Succession of physiological stages hallmarks the transcriptomic response of the fungus Aspergillus niger to lignocellulose

Abstract

Background

Understanding how fungi degrade lignocellulose is a cornerstone of improving renewables-based biotechnology, in particular for the production of hydrolytic enzymes. Considerable progress has been made in investigating fungal degradation during time-points where CAZyme expression peaks. However, a robust understanding of the fungal survival strategies over its life time on lignocellulose is thereby missed. Here we aimed to uncover the physiological responses of the biotechnological workhorse and enzyme producer Aspergillus niger over its life time to six substrates important for biofuel production.

Results

We analysed the response of A. niger to the feedstock Miscanthus and compared it with our previous study on wheat straw, alone or in combination with hydrothermal or ionic liquid feedstock pretreatments. Conserved (substrate-independent) metabolic responses as well as those affected by pretreatment and feedstock were identified via multivariate analysis of genome-wide transcriptomics combined with targeted transcript and protein analyses and mapping to a metabolic model. Initial exposure to all substrates increased fatty acid beta-oxidation and lipid metabolism transcripts. In a strain carrying a deletion of the ortholog of the Aspergillus nidulans fatty acid beta-oxidation transcriptional regulator farA, there was a reduction in expression of selected lignocellulose degradative CAZyme-encoding genes suggesting that beta-oxidation contributes to adaptation to lignocellulose. Mannan degradation expression was wheat straw feedstock-dependent and pectin degradation was higher on the untreated substrates. In the later life stages, known and novel secondary metabolite gene clusters were activated, which are of high interest due to their potential to synthesize bioactive compounds.

Conclusion

In this study, which includes the first transcriptional response of Aspergilli to Miscanthus, we highlighted that life time as well as substrate composition and structure (via variations in pretreatment and feedstock) influence the fungal responses to lignocellulose. We also demonstrated that the fungal response contains physiological stages that are conserved across substrates and are typically found outside of the conditions with high CAZyme expression, as exemplified by the stages that are dominated by lipid and secondary metabolism.

Background

Plant biomass is a renewable natural resource that is highly abundant, and is increasingly attracting attention for use of its components in the production of fuels, platform chemicals and materials [1,2,3]. Effective deconstruction of lignocellulose from dedicated feedstock crops such as Miscanthus, or agricultural waste such as sugar cane bagasse and wheat straw is becoming increasingly urgent as availability of fossil resources is decreasing [4]. Industrial scale use of plant biomass for biotechnologies is hampered by biomass recalcitrance [5, 6], whereas nature evolved effective microbial systems to recycle lignocellulose [7,8,9]. To fully exploit such natural deconstruction systems in biotechnology we first need a system-level understanding of microbial actions in response to lignocellulose.

In nature, saprophytic fungi are effective degraders of plant biomass [10,11,12]. Through tightly regulated responses to nutritional and environmental cues they can secrete a large array of hydrolytic enzymes and other effectors [7, 13,14,15,16,17]. Fungi such as Trichoderma reesei and Aspergillus niger are widely used in commercial production of these enzymes, due to their high titer secretion of degradative enzymes under industrial conditions [18, 19]. In addition A. niger has a firmly established role as industrial workhorse in the production of organic acids and other high value compounds [19]. It is therefore imperative to increase the understanding of the saprotrophic biology of these particular fungi.

In biotechnology, strategies to overcome substrate recalcitrance include engineering of enzyme cocktails and microbial strains for improved degradation as well as improving biomass digestibility via engineering of its composition or structure, or via chemical or physical pretreatment [6, 20,21,22]. Pretreatments, currently essential for effective lignocellulose processing, increase biomass digestibility via various mechanisms. Hydrothermal treatments remove hemicellulose components and break linkages between lignin and polysaccharides [23, 24] while ionic liquids solubilise wall components and generate, after recovery, insoluble but highly digestible polymers [3, 25]. However, even small differences in composition and structure of plant cell walls can significantly affect their digestibility [20, 26]. As ascomycetes have been described to strongly respond to differences in polysaccharides used as carbon source [27,28,29] and also exhibit partially conserved (substrate independent) responses to complex lignocellulose [30,31,32,33], here we investigated the effect of both hydrothermal and ionic liquid pretreated biomass as well as untreated feedstocks on the system-level response of A. niger.

The production of plant cell wall hydrolytic enzymes by A. niger is a highly orchestrated process. Initial responses of the fungi to biomass include alleviation of carbon catabolite repression and production of a set of scouting enzymes that scavenges for carbon sources [32, 34,35,36,37,38]. These then mediate release of small soluble carbohydrates from biomass, thereby inducing expression of an appropriate broad set of degradative enzymes via a network of transcriptional regulators [7, 14, 39,40,41,42]. Subsequent degradation of lignocellulose components can be expected to reflect the evolved fungal biochemical capacity [27, 43]. Fungal production of CAZymes and signalling responses to different sources of complex carbohydrates have been documented extensively, often capturing a snapshot of events at one or a selected few moments, such as when the CAZyme expression was peaking, in the cultivation time [28,29,30,31,32,33,34, 43,44,45,46]. Knowledge on the wider, and potentially time-staged, physiological responses and degradative strategies of A. niger is however still very limited.

To overcome this limitation, here we aimed to generate a systems level understanding of physiological responses of A. niger to lignocellulose. Previously, we found that the expression of A. niger CAZymes was determined by compositional changes in wheat straw generated by hydrothermal or ionic liquid pretreatments. We demonstrated how a high-resolution time series analysis captured a comprehensive view of the production of CAZymes in response to substrate compositional changes in wheat straw but did not examine other physiological responses [31]. Here we build on this approach to elucidate the effect of feedstock identity as well as pretreatment and cultivation time on the wider A. niger transcriptional landscape. We applied the high resolution time-staged transcriptome approach to cultivations with Miscanthus lignocellulose, both untreated and pretreated with ionic liquid or hydrothermally. Using this first dataset of the Aspergillus transcriptome on Miscanthus we furthermore aimed to uncover conserved fungal cell biological responses that are important to biomass degradation. The physiological responses to lignocellulose identified here will improve understanding of fungal degradative strategies, and potentially aid understanding on the requirements for development of Aspergillus strains for bio-processing.

Results and discussion

Aspergillus niger cultivation over time is affected by pretreatment of Miscanthus that influences its composition and digestibility

Aspergillus niger was exposed to knife-milled Miscanthus (KMM) as well as ionic liquid (IL) and hydrothermally (HT) pretreated Miscanthus for up to 5 days to identify the system-level fungal responses upon exposure to these substrates (Fig. 1a). Time-courses were terminated earlier where RNA of sufficient quantity and quality could no longer be obtained, i.e. after 24 h exposure of mycelium to HT pretreated Miscanthus (HTM) or 2 days to KMM, which suggested reduced fungal viability under these conditions.

Fig. 1

Schematic representation of experimental setup. aA. niger was cultivated in shake flasks with glucose as carbon source, after which mycelium was transferred to shake flask containing either knife milled Miscanthus (KMM), or KMM pretreated with ionic liquid (ILM) or hydrothermal treatment (HTM). Mycelium was harvested at indicated time points after which RNA was isolated and sequenced. This dataset was compared to data obtained with a similar experiment b using knife milled wheat straw (KMS) and pretreated straw (ILS, HTS) [31]. b is reproduced from [31] under a CC BY ((https://creativecommons.org/licenses/by/4.0/) license

HT pretreatment of Miscanthus reduced the arabinose and galactose content ≥ 10-fold and the xylose content twofold (Additional file 1: Fig. S1a), indicating loss of hemicellulose and pectin, while IL pretreatment reduced cellulose crystallinity (Additional file 1: Fig. S1b). As expected both pretreatments improved substrate saccharification with a standard commercial enzyme cocktail (Additional file 1: Fig. S1c). Fungal degradation of HTM appeared to be less than that achieved with the commercial enzyme saccharification. The xylan detected in the solid fraction retrieved from cultures of A. niger with KMM reduced by 26% over the first 24 h of cultivation, whereas in HTM it reduced by only 18%, suggesting less degradation of HTM than KMM during cultivations (Additional file 1: Fig. S1d). In contrast, xylan content of ILM decreased by 71% after 24 h cultivation, indicating that its increased digestibility in enzymatic saccharification also occurred in the fungal cultivations. Levels of monosaccharides in culture filtrates support increased ILM digestibility in fungal cultivations compared to KMM. In the 9-24 h time points of exposure to ILM, up to 2.2 mM of xylose and arabinose was detected (Additional file 1: Fig. S1e), whereas monosaccharide concentrations in other conditions were < 100 μM. This suggests high sugar release via effective ILM digestion and accumulation to levels reported to be high enough [47] to activate carbon catabolite repression.

The overall effect of HT and IL pretreatment on Miscanthus feedstock composition and digestibility, and its subsequent degradation in fungal cultures, was similar to our wheat straw study [31]. However, feedstock identity affected the end result of pretreatments and subsequent fungal cultivation; Miscanthus was less digestible than knife milled wheat straw (KMS) and pretreatments only partly remedied this. For example, saccharification with commercial enzymes released fourfold less sugar from KMM than from KMS but twofold less from HTM than from HT pretreated straw (HTS), with no difference for ILM and IL pretreated straw (ILS).

Overview of gene expression during cultivation on complex substrates

We investigated the transcriptional landscape of A. niger to untreated and pretreated Miscanthus to elucidate the effect of feedstock and pretreatment on gene expression, and to identify the system-level fungal responses to cultivation. We analysed the data together with that of cultivations on untreated and pretreated wheat straw from our previous study [31] (Fig. 1b), where the broader physiological responses beyond CAZymes had not yet been analysed. This resulted in a data matrix of 47 conditions of A. niger cultured with a particular (pre- or untreated) substrate (Fig. 1).

Hierarchical clustering of the transcript levels of all genes (Fig. 2a), showed that pretreatment had the most discriminatory effect, as demonstrated by clustering of selected IL pretreated conditions with glucose-grown conditions (Cluster 1, C1, indicated in blue). The second most influential contribution to differences between samples was cultivation time; early (3-6 h) time points clustered together, as do 9–12 h conditions, and together these group (C3, red) separately from later time points (C2, green). Surprisingly, lignocellulose feedstock identity (i.e. Miscanthus or wheat straw) had an overall minor effect on clustering pattern. Subsequent PCA confirmed separation of conditions on the first to third principal component (Fig. 2b, c) corresponded mainly to effects of pretreatment and cultivation time, with follow-on detailed analysis revealing feedstock effects.

Fig. 2

Hierarchical clustering and PCA analysis illustrate time, pretreatment and substrate dependant clustering patterns. With a hierarchical clustering of all conditions using the log transformed and quartile normalised mean FPKM values that were ≥ 1 in any of the time points on any media, b, c results from PCA showing separation in distinct clusters over the first 3 components in 2D and 3D respectively. Ellipses indicate 95% confidence intervals. Clusters that are discussed in the main text are labelled C1–C3 and indicated with coloured panels, with dashed lines representing sub-clusters. d Summarises relative abundance of GO-terms identified as linked to the clusters via word clouds

The genes and processes underpinning the observed variation in gene expression between sample groups were identified via analysis of the PCA loadings, (which identifies genes whose expression levels contributed highest to the sample group distribution) via gene transcription patterns and GO-term enrichments. Samples from cultures grown on glucose and those cultivated on IL pretreated substrates for 9 h, 12 h and 24 h, cluster together (C1) on PCA1 and PCA2. Genes with highest loadings for this group, whose expression levels contributed highest to the clustering, were enriched for those annotated with GO-terms for respiration, e.g. mitochondrial function including transport, organisation, ATP metabolism, as well as biosynthesis of compounds as amino acids and organic acids. IL 9 h and 12 h time points distinguished themselves (Fig. 2b, c) from the glucose grown samples via enrichment of GO-terms associated with translation. Identified GO-terms are summarised in word clouds (Fig. 2d).

The 3 h to 12 h time points on lignocellulose (except ILM and ILS 9-12 h) formed cluster C3. Genes with the highest loadings were enriched for those annotated with GO-terms for carbohydrate metabolism, cellular respiration, protein transport and peroxisome function or organisation. Closer inspection revealed that 3 h and 6 h time points on lignocellulose formed a subcluster C3a (Fig. 2b, indicated in dashed red), which largely consisted of the genes with the highest loading enriched for those annotated with GO-terms for fatty acid beta-oxidation, peroxisome organisation, cell division and cytoskeleton (re)organisation.

Lignocellulose samples at 24 h or later (except ILM and ILS 24 h) clustered together (Cluster C2). Genes with the highest (≥ 0.1) loadings were enriched for those annotated with GO-terms for catabolism of polysaccharides, aromatic amino acids and other aromatic compounds, the biosynthesis of secondary metabolites as terpenoids and mycotoxins, modification of fungal cell walls, chromosome organisation and regulation of metabolic processes. This coincides with enrichment of oxidoreductase and glycoside hydrolase activity associated terms.

Initial response to lignocellulose involved increased fungal lipid metabolism

The initial transcriptional response to lignocellulose was conserved over feedstocks and pretreatment, as demonstrated by clustering of the early (3–6 h) time points in PCA and hierarchical clustering. For genes with reduced transcript levels, GO terms for RNA helicase activity were enriched at the 3 h time-point (Additional file 2: Table S1), as well as those for RNA processing—specifically processing and maturation of non-coding and ribosomal rRNA—and ribosomal subunit biogenesis, assembly and localisation, together suggesting a conserved reduction in overall translational activity upon transfer to lignocellulose. For genes with increased transcript levels, enriched GO-terms (Additional file 3: Table S2) strongly represented lipid catabolism, specifically fatty acid beta-oxidation and catabolism of the short chain fatty acid acetate (Fig. 3a). This is in accordance with the results from the PCA. In addition, the GO-term for branched-chain amino acid breakdown was enriched at the 3 h time points, which could be caused by breakdown products of valine and isoleucine which are processed via fatty acid beta-oxidation [48]. Metabolic mapping highlighted significant enrichment of the butanoate metabolic pathway, a short chain fatty acid, in the early time points, as well as gluconeogenesis pathway enrichment (Additional file 4: Table S3).

Fig. 3

Lipid metabolism as general response in early time points. With a enrichment of selected GO-terms in early time points, b transcript levels of genes involved in fatty acid β-oxidation and acetate metabolism, as detected by RNASeq, c cbhA, GH61a (An12g04610), GH5 (An01g11670) and xynB transcript levels in the wild-type (black) and farA (grey) deletion strain grown on glucose or wheat straw (9 and 24 h) as detected by qRT-PCR, mean indicated by bar, results from statistical testing is indicated by ns (p > 0.05), * (p < 0.05), ** (p < 0.01) or *** (p < 0.001)

Fatty acid metabolism has been investigated most thoroughly in A. nidulans, and many orthologous genes are present in A. niger. Short chain fatty acids are oxidized to produce acetyl-CoA in the mitochondrion by, among others, ScdA and EchA, while long chain fatty acids are oxidised in the peroxisome by FaaB and FoxA [48, 49]. Transcript levels of corresponding A. niger genes (Fig. 3b) showed a peak at 3 h, indicating that both short-chain and long-chain fatty acid beta-oxidation is activated. Acetyl-CoA is metabolised via the tricarboxylic acid (TCA) cycle in the mitochondrion. TCA cycle intermediates are used for gluconeogenesis, and are replenished in the peroxisome by AcuD and AcuE. To enable acetyl-CoA shuttling between cell compartments, carnitine acetyltransferases FacC, AcuJ [50, 51] interconvert acetyl-CoA and acetylcarnitine, which can cross membranes using transporters such as AcuH. The A. niger genes encoding these proteins, as well as An11g02590.Aspni5_55954, encoding PexK required for peroxisome biogenesis during fatty acid β-oxidation [52, 53], all show a similar expression profile with a peak at 3 h (Fig. 3b). Genes specific for acetate metabolism, facC [54], and acuA (ortholog of A. nidulans facA), also show a peak in transcript levels at the 3 h (Fig. 3b) indicating that the acetate metabolism pathway is specifically activated during early exposure to complex lignocellulose substrates.

Taken together these data strongly suggest that directly upon exposure to lignocellulose, fatty acid beta-oxidation and subsequent metabolism is increased. During the carbon-starved conditions experienced at that point, this could provide the fungus with a source of energy and result in the generation of glucose via gluconeogenesis. The response is conserved over all substrates analysed here, indicating a universal process consistent with the carbon starvation experience. In line with this, enrichment of almost all of these GO-terms was also observed in the 5 days glucose culture where glucose is depleted and increased transcription of genes linked to fatty acid metabolism in response to carbon starvation has been reported for Aspergilli [16, 55] as well as in other fungi [56, 57]. A role for carbon starvation in induction of plant polysaccharide degradative enzymes was previously established in A. niger [34, 38], A. nidulans [36, 37, 58] and N. crassa [28, 59], which involved low-level upregulation of a subset of genes encoding CAZymes active on plant polysaccharides, the ‘scouting enzymes’. Besides generating energy for cell maintenance, fatty acid metabolism may generate energy required for production of these scouting enzymes or assist in establishing fungal colonisation of lignocellulose. To test this hypothesis, we analysed the role of fatty acid metabolism in hydrolase production via deletion mutants of the ortholog of the A. nidulans gene farA, encoding the inducible regulatory activator responsive to both short- and long-chain fatty acids FarA [60].

The ortholog of A. nidulans FarA contributes to timely establishment of hydrolytic capacity

The expression of farA strongly increased at 3 and 6 h on all lignocellulose substrates, and in addition at 9 h for KMM (Fig. 3b). As this gene is the ortholog of the confirmed regulatory activator of fatty acid metabolism in A. nidulans, it was an ideal candidate to assess an influence of fatty acid metabolism in establishing hydrolytic capacity via farA deletion.

Gene expression in two independently generated farA deletion mutants and its wild-type parent strain were analysed by qRT-PCR after exposure to lignocellulose. After 9 h, the farA deletion strains had significantly (p < 0.05) lower transcript levels for the cellulolytic enzyme encoding genes cbhA and GH61a (An12g04610): compared to the wild-type, mean levels of transcripts were 10-fold and twofold less induced respectively (Fig. 3c). After 24 h transcription levels of these genes were more similar with < twofold difference. No pronounced differences (p > 0.05 and/or < twofold change) were found for transcription of xynA or xynB at 9 h or 24 h, while transcript levels of GH5 An01g11670 were similar in both strains at 9 h but were increased threefold in expression in the farA deletion strain after 24 h. These results indicate that the initial substrate-independent lipid metabolism via FarA affects the subsequent induction of a subset of genes encoding CAZymes during exposure to lignocellulose. The results also suggest the mechanism may not be via a direct effect on scouting enzyme transcript levels in early time points, as no difference for initial transcription levels of a GH5 scouting enzyme was detected at 9 h. An indirect effect would be consistent with the lack of FarA binding sites (CCTCGG [60]) in the 1 kbp upstream promoter regions of the cbhA and GH61a genes. Rather, the effect could be through translational effects, or cumulative effects that eventually result in a delayed fungal lignocellulose colonisation and thus delayed induction of genes that encode enzymes routinely observed only under full induction on this substrate [34, 38], such as cbhA and GH61a.

Pretreatment and feedstock effects dominate gene expression in middle-stage cultures on lignocellulose

Our initial multivariate analysis indicated that the 9–12 h time points on all substrates represent established (middle stage) cultures on lignocellulose with the most active carbohydrate degradation and metabolism, cellular respiration and protein transport (Fig. 2). PCA analysis of gene expression levels indicated at least 27% of variation in gene expression in these time points resulted from differences between IL versus other cultures, with ≥ 9% of variation explained by HT versus other cultures and ≥ 5% by feedstock (Fig. 4a).

