The carbon starvation response of Aspergillus niger during submerged cultivation: Insights from the transcriptome and secretome

Background Filamentous fungi are confronted with changes and limitations of their carbon source during growth in their natural habitats and during industrial applications. To survive life-threatening starvation conditions, carbon from endogenous resources becomes mobilized to fuel maintenance and self-propagation. Key to understand the underlying cellular processes is the system-wide analysis of fungal starvation responses in a temporal and spatial resolution. The knowledge deduced is important for the development of optimized industrial production processes. Results This study describes the physiological, morphological and genome-wide transcriptional changes caused by prolonged carbon starvation during submerged batch cultivation of the filamentous fungus Aspergillus niger. Bioreactor cultivation supported highly reproducible growth conditions and monitoring of physiological parameters. Changes in hyphal growth and morphology were analyzed at distinct cultivation phases using automated image analysis. The Affymetrix GeneChip platform was used to establish genome-wide transcriptional profiles for three selected time points during prolonged carbon starvation. Compared to the exponential growth transcriptome, about 50% (7,292) of all genes displayed differential gene expression during at least one of the starvation time points. Enrichment analysis of Gene Ontology, Pfam domain and KEGG pathway annotations uncovered autophagy and asexual reproduction as major global transcriptional trends. Induced transcription of genes encoding hydrolytic enzymes was accompanied by increased secretion of hydrolases including chitinases, glucanases, proteases and phospholipases as identified by mass spectrometry. Conclusions This study is the first system-wide analysis of the carbon starvation response in a filamentous fungus. Morphological, transcriptomic and secretomic analyses identified key events important for fungal survival and their chronology. The dataset obtained forms a comprehensive framework for further elucidation of the interrelation and interplay of the individual cellular events involved.

http://www.biomedcentral.com/1471-2164/13/380 [1], suggesting that the regulation of asexual development is conserved. The first step in conidiophore development is the activation of the transcriptional regulator BrlA, which induces the expression of a number of conidiationspecific genes. BrlA expression is autoregulated, resulting in a strong accumulation of its mRNA during asexual development [8]. Although most conidiation studies are performed at a substrate/air interface, conidiation can also be induced in submerged cultures by nutrient limitation such as severe carbon limitation [9][10][11]. Under these conditions, carbon from endogenous resources becomes mobilized to fuel maintenance and self-propagation. Consequently, the fungal mycelium becomes highly heterogeneous, bearing empty compartments and those that are committed to conidiation [11,12]. While this strategy is beneficial for self-propagation and the exploitation of new substrate sources during saprophytic growth, it may result in a decrease of the active biomass fraction during carbon-limited industrial production processes.
Only a few studies have been conducted to investigate different aspects of aging carbon-limited fungal cultