants (Hymenoptera: formicidae: myrmicinae: attini)
    Myrmecological News 13:37–55.
  63. Book
    1. Smith JM
    2. Szathmáry E
    (1997)
    The Major Transitions in Evolution
    Oxford University Press.
    1. Wu JX
    2. Li J
    (2015)
    The Arabidopsis ceramidase AtACER functions in disease resistance and salt tolerance
    Plant Journal 81:767–780.

Decision letter

  1. Meredith C Schuman
    Senior Editor; University of Zurich, Switzerland
  2. Dieter Ebert
    Reviewing Editor; University of Basel, Switzerland
  3. Maridel Fredericksen
    Reviewer; University of Basel, Switzerland
  4. Heike Feldhaar
    Reviewer; University of Bayreuth, Germany
  5. Michiel Op De Beeck
    Reviewer

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

The study by Schiøtt and Boomsma moves our understanding of the intricate mutualistic interaction between leaf-cutter ants and their fungus a significant step forward. Using a combination of proteomics and subsequent specific chemical tests to understand the nature of leaf degradation the authors present compelling evidence that a Fenton reaction plays a crucial role in the degradation of the plant material, which is otherwise hard to break down. The chemicals for the Fenton reaction are provided in a unique way by both the fungus and the ants, stressing once more the very tight mutualism between these species.

Decision letter after peer review:

Thank you for submitting your work entitled "Proteomics reveals synergy between biomass degrading enzymes and inorganic Fenton chemistry in leaf-cutting ant colonies" for consideration by eLife. Your article has been reviewed by four peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Meredith Schuman as the Senior Editor.

Four reviewers have read this study and they are all enthusiastic about the results. All reviewers made a number of comments that should be addressed before publications.

The most prominent suggestions are:

1) A recurrent issue in several comments raised by the reviewers is a lack of clarity. Given the interdisciplinary nature of the study, particular care should be given to make the material better accessible for non-specialists. This requires a better organisation of the way certain aspects of the material is presented (including the logical steps from proteome analysis), explanation of terminology and better explanation of aspects that are specific to the chemical details of the finding.

2) All reviewers had issues with specific aspects of the description of the chemical aspects of the study. Here consistency and clarity is of particular importance.

3) The title should be more descriptive with regard to the study.

4) The study refers at various places to brown rot fungi. The relevance of this, the physiological, biochemical and phylogenetic link to the ant-fungus is not always clear.

All the comments by the four reviewers are listed below.

Reviewer #1:

The study by Schiøtt and Boomsma moves our understanding of the intricate mutualistic interaction between leaf-cutter ants and their fungus a significant step forward. Using a combination of proteomics and subsequent specific chemical tests to understand the nature of leaf degradation the authors present compelling evidence that a Fenton reaction plays a crucial role in the degradation of the plant material, which is otherwise hard to break down. The chemicals for the Fenton reaction are provided in a unique way by both the fungus and the ants, stressing once more the very tight mutualism between these species.

I am not good enough in chemistry to judge if the chemical tests presented are as clear in their interpretation as it is presented here; this needs to be assessed by a biochemist. Other than this I have little comments.

My main comment is to make the study more accessible to non-specialist readers. The text meanders at places and needs more of a clear line. Terminology needs to be explained better. I had to consult Wikipedia several times while reading this study.

The title is not very clear. As it stands it misses, what I consider a main point, namely that biomass degradation is achieved by a synergy of fungal and ant chemistry.

Reviewer #2:

The authors have conducted a thorough study of the fecal fluid proteome in Acromyrmex echinator and its potential adaptive function in maintaining the ants' symbiosis with the fungus Leucocoprinus gongylophorus. The authors provide strong evidence that Fenton reactions occur within pellets of chewed plant matter mixed with ant fecal fluid, and they suggest these reactions enable the efficient digestion of plant material. Weaker evidence suggests both ant and fungal proteins maintain the Fenton reactions. Overall, this study advances our understanding of chemical mechanisms permitting a textbook example of symbiosis, and it raises intriguing questions for future study regarding the evolutionary origins of this and other leaf-cutter ant – fungal symbioses.

1) Several points regarding glucose dehydrogenase and its proposed role in the Fenton reaction need to be clarified

– How did the researchers choose to focus on this particular protein as the candidate for the reducing agent in the Fenton reaction? Why did this protein stand out from the list (besides being the most abundant of the ant proteins)? Does it play a similar role in other systems?

– In the Discussion, the reader learns that laccases are apparently an alternative to glucose dehydrogenase in a Fenton reaction, and these laccases are also present in the fecal fluid, so this raises even more questions as to why glucose dehydrogenase was chosen as the candidate.

The story with the laccases, including the preliminary experiments, should be introduced sooner (e.g. when introducing the hypothesis involving glucose dehydrogenase)

– What is the relationship between "glucose dehydrogenase" and "GMC oxidoreductase" (from the protein table, it seems that glucose dehydrogenase is a type of GMC oxidoreductase, but this is not clear in the text). Also, what is its relationship to aryl alcohol oxidases, and is it also present in brown-rot fungi?

– Is glucose dehydrogenase the only enzyme that would increase its activity in response to additional glucose? (i.e. could the effect of adding glucose be caused by another candidate enzyme?)

2) It is not clear how long the Fenton reactions can be maintained, given the iron recycling which was demonstrated. The authors imply that Fe2+ is not the limiting reagent, but the evidence for this is not apparent. Iron reduction was measured at "considerable capacity", but in relation to what? The comparison to human blood (Results and Discussion) is not relevant without additional context: How does it compare to the wood substrate of brown-rot fungi? How much is needed to maintain the Fenton reaction? How does the recycling rate compare to the consumption rate?

3) The authors argue that the fecal fluid of A. echinator contains a "specialized" set of proteins that have been adaptively selected to escape digestion. The main evidence they give is that the proteins consistently produce the same banding pattern on gels, and that most of the proteins were found in three of four colony samples. However, the reader is given no phylogenetic context for this comparison: do other ant species (or any other species) have fecal proteomes that appear more "random" on gels? Moreover, we know these ants always eat the same food, so why should we expect their fecal proteomes to vary over time or between colonies?

4) Language: Many sentences are quite long and complex. I recommend keeping sentences under 30 words as a rule of thumb. Also, there are several errors with grammar and punctuation, as well as misplaced modifiers (e.g. Introduction, Results), which make the paper laborious to follow. I recommend consulting with a professional editor to improve language use and overall clarity of the manuscript.

Reviewer #3:

Already decades ago it was suggested that the fecal fluid of leaf-cutter ants contains enzymes required for the breakdown of plant material that are produced by the fungal cultivar. While this was corroborated in recent studies on the proteome of the fecal fluid as well as genome evolution of both, the ants and the fungal cultivar, the mechanism of the actual breakdown of the plant material still remained a black box.

Based on a more in-depth proteome study of the fecal fluid of the leaf-cutter ant Acromyrmex echinator the authors can show that the fecal fluid predominantly contains fungal-produced enzymes. While fungal laccases were already known from the fecal fluids, here the authors present data that many enzymes may contribute to breakdown of plant material. Based on the idea that abundantly occurring oxidoreduxctases produce hydrogen peroxide the authors test the hypothesis that this molecule may react with iron to produce reactive oxygen radicals, i.e. the inorganic Fenton reaction would proceed after ants place a droplet of fecal fluid on ant-produced pellets of munched leaf material that are placed onto the surface of the fungal cultivar (“Fenton ball”). The reactive oxygen species may then damage plant cell walls to an extent that enables the fungal cultivar to enter into the cells and obtain nutrients.

Taken together I think that the experimental evidence showing that the Fenton reaction proceeds in the “Fenton balls” on the surface of the fungal cultivar is very strong. Overall, the study is of very high quality and is well worth publishing.

1) I suggest that the authors explain whether there is a link between brown rot fungi and the ant cultivar. Brown rot fungi typically use hydrogen peroxide and oxygen radicals to aid in breakdown of cellulose and hemicellulose. Is there a phylogenetic relationship of the ant cultivar to brown rot fungi? Or has the trait of producing H2O2 and the entailing reactions evolved multiple times?

2) In brown rot fungi aromatic alcohols play an important role. These alcohols are oxidized in order to produce H2O2. The aromatic alcohol veratryl alcohol produced by brown rot fungi was also used here in experimental procedures, where the authors showed that H2O2 is produced in larger quantities when it is present. Is anything known about the concentration of such aromatic alcohols in the fecal fluid of leaf-cutter ants? How would it be produced (and by whom?)?

3) Is there any evidence for a high concentration of citrate in the gut of leaf-cutter ants? Indirect evidence is presented by the authors (Results) with a Mollicute bacterium that has a citrate catabolizing pathway. However, is this bacterium part of the core microbiome? Any possibility to provide more direct evidence here? The Mollicute bacterium would only be an opportunist scavenging the citrate? In Sapountzis et al., 2018, the information is given that the leaf-cutter workers would consume relatively high levels of citrate with juices from freshly cut plant material. I think that this link may be more convincing.

4) I really like the point of the Discussion where the authors point out that the Fenton balls are spatially separated bioconversion reactors that prevent tissue damage of either ant of fungus when high levels of reactive oxygen is presumably produced.

5) A very interesting point that the authors touch for future studies is that other higher leaf-cutter ants of genus Atta do not seem to make use of mechanism for the breakdown of plant material studied here for Acromyrmex. Is this really a total black box? Or could the authors at least provide an idea how it works in Atta?

6) In contrast, I think the last section of the Discussion is only partially helpful. While I agree with the authors that the high level of physiological complementarity and the large proportion of fungal enzymes in the fecal fluid renders the term “manure” inappropriate I do think that the suggested term “bloodstream” is not very helpful. The superorganism does not have a bloodstream (and also not a joint symbiotic one). I am not sure whether this highly anthropomorphic analogy is helpful in the future. I suggest to tone down this paragraph and stress the functional physiological complementarity more.

Reviewer #4:

The work presented by Schiøtt and Boomsma is a very interesting study investigating how the Fenton reaction assists enzymatic digestion of plant material collected by leaf-cutting ants in fecal substrate pellets.

The text is very enjoyable to read.

The methodologies are well-described and the findings are of interest.

1) In the Introduction, it is stated that saprotrophs are not well adapted to efficiently degrade fresh plant material. I think there are many sparotrophic fungi well-adapted to do this.

2) Throughout the work, oxidoreductases have been categorized separately from other CAZY enzymes. For example, in the Results and in Table 1 and Figure 1. However, enzymes such as laccase (AA1), GMC oxidoreductase/aryl-alcohol oxidase (AA3), etc. are also classified as CAZY enzymes, although as having auxiliary activities. Could oxidoreductases be structured as a sub-category of CAZY enzymes to distinguish them from hydrolytic enzymes?

3) In the Results and Discussion, it is stated that brown rot fungi rely on iron already present in the wood to employ the Fenton reaction. This idea may be somewhat outdated as XRF studies in Serpula lacrymans have shown that this brown rot fungus actively increases the iron concentrations in wood before decomposition. See Kirker et al., 2017. Synchrotron-based X-ray fluorescence microscopy enables multiscale spatial visualization of ions involved in fungal lignocellulose deconstruction. Scientific Reports 7, Article number: 41798.

4) Discussion. I do not see how laccases in the fecal fluid can be considered functional analogs of brown rot recycling mechanisms of Fe(III) to Fe(II) directly. Laccases in the system described by Guillen et al., 2000 oxidise quinones. Reduced quinones in that system are responsible for reducing Fe(III) to Fe(II) and the production of H2O2. This sentence reads as if the laccases perform this function instead of the quinones.

5) Results. Not only enzymes can reduce Fe(III) to Fe(II) in brown rot fungi. The recycling of Fe(III) to Fe(II) in brown rot systems can be attributed to quinones as well.

https://doi.org/10.7554/eLife.61816.sa1

Author response

Four reviewers have read this study and they are all enthusiastic about the results. All reviewers made a number of comments that should be addressed before publications.

The most prominent suggestions are:

1) A recurrent issue in several comments raised by the reviewers is a lack of clarity. Given the interdisciplinary nature of the study, particular care should be given to make the material better accessible for non-specialists. This requires a better organisation of the way certain aspects of the material is presented (including the logical steps from proteome analysis), explanation of terminology and better explanation of aspects that are specific to the chemical details of the finding.

2) All reviewers had issues with specific aspects of the description of the chemical aspects of the study. Here consistency and clarity is of particular importance.

3) The title should be more descriptive with regard to the study.

4) The study refers at various places to brown rot fungi. The relevance of this, the physiological, biochemical and phylogenetic link to the ant-fungus is not always clear.

We have endeavoured to clarify along the lines summarized here, but could not find an alternative title that could capture more of the key focus of our paper and still remain within the maximally allowed number of characters. See below for a more detailed response.

All the comments by the four reviewers are listed below.

Reviewer #1:

[…] I am not good enough in chemistry to judge if the chemical tests presented are as clear in their interpretation as it is presented here; this needs to be assessed by a biochemist. Other than this I have little comments.

My main comment is to make the study more accessible to non-specialist readers. The text meanders at places and needs more of a clear line. Terminology needs to be explained better. I had to consult Wikipedia several times while reading this study.

The title is not very clear. As it stands it misses, what I consider a main point, namely that biomass degradation is achieved by a synergy of fungal and ant chemistry.

We struggled with the stringent limitation of maximally 120 characters being allowed for the title.

If we understand the reviewer well, (s)he finds that our present title:

“Proteomics reveals synergy in biomass conversion between fungal enzymes and inorganic Fenton chemistry in leaf-cutting ant colonies”

lacks the word “ant-chemistry”. However, all three keywords “synergy”, “fungal” and “ant” are there, so it is largely a matter of semantics whether one finds this title captures the key elements of our study. We have toyed with various alternatives but found none of them were improvements. This exercise also reminded us that we arrived at the present title on purpose because the chemistry participation of the ants is in fact rather minor and indirect compared to the more major and direct involvement of the fungal cultivar. The ant roles are of key significance but that is because of vectoring of fecal droplets and fine-tuned behavioral facilitation. Also the majority of the fecal fluid proteome is not ant derived (86 fungal/15 ant) and only a single enzyme involved in the Fenton reaction was encoded by the ant genome. We therefore concluded in the end that we prefer to keep the title unchanged because it will, together with the Abstract, convey the take-home messages of this study quite accurately.

The only alternative title that we could think of:

“Proteomics reveals that leaf-cutting ants and their fungal symbiont synergistically combine Fenton chemistry with enzymatic degradation of plant biomass”

exceeds the character limit, so the question is whether you feel this is a sufficient improvement to give dispensation from that rule.

Reviewer #2:

[…] 1) Several points regarding glucose dehydrogenase and its proposed role in the Fenton reaction need to be clarified

o How did the researchers choose to focus on this particular protein as the candidate for the reducing agent in the Fenton reaction? Why did this protein stand out from the list (besides being the most abundant of the ant proteins)? Does it play a similar role in other systems?

As mentioned in the text, we conjectured it there should be a mechanism to recycle oxidized iron into reduced iron in order to keep the Fenton reaction running. Scrutinizing Table 1 for proteins that could have such a function, the glucose dehydrogenase stood out because: 1) It is an oxidoreductase and therefore capable of mediating redox reactions. 2) Using glucose as an electron donor would mean that it would constantly be supplied with substrate because the degradation of plant material liberates glucose molecules in a constant flow. 3) It appeared to always be present in high quantities. We have not been able to find other examples in the literature that document the use of glucose dehydrogenase for this purpose, and we also do not claim to have proved beyond doubt that this is the prime function of the identified glucose dehydrogenase. But the fact that addition of glucose increased the iron reduction capability of the fecal fluid was consistent with this hypothesis, and glucose dehydrogenase has also been suggested to be used for glucose liberation functions in industrial Fenton chemistry, once more consistent with such a function being an obvious possibility. We note that glucose dehydrogenase may not directly catalyze the iron reduction reaction, but may depend on other enzymes catalyzing intermediate steps in the process, with the ultimate result of reducing oxidized iron. Finally, cellobiose dehydrogenase is known to be able to reduce oxidized iron with cellobiose as substrate, which adds to the plausibility of glucose dehydrogenase have a similar function.

– In the Discussion, the reader learns that laccases are apparently an alternative to glucose dehydrogenase in a Fenton reaction, and these laccases are also present in the fecal fluid, so this raises even more questions as to why glucose dehydrogenase was chosen as the candidate.

We explicitly reported in the manuscript that simple experiments adding laccase substrate to the fecal fluid did not affect its iron reducing capabilities, which suggests that laccases have roles that are independent of Fenton chemistry. We admit that more elaborate experiments on this would have been nice to do, but they are very time consuming and were thus beyond the scope of the present study.

So, while we do not rule out the possibility of laccase activity somehow interacting with Fenton chemistry, it does not seem an obvious possibility. We note that Fenton pellets are typically produced when colonies receive leaf-resources ad libitum, so it may be that laccase decomposition primarily works when resources are less abundant and glucose becomes a more valuable resource.

The story with the laccases, including the preliminary experiments, should be introduced sooner (e.g. when introducing the hypothesis involving glucose dehydrogenase)

A sentence on the preliminary laccase experiments has now been added to the Results section.

– What is the relationship between "glucose dehydrogenase" and "GMC oxidoreductase" (from the protein table, it seems that glucose dehydrogenase is a type of GMC oxidoreductase, but this is not clear in the text). Also, what is its relationship to aryl alcohol oxidases, and is it also present in brown-rot fungi?

It is correct that glucose dehydrogenase belongs to the GMC oxidoreductase protein family, but as it does not produce hydrogen peroxide, it cannot alone be mediating the Fenton reaction. Also, being an ant-encoded protein, it was not meaningful to include it into the gene tree in Figure 3, which had only fungal proteins. Glucose dehydrogenases are closely related to glucose oxidases (GOX), which means that they are also (but more distantly) related to aryl alcohol oxidases. They are present in both white rot fungi and brown rot fungi. As our manuscript was already quite long, we chose not to prioritize detailed further descriptions of these issues as that would have mostly distracted from the overall interpretational narrative. We have, however, now added the sentence: “Note that the ant encoded glucose dehydrogenase is also a GMC Oxidoreductase, but would not fit into this phylogeny of fungal proteins” to the legend of Figure 3 to make clear that we were aware of this issue.

– Is glucose dehydrogenase the only enzyme that would increase its activity in response to additional glucose? (i.e. could the effect of adding glucose be caused by another candidate enzyme?)

To our knowledge, glucose dehydrogenase is the only enzyme in the fecal fluid proteome that can use glucose as a substrate. However, there are a few hypothetical proteins in the list for which we do not know the function. Although our proteome is very complete, it also remains possible that some proteins somehow escaped being identified by the methods that we used.

2) It is not clear how long the Fenton reactions can be maintained, given the iron recycling which was demonstrated. The authors imply that Fe2+ is not the limiting reagent, but the evidence for this is not apparent.

We have not done experiments to look into this issue, partly because the fecal fluid material is only available in small quantities, which limits what kind of experiments can be feasibly done. In principle we believe the reaction can go on as long as oxygen, aromatic alcohols, and glucose are all present, of which the amount of alcohols will probably be the limiting component. However, as the fecal fluid also contains a lot of proteases, the limiting factor may in reality be proteolytic degradation of the relevant enzymes. To look into these issues, it will probably be needed to mix the various components in vitro at a much larger scale, which we have not tried yet. We have now deleted the sentence stating that the Fenton reaction appeared not to be limited by metal ions.

Iron reduction was measured at "considerable capacity", but in relation to what?

Unfortunately it is very difficult to compare this value to other studies, as we do not know the amount of the relevant enzymes in the fecal fluid. But what matters in this context is that we can conclude that iron reduction takes place in the fecal fluid and that the velocity of the reaction increases after adding glucose. The exact values are therefore not so important. We have now changed “considerable capacity” to “measurable capacity”.

The comparison to human blood (Results and Discussion) is not relevant without additional context: How does it compare to the wood substrate of brown-rot fungi?

We have now revised the relevant paragraphs, removing the comparison to human blood and instead focusing on how this value compares to other systems in which Fenton chemistry plays a role (Results and Discussion).

How much is needed to maintain the Fenton reaction? How does the recycling rate compare to the consumption rate?

As mentioned above, even though these kinds of experiments are relevant to do, the small amounts of fecal fluid material that can be obtained by pressing it out of individual workers and the possible interactions with many other substances in the fluid makes such experiments difficult without having an established in vitro model system. Also here, we do not claim to have solved all questions related to the proposed scenario, but present what we think is compelling cumulative evidence to suggest that Fenton chemistry plays a key role in the degradation of plant substrate in leaf-cutting ant fungus gardens. Thereby we will hopefully pave the way for further investigation of this subject in the same or other model systems.

3) The authors argue that the fecal fluid of A. echinator contains a "specialized" set of proteins that have been adaptively selected to escape digestion. The main evidence they give is that the proteins consistently produce the same banding pattern on gels, and that most of the proteins were found in three of four colony samples. However, the reader is given no phylogenetic context for this comparison: do other ant species (or any other species) have fecal proteomes that appear more "random" on gels? Moreover, we know these ants always eat the same food, so why should we expect their fecal proteomes to vary over time or between colonies?

First of all, proteins are very valuable sources of nitrogen, and most organisms would degrade and assimilate as much food protein as possible. That we repeatedly find the same set of proteins, and that these proteins are still functional after gut passage, and with functional roles relevant for degradation of plant biomass, cannot be a coincidence. It is true that the diet of leaf-cutting ants mostly consists of fungal hyphae, but all we say is that if the escape from digestion was just a coincidence, the banding pattern on the gels should be random, which it is definitely not. Consequently there must be a predictable reason why these proteins are not digested, and given that their putative functions make sense when interpreted as mediating a sequence of substrate degradation steps, it seems fair to interpret this phenomenon as an advanced symbiotic adaptation. We are not aware of studies looking into fecal proteomes of other ants, and studies on humans and other mammals typically find bacterial proteins and host proteins involved in inflammation. As we illustrated in the final Discussion paragraph, it appears that the fungus-farming symbiosis makes it ambiguous whether ant feces can still be considered as waste.

4) Language: Many sentences are quite long and complex. I recommend keeping sentences under 30 words as a rule of thumb. Also, there are several errors with grammar and punctuation, as well as misplaced modifiers (e.g. Introduction, Results), which make the paper laborious to follow. I recommend consulting with a professional editor to improve language use and overall clarity of the manuscript.

The manuscript has been revised to generally improve clarity of communication.

Reviewer #3:

[…] 1) I suggest that the authors explain whether there is a link between brown rot fungi and the ant cultivar. Brown rot fungi typically use hydrogen peroxide and oxygen radicals to aid in breakdown of cellulose and hemicellulose. Is there a phylogenetic relationship of the ant cultivar to brown rot fungi? Or has the trait of producing H2O2 and the entailing reactions evolved multiple times?

The brown rot and white rot fungi terminology only applies to wood degrading fungi found within the basidiomycete subdivision Agaricomycotina. It is a functional categorization based on the ability to degrade lignin, rather than a phylogenetic categorization. The mechanism used by brown rot fungi to degrade wood is probably the default one, used by lineages that have not developed the white rot mechanism of active lignin degradation or have secondarily lost it. The attine ant fungal symbionts belong to the order Agaricales within the Agaricomycotina and we have shown previously (Nygaard et al., 2016) that the leaf-cutting ant cultivar has lost the (white-rot-analogous) lignin peroxidases that the cultivars of other, phylogenetically more basal, attine ant species have. It thus appears like the leaf-cutting ant cultivar has reverted to a brown rot-like mechanism from a white rot-like mechanism used by its ancestors. It thus seems reasonable to expect that this loss only happened after alternative and possibly more sophisticated mechanisms for lignin degradation had evolved. We have now explained this more explicitly (Results).

2) In brown rot fungi aromatic alcohols play an important role. These alcohols are oxidized in order to produce H2O2. The aromatic alcohol veratryl alcohol produced by brown rot fungi was also used here in experimental procedures, where the authors showed that H2O2 is produced in larger quantities when it is present. Is anything known about the concentration of such aromatic alcohols in the fecal fluid of leaf-cutter ants? How would it be produced (and by whom?)?

To our knowledge this issue has never been investigated. As written in the manuscript, we conjecture that the fungal cultivars provide these alcohols, which would explain the upregulation of the phenylalanine/tyrosine pathway in the gongylidia (Fine Licht et al., 2014), the swollen hyphal tips that the ants ingest, because phenylalanine is the precursor of veratryl alcohol. This is clearly a subject that deserves to be looked into in more detail in the future.

3) Is there any evidence for a high concentration of citrate in the gut of leaf-cutter ants? Indirect evidence is presented by the authors (Results) with a Mollicute bacterium that has a citrate catabolizing pathway. However, is this bacterium part of the core microbiome? Any possibility to provide more direct evidence here? The Mollicute bacterium would only be an opportunist scavenging the citrate? In Sapountzis et al., 2018, the information is given that the leaf-cutter workers would consume relatively high levels of citrate with juices from freshly cut plant material. I think that this link may be more convincing.

To our knowledge the concentration of citrate in the guts of leaf-cutting ants has not been measured. We consider Mollicutes as part of the core gut microbiome of leaf-cutting ants, and in addition they are often found as intracellular bacteria in gut cells and Malpighian tubules. It is thus hard to imagine that they would only be opportunist bacteria. Several aspects of the sequenced genomes of these Mollicutes also indicate specialized functions as co-adapted symbionts (cf. Sapountzis et al., 2018). The Mollicute strain found in leaf-cutting ants has a citrate catabolizing pathway that is very rare among Mollicutes, and it thus seems most parsimonious to assume that they have acquired this pathway because they assimilate citrate from the gut lumen. However, the paragraph in question has now been removed because the possible link between Fenton chemistry and citrate catabolism became less important after revising the text on iron content of fecal fluid being surprisingly high.

4) I really like the point of the Discussion where the authors point out that the Fenton balls are spatially separated bioconversion reactors that prevent tissue damage of either ant of fungus when high levels of reactive oxygen is presumably produced.

We also find this an important part of the story.

5) A very interesting point that the authors touch for future studies is that other higher leaf-cutter ants of genus Atta do not seem to make use of mechanism for the breakdown of plant material studied here for Acromyrmex. Is this really a total black box? Or could the authors at least provide an idea how it works in Atta?

We have not looked into this, and would prefer not make unsolicited hypotheses in this manuscript. We suspect that it will take a complete fecal fluid proteome to begin understanding the adaptations that Atta must have evolved independently and differently. Although Fenton chemistry made sense after we put all the Acromyrmex puzzle pieces together, it would not be surprising if a different innovation of similar or even higher efficiency in Atta would turn out to be equally surprising.

6) In contrast, I think the last section of the Discussion is only partially helpful. While I agree with the authors that the high level of physiological complementarity and the large proportion of fungal enzymes in the fecal fluid renders the term “manure” inappropriate I do think that the suggested term “bloodstream” is not very helpful. The superorganism does not have a bloodstream (and also not a joint symbiotic one). I am not sure whether this highly anthropomorphic analogy is helpful in the future. I suggest to tone down this paragraph and stress the functional physiological complementarity more.

We appreciate this comment and can see that the “bloodstream” analogy was not particularly helpful. In fact the reason for adding this paragraph was, precisely as the reviewer writes, to be careful with anthropomorphic language. We have now replaced that term by “higher-order circulation system”, which we hope will be less distracting and help focus on the real issue, which is that ant farming has very few deep analogies with human farming even though established terminology continues to suggests such analogies. The second author now has an “in press” paper working out the details of these semantic issues that deserve to be reappreciated, which has now been added as source. We hope you can agree that ending the paper this way is reasonable (Discussion).

Reviewer #4:

[…] 1) In the Introduction, it is stated that saprotrophs are not well adapted to efficiently degrade fresh plant material. I think there are many sparotrophic fungi well-adapted to do this.

What we mean here is that saprotrophic fungi (in contrast to biotrophic and necrotrophic fungi) are not adapted to cope with live plant tissues that are still able to defend themselves against pathogens. If they were able to do that, they would most likely qualify as necrotrophic fungi. It is of course debatable whether the leaf substrate collected by the ants is still able to mount induced anti-herbivory defenses. However, given that many plants can be grown from plant cuttings we believe that cells in the leaf fragments are likely to remain alive for a considerable amount of time after harvest, and would thus probably still be able to produce substances to deter herbivores.

2) Throughout the work, oxidoreductases have been categorized separately from other CAZY enzymes. For example, in the Results and in Table 1 and Figure 1. However, enzymes such as laccase (AA1), GMC oxidoreductase/aryl-alcohol oxidase (AA3), etc. are also classified as CAZY enzymes, although as having auxiliary activities. Could oxidoreductases be structured as a sub-category of CAZY enzymes to distinguish them from hydrolytic enzymes?

It is correct that oxidoreductases are listed in the CAZy database in the category Auxiliary Activities. However, in the description of this category it is stated that it covers “redox enzymes that act in conjunction with CAZymes”, which implies that they are not CAZymes themselves. Also, since some of the oxidoreductases listed in Table 1 are not covered by the CAZy database, we find it most appropriate to keep them separate from the classical carbohydrate active enzymes.

We have now added the following sentence in the legend of Table 1 to clarify this point: “Although a number of oxidoreductases are listed in the CAZy database in the subcategory of Auxiliary Activities we have for the present study kept them separate. […] This separation also resolved the problem that some of the oxidoreductases listed in Table 1 are not covered by the CAZy database.” In addition we have inserted a note in the Oxidoreductase field in the table indicating that this category includes CAZymes with auxiliary activity.

3) In the Results and Discussion, it is stated that brown rot fungi rely on iron already present in the wood to employ the Fenton reaction. This idea may be somewhat outdated as XRF studies in Serpula lacrymans have shown that this brown rot fungus actively increases the iron concentrations in wood before decomposition. See Kirker et al., 2017. Synchrotron-based X-ray fluorescence microscopy enables multiscale spatial visualization of ions involved in fungal lignocellulose deconstruction. Scientific Reports 7, Article number: 41798.

We were not aware of this interesting publication and thank the reviewer for pointing it out to us. We have now rephrased the relevant sentence to include the findings of this paper (Results).

4) Discussion. I do not see how laccases in the fecal fluid can be considered functional analogs of brown rot recycling mechanisms of Fe(III) to Fe(II) directly. Laccases in the system described by Guillen et al., 2000 oxidise quinones. Reduced quinones in that system are responsible for reducing Fe(III) to Fe(II) and the production of H2O2. This sentence reads as if the laccases perform this function instead of the quinones.

In the system described by Guillen et al., 2000, laccase is needed to oxidize hydroquinone into semiquinone that is subsequently able to convert Fe3+ to Fe2+. In the manuscript sentence referred to, we did not mention that hydroquinone is used as a substrate in the attine symbiotic reaction as this information had already been given before (Discussion) and was not an essential part of the present argument. However, to avoid misunderstanding we have now changed the sentence to: “The high concentrations of laccase LgLcc1 (Protein ID 97 in Table 1) in the ant fecal fluid suggest a similar mechanism of maintaining reliable conversion of Fe3+ back to Fe2+ (using hydroquinone as an electron donor) to the one known from brown-rot fungi (Guillen et al., 2000)”.

5) Results. Not only enzymes can reduce Fe(III) to Fe(II) in brown rot fungi. The recycling of Fe(III) to Fe(II) in brown rot systems can be attributed to quinones as well.