Fig. 4

Feedstock and pretreatment effects in established cultures on Miscanthus include nitrate assimilation and CAZyme expression. a PCA analysis of variation in gene expression in 9 h and 12 h Miscanthus and straw cultivations, demonstrating pretreatment (PC1, PC2) and feedstock (PC4) effects. Ellipses indicate 95% confidence intervals. b Expression of nitrate assimilation genes niaD and niiA identified via RNA-Seq, values are mean ± st. error, c Venn diagram illustrating overlaps in total number of induced plant polysaccharide- active CAZyme encoding genes during exposure to Miscanthus and wheat straw, d and between induced genes on untreated and pretreated Miscanthus substrates, e total expression, as sum of FPKM, for genes encoding hydrolytic enzyme activities over the 9 h and 12 h Miscanthus time courses. See Additional file 7: Table S5 for underlying data

Genes with positive loadings in the PCA first component that distinguishes IL pretreatment from the other conditions, revealed high expression of a cluster of genes encoding proteins responsible for nitrate uptake and intracellular conversion to ammonium, including nitrate reductase NiaD and nitrite reductase NiiA [61]. High expression of this gene cluster was shared with the glucose cultures and with early KMS time points (Fig. 4b). When using nitrogen assimilation activity as a proxy for growth [62], this indicated fungal growth on IL pretreated substrates from 3 to 24 h, reflecting high digestibility of the IL pretreated substrates (Additional file 1: Fig. S1). Genes associated with HT pretreatment, with negative loadings in the second principle component, were enriched in GO-terms for phenylalanine catabolism and oxidoreductase activity.

Genes linked to complex carbohydrate degradation had high loadings on the first principle component (IL pretreatment) and the fourth principle component, separating Miscanthus and wheat straw feedstocks. We therefore compared expression of genes encoding plant-polysaccharide degrading CAZymes that showed at least in one of the time-points a high-level induction compared to the glucose preculture (i.e. with a DESeq p_adj < 0.05, FPKM of ≥ 50 on lignocellulose and log2 FC of ≥ 3) (Fig. 4c). The sum of transcript levels of genes encoding plant CAZymes formed a dome-shaped curve during exposure to lignocellulose (Additional file 5: Fig. S2b), with a dip in expression levels during the 12-24 h time points in ILM cultures similar as reported for ILS cultures [31], as high IL-pretreated substrate digestibility resulted in accumulation of xylose to levels responsible [47] for carbon catabolite repression (Additional file 1: Fig S1). While the overall number of induced CAZyme-encoding genes was very similar on wheat straw and Miscanthus, the total expression level on KMM was reduced compared to KMS (Additional file 5: Fig. S2, Additional file 6: Table S4).

Over the time-course on Miscanthus, 70 CAZyme-encoding genes were induced (Fig. 4c, Additional file 7: Table S5). A core set of 51 genes encoding CAZymes with (putative) plant-polysaccharide degrading activity were induced both in cultures grown with untreated and pretreated Miscanthus (Fig. 4d). We detected nine genes that were specifically induced on KMM, including six pectinolytic genes. ILM specific induced genes included the α-galactosidase encoding aglB, while the α-xylosidase encoding axlA was ILM and HTM specific.

We compared the expression levels of the plant encoding CAZy genes induced at the 9 h and 12 h time points as these established cultures showed minimal transcriptional repression by CCR and overall high CAZyme expression levels. Considerably higher average expression levels were found after exposure to ILM and/or HTM when compared to KMM, with induction between 5- and 80-fold (Additional file 7: Table S5). Investigating the induced activities (Fig. 4e), a broad set of genes encoding hemicellulosic and cellulosic enzymes had much higher transcript abundance on pretreated substrates when compared to KMM, with ILM displaying the highest transcript abundance of genes encoding hemicellulosic enzymes and HTM the highest transcript abundance of genes encoding cellulosic enzymes. KMM on the other hand displayed the overall highest transcript abundance of genes encoding pectinolytic enzymes, in line with the number of pectinase-encoding genes expressed in this condition. Confirmation that these transcript abundance differences also translated into relative protein abundance and thus degradative capacity was obtained by targeted proteomics. Relative abundance of xylanolytic enzymes XynA and XynB, but not XlnD was found to be higher in culture filtrates of ILM and HTM compared to KMM (Additional file 8: Fig. S3). Cellulolytic CbhA, EglA and an AA9 putative LPMO, but not CbhB and EglC, were more abundant in filtrates of HTM than of KMM cultures. Pectinase PelA was most abundant in filtrates from KMM cultures (Additional file 8: Fig. S3).

The relatively abundant pectinase production on untreated Miscanthus was also observed on wheat straw (KMS) [31] and may be related to the reduction of pectin content by the pretreatments, and resultant lack of inducers for pectinase encoding genes [41]. HT-dependent higher production of cellulases could be explained by the relatively higher cellulose content of this substrate and its increased digestibility compared to KMM.

Feedstock-specific enrichment for mannan degradation

Comparing CAZy gene expression over the time course of cultivations on both wheat straw and Miscanthus feedstocks (Fig. 4c), mannan degradation was found to be feedstock-specific. Enrichment of its GO-term (GO:0046355) was specific to the untreated and pretreated wheat straw cultures. Four of five genes annotated with the GO term for mannan degradation were responsible for this enrichment (Additional file 3: Table S2). When looking at the expression profile of these genes in detail, induction of transcription of β-mannosidase mndA was straw specific, while endo-β-mannanase encoding man5A had much higher transcript levels on wheat straw, especially KMS and HTS, compared to Miscanthus (Fig. 5a). Targeted proteomics showed a similar trend for the abundance of this protein in the culture filtrate (Fig. 5b) and endo-mannanase activity was detected at higher levels in wheat straw culture filtrates compared to the Miscanthus culture filtrates (Fig. 5c). These results confirm the up-regulation of endo-β-mannanase activity is feedstock-specific. Three out of four of the genes responsible for enrichment of the GO term are inducible by mannose or mannan [41]. Mannan content of wheat straw and Miscanthus is very low [63] and perhaps instead increased accessibility of the mannan in the wheat straw could be a contributory factor to the observed distinct activation of mannan degradation. Mannan-derived compounds such as mannobiose may induce genes encoding mannan-degrading enzymes via ClrB [64], the ortholog of Aspergillus oryzae ManR, A. nidulans ClrB and Neurospora crassa CLR-2. As the level of transcriptional response of this regulator to inducers does not seem to be conserved between fungal species and the regulatory pathway overlaps with that of cellulose degradation [13, 59, 65, 66], it is not straightforward to link A. niger clrB transcription to expression of its regulatory targets and we have in our experimental conditions not identified such a correlation.

Fig. 5

Mannan degradation is feedstock specific. a gene expression levels of the endo-β-mannanase encoding gene An05g01320.Aspni5_50378, b protein levels of the endo-β-mannanase in the culture filtrate and c endo-mannanase activity detected on AZCL-galactomannan. Error bars represent standard errors

Transition to starvation and sporulation

The high-level transcription of plant biomass degrading CAZy genes was not sustained across the time-course. A feedstock independent transition to reduced enzyme expression and activation of starvation, sporulation and secondary metabolism occurred between 12 and 24 h (KM and HT) or 24 h–2 days (IL), as indicated by hierarchical clustering and PCA (Fig. 2).

The transcriptomic data indicated limited growth and primary metabolic activity at the later time points. Expression of nitrogen assimilation related genes such as niiA and niaD was limited (Fig. 4b) and expression of the genes encoding key biosynthesis enzymes for the fungal cell membrane component ergosterol (erg2, erg3, erg11 and erg25) was considerably reduced (Fig. 6a). In agreement, steroid biosynthesis (KEGG rn00100) was enriched in the metabolic mapping in pre-, but not post-24 h time points. As ergosterol is a marker for fungal biomass [67] the depletion of transcripts suggests limited fungal growth after the transition point. During late (≥ 24 h) time points GO-terms suggested a low metabolic activity: terms enriched in genes with reduced transcript levels included aerobic respiration, specifically citric acid (TCA) cycle and acetyl coenzyme A degradation, oxidative phosphorylation and ATP synthesis (Fig. 6b).

Fig. 6

Late time point cultures are starving and induce secondary metabolism. a Transcript levels of genes encoding ergosterol biosynthesis enzymes, b enrichment of selected GO-term in late time points, c transcript levels of genes encoding starvation-induced fungal cell wall acting hydrolases, d protein level of corresponding fungal cell wall acting hydrolases in the culture filtrates, e transcript levels of genes encoding a sporulation specific transcriptional regulator and a protein involved in spore pigmentation and f transcript levels of genes encoding the putative PKS of a sporulation specific secondary metabolite cluster (left), and the NRPS of the gliotoxin cluster (right), with 2 and 7 respectively of the other cluster genes sharing this profile. Error bars represent standard errors

Indications for starvation and stress in the late time point cultures were found through enrichment of the following GO-terms among genes with increased transcript levels: retrograde vesicle recycling in the Golgi, (non-glycolytic) fermentation and ionic osmotic (saline) stress responses, cell death, DNA modification, lipid metabolism and fungal cell wall acting enzymes (Fig. 6b). Starvation and stress occurred later in the IL pretreated compared to HT or untreated cultures, consistent with ILS and ILM having increased digestibility. Fungal cell wall-acting CAZymes encoding genes that are responsive to carbon starvation, such as cfcA/chiB and engA have high transcript levels from 12 h (HT, KMM), 24 h (KMS) or 2 d (IL) (Fig. 6c) with corresponding accumulation of proteins (Fig. 6d). Starting from 24 h (HT, KM) or 2 days (IL) cultures, transcript levels were strongly increased for genes encoding sporulation transcriptional activators (brlA, abaA), pigment synthetic enzymes (fwnA, brnA, olvA), hydrophobins (predominantly hfbA, hfbB, hyp1) and sporulation-specific cell wall modifying CAZymes (ctcB, bgtD), indicating sporulation was taking place (Fig. 6e). GO-terms for breakdown of aromatic amino acids (phenylalanine, tyrosine, tryptophan) were enriched at most of the time-points without high monomeric sugar levels. These GO-terms share common genes that belong to the tyrosine-phenylalanine metabolism gene cluster. Transcription of these genes was co-regulated and was most abundant on later (≥ 24 h) time points. Together this indicates that the late time points represent carbon-limited fungal cultures with limited growth, which activated survival strategies such as sporulation and secondary metabolism.

Novel and known secondary metabolite gene clusters induced at later time points

Fungal secondary metabolites attract high interest due to their potential bioactive properties [68]. A. niger contains many secondary metabolite gene clusters [68, 69] of which for many their products are so far unknown. The production of secondary metabolites has been reported in some instances to be linked to starvation and sporulation but for a large number of metabolite clusters no conditions have been identified under which they are expressed. Our lignocellulose time-courses represent a range of novel conditions that could stimulate the fungus to generate secondary metabolites and we therefore assessed expression of secondary metabolite generating gene clusters predicted in the A. niger genome (Additional file 9: Table S6).