As we have mentioned in a response to a previous comment above, it is believed – consistent with Guillen, 2000, – that in natural systems Fe3+ will always be bound by metal chelators, which makes Fe3+ unavailable to be directly reduced by hydroquinone. It thus needs laccase for this purpose. The affinity of the metal chelators will, however, depend on ambient pH, such that at very low pH it might be possible that Fe3+ can be directly reduced by hydroquinone. We prefer, however, not to go into this discussion, as the manuscript is complicated enough already.

https://doi.org/10.7554/eLife.61816.sa2

Article and author information

Author details

  1. Morten Schiøtt

    Centre for Social Evolution, Department of Biology, University of Copenhagen, Universitetsparken, Copenhagen, Denmark
    Contribution
    Conceptualization, Formal analysis, Investigation, Methodology, Writing - original draft, Writing - review and editing
    For correspondence
    mosch@aqua.dtu.dk
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4309-8090

  2. Jacobus J Boomsma

    Centre for Social Evolution, Department of Biology, University of Copenhagen, Universitetsparken, Copenhagen, Denmark
    Contribution
    Conceptualization, Funding acquisition, Writing - original draft, Project administration, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-3598-1609

Funding

Danmarks Grundforskningsfond (DNRF57)

  • Jacobus J Boomsma

H2020 European Research Council (Advanced Grant 323085)

  • Jacobus J Boomsma

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank the Smithsonian Tropical Research Institute (STRI), Panama, for providing logistic help and facilities during our field work in Gamboa.

Senior Editor

  1. Meredith C Schuman, University of Zurich, Switzerland

Reviewing Editor

  1. Dieter Ebert, University of Basel, Switzerland

Reviewers

  1. Maridel Fredericksen, University of Basel, Switzerland
  2. Heike Feldhaar, University of Bayreuth, Germany
  3. Michiel Op De Beeck

Publication history

  1. Received: August 5, 2020
  2. Accepted: January 9, 2021
  3. Accepted Manuscript published: January 12, 2021 (version 1)
  4. Version of Record published: February 11, 2021 (version 2)

Copyright

© 2021, Schiøtt and Boomsma

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 574
    Page views
  • 61
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Categories and tags

Further reading

    1. Ecology

    Plant-associated CO2 mediates long-distance host location and foraging behaviour of a root herbivore

    Carla CM Arce et al.
    Research Article

    Insect herbivores use different cues to locate host plants. The importance of CO2 in this context is not well understood. We manipulated CO2 perception in western corn rootworm (WCR) larvae through RNAi and studied how CO2 perception impacts their interaction with their host plant. The expression of a carbon dioxide receptor, DvvGr2, is specifically required for dose-dependent larval responses to CO2. Silencing CO2 perception or scrubbing plant-associated CO2 has no effect on the ability of WCR larvae to locate host plants at short distances (<9 cm), but impairs host location at greater distances. WCR larvae preferentially orient and prefer plants that grow in well-fertilized soils compared to plants that grow in nutrient-poor soils, a behaviour that has direct consequences for larval growth and depends on the ability of the larvae to perceive root-emitted CO2. This study unravels how CO2 can mediate plant–herbivore interactions by serving as a distance-dependent host location cue.

    1. Computational and Systems Biology
    2. Ecology

    Nutrient dominance governs the assembly of microbial communities in mixed nutrient environments

    Sylvie Estrela et al.
    Research Article

    A major open question in microbial community ecology is whether we can predict how the components of a diet collectively determine the taxonomic composition of microbial communities. Motivated by this challenge, we investigate whether communities assembled in pairs of nutrients can be predicted from those assembled in every single nutrient alone. We find that although the null, naturally additive model generally predicts well the family-level community composition, there exist systematic deviations from the additive predictions that reflect generic patterns of nutrient dominance at the family level. Pairs of more-similar nutrients (e.g. two sugars) are on average more additive than pairs of more dissimilar nutrients (one sugar–one organic acid). Furthermore, sugar–acid communities are generally more similar to the sugar than the acid community, which may be explained by family-level asymmetries in nutrient benefits. Overall, our results suggest that regularities in how nutrients interact may help predict community responses to dietary changes.

    1. Ecology

    Relative demographic susceptibility does not explain the extinction chronology of Sahul’s megafauna

    Corey JA Bradshaw et al.
    Research Article Updated

    The causes of Sahul’s megafauna extinctions remain uncertain, although several interacting factors were likely responsible. To examine the relative support for hypotheses regarding plausible ecological mechanisms underlying these extinctions, we constructed the first stochastic, age-structured models for 13 extinct megafauna species from five functional/taxonomic groups, as well as 8 extant species within these groups for comparison. Perturbing specific demographic rates individually, we tested which species were more demographically susceptible to extinction, and then compared these relative sensitivities to the fossil-derived extinction chronology. Our models show that the macropodiformes were the least demographically susceptible to extinction, followed by carnivores, monotremes, vombatiform herbivores, and large birds. Five of the eight extant species were as or more susceptible than the extinct species. There was no clear relationship between extinction susceptibility and the extinction chronology for any perturbation scenario, while body mass and generation length explained much of the variation in relative risk. Our results reveal that the actual mechanisms leading to the observed extinction chronology were unlikely related to variation in demographic susceptibility per se, but were possibly driven instead by finer-scale variation in climate change and/or human prey choice and relative hunting success.