The GO-term for secondary metabolism was enriched most notably in the ILM and ILS time courses (Fig. 6b). Eight genes from the gene cluster encoding proteins for production of the toxin fumonisin (An01g06810-An01g06970), although not the PKS-encoding gene, were strongly induced on the 2 d ILS cultures and to a more moderate level in 2 d ILM (Additional file 9: Table S6). Seven genes from cluster An09g00520-An09g00620, including the gene with strong similarity to gliC of the A. fumigatus gliotoxin synthesis gene cluster [70], showed a pronounced increase in transcription levels at the 9 h KMS and ILM 2 days time points, as well as some increases on ILS 2 days (Fig. 6f). This peculiar transcript profile was shared with two genes (An01g06150 and An08g06810) involved in the shikimate pathway, which generates chorismate that feeds into biosynthesis of aromatic amino acids such as phenylalanine, one of the two amino acids that are condensed to produce gliotoxin. Metabolic mapping also pointed towards enrichment of the shikimate pathway (part of KEGG pathway rn00400). The secondary metabolite cluster An11g06420-An11g06490 showed strong induction from IL 2 and 3 days, as well as some expression on later IL time points and on KMS from 24 h onwards. This cluster was also induced in A. niger in mixed-species cultures on wheat straw [71]. An orthologous hybrid PKS/NRPS is required for production of pseurotin A in A. fumigatus, and is induced in response to hypoxia and the anti-fungal drug voriconazole [69, 72, 73]. The uncharacterised NRPS encoding An13g03040 as well as seven genes of this cluster expressed specifically in the 2 days IL and 2 d KMS conditions. Expression of the uncharacterised cluster An12g07060-An12g07110 [74] increased to considerable levels on late cultivation time points (Fig. 6f). The transcription profile of this cluster’s PKS-encoding gene correlated with that encoding the FwnA PKS involved in production of spore pigmentation, as did backbone synthase genes from five other gene clusters (Pearson correlation coefficient 0.78 ± 0.13 (mean ± stdev)). This suggests their transcription may be sporulation-related, which is consistent with conditions in which Pyranonigrin A has been detected [75], the antioxidant generated by the only characterised gene cluster of this group. Overall, the transcriptomic data indicated activation of the production of potentially new as well as known secondary metabolites, confirming that cultivation with plant biomass can provide unique conditions for the induction of secondary metabolite gene clusters in fungi, which can be further evaluated as candidates for exploitation in biotechnology. In the natural habitat of A. niger, such metabolites may function in interaction with other organisms [76], as e.g. signalling molecule or antibiotic, or provide protection against harmful conditions (e.g. antioxidants), and thereby contribute to fungal survival.

Conclusions

The transcriptional response of A. niger to lignocellulose adapted over the cultivation time-course, and could overall be divided into three stages: an initial (3–6 h) conserved response, followed by a substrate and pretreatment-specific response (generally 9–24 h) and a final largely conserved response (generally 24 h–5 days) albeit with variable elements, as summarised in Fig. 7.

Fig. 7

Illustration of physiological stages of the A. niger response to (pretreated) substrates including both conserved and substrate-specific responses. After transfer to lignocellulose, A. niger reacted with a general carbon starvation response, followed by pretreatment and feedstock specific elements such as mannan degradation on straw but not Miscanthus, followed by a general carbon-limitation induced response with sporulation and induction of secondary metabolite clusters. Stand-out features of the response are indicated for each condition. Green and red arrows represent processes that were increased or decreased, respectively. CAZy transcript levels and pentose concentration are represented as average (for all or stated conditions) percentage of the maximum with identical y-axes in all instances. The width of blue and grey panels represent the covered time-period, marked on the time-scale as h after transfer to lignocellulose

Initially fatty-acid β-oxidation was activated and a role in timely establishment of lignocellulose degradative capacity was identified. In nature, this response may be part of a mechanism contributing to fungal survival in a habitat with a highly fluctuating and often limited availability of carbon sources. Subsequently a substrate hydrolytic stage was activated that was mainly directed by substrate properties such as composition and digestibility. Subsequently, sporulation was activated, and the transcriptome data indicated this was in response to reduced metabolic activity and carbon limitation in the culture. This study emphasises the value of high-resolution time-courses when studying fungal responses to lignocellulose, to ensure confidence in identification of all involved enzymes that may be expressed, and importantly, to understand more broadly the range and dynamics of the fungal physiology on a systems level. Ultimately, improved understanding of these responses should improve the efficiency of strain and enzyme developments for renewables-based biotechnology.

We identified a range of secondary metabolite gene clusters that were activated during exposure to lignocellulose, including multiple clusters so far without product characterisation. Secondary metabolite gene clusters attract significant interest for generating novel potentially bioactive compounds and as a source of biocatalytic enzymes. We are therefore confident that the expression conditions for these novel clusters identified here will receive interest by providing opportunities for their further characterisation and exploitation.

Methods

Substrates and pretreatments

Miscanthus (cv. Giganteus) was ground to pass an 0.5 mm sieve as described previously [31]. The ionic liquid (IL) pretreatment was performed using 1-ethyl-3-methylimidazolium acetate ([C2mim][OAc]) for 2 h at 160 °C and hydrothermal treatment was performed at 200 °C for 2 h, as described in detail previously for wheat straw [31]. The compositional and structural analysis of substrates was performed as described in detail previously for wheat straw [31], employing total acid hydrolysis assay followed by analysis of soluble sugar monomers in the hydrolysate by high-performance anion exchange chromatography with pulsed amperometric detection (HPAEC-PAD), and wide-angle X-ray diffraction. Digestibility of the substrates was assessed in saccharification assays with a commercial enzyme mixture (Celluclast (Sigma) and Novozyme 188 (Sigma), 1:1 v/v), as well as by measurement of residual xylose in solids of shake-flask cultures during fungal exposure [31].

Fungal cultivation conditions

Cultivations were performed as described before [31]. Briefly, 1% w/v glucose cultures were inoculated with A. niger N402 spores at a final concentration of 1 × 10⁶ ml⁻¹ and incubated at 28 °C for 48 h. The mycelia were collected and 1.5 g wet weight was transferred to flasks containing one of the following carbon sources at a 1% w/v final concentration in AMM: glucose, untreated Miscanthus, ionic liquid pretreated Miscanthus or hydrothermally pretreated Miscanthus. The cultures were incubated for 3 h, 6 h, 9 h, 12 h, 24 h, 2 d, 3 d, 4 d or 5 d. Mycelia were collected for RNA extraction and the filtrate of the culture was flash frozen in liquid nitrogen and stored at − 80 °C for subsequent sugar and targeted proteomics analyses, both performed as described [31].

Gene expression analysis

RNA extraction, Illumina paired-end sequencing, read mapping to the A. niger ATCC 1015 strain and statistical analysis were all performed as described in detail before using a DESeq 2 p_adj value < 0.05 to identify differentially expressed genes between pairs of conditions [31]. The obtained data were analysed together with those obtained from cultivation with untreated and pretreated wheat straw [31] under identical conditions. Additional file 10: Table S7 contains the FPKM values for each sample with Miscanthus as feedstock as well as the DESeq 2 analysis for the comparison of each condition with the Glu 48 h control. Clustering and Gene Ontology (GO) enrichment analysis was performed as described before [31] using the combined dataset including wheat straw and Miscanthus derived samples. In short, hierarchical clustering of log transformed gene expression values (FPKM) of all 47 conditions with R package gplots. GO enrichment analysis was performed using the Bioconductor package GOseq (v1.18.0) [77] using GO annotations from the FetGOat resource (http://www.broadinstitute.org/fetgoat/) [78], and comparing to the Glu 48 h control for each condition the genes with a log2 fold change (i.e. DESeq 2 of ≤ − 1 or ≥ 1), p_adj ≤ 0.05 and an FPKM ≥ 1 in at least 1 condition. PCA analysis was performed using the MetaboAnalyst package [79, 80], using log transformed, mean centered FPKM values of triplicate conditions. Genes with high calculated loadings are analysed for GO-enrichment using g:Profiler [81], results were summarised using REVIGO [82] and subsequently visualised using https://www.jasondavies.com/wordcloud/ to generate word clouds.

Mapping transcript level data to a genome-wide metabolic model

In order to analyse gene expression data on a metabolic network we used a CoReCo created genome wide metabolic model of A. niger [83]. For each enzyme commission (EC) number in the model an arbitrary enzyme expression value was created by taking the expression value of the responsible gene or taking the highest value when multiple genes were mapped to one EC number. The resulting EC expression data matrix was then explored using principal component analysis and Bayesian Hierarchical Clustering [84], followed by mapping of the EC cluster content to pre-defined pathways in KEGG [85] and Metacyc [86] databases. Enrichment of EC cluster members on predefined pathways was quantified with the hypergeometric test.

Analysis of gene expression in farA deletion strain

The wild-type strain N402 and two independent farA deletion strains (MA537.3 and MA537.4, kindly provided by Dr A. Ram, University of Leiden, the Netherlands) were grown in liquid cultures in triplicate as described above using 1% ball milled wheat straw as substrate. After 9 h and 24 h, mycelium was harvested and RNA was isolated as described above. qRT-PCR was performed as described previously [31] but with PCR products as a standard, using 0.5 µl of cDNA and the primers specified in Additional file 11: Table S8. Gene expression values are normalised via the expression of actA and sarA, two genes that were previously reported to be stable reference genes suitable for data normalisation in qRT-PCR experiments [87, 88]. Statistically significance of differences in gene expression was tested with a one-way ANOVA with post hoc multiple comparisons test.

Enzyme assays

For the enzyme assays with AZCL dyed galactomannan (I-AZGMA, Megazyme), appropriate volumes of the concentrated culture filtrates were incubated for 3 h at 28 °C with 150 rpm shaking with 1% w/v concentration of the dyed substrate in 100 mM sodium acetate pH 4.5 in a total volume of 250 μl in 96 well plates. Reactions were stopped with 2% Trizma base as per manufacturer’s instructions and absorbance was measured at 590 nm with a plate reader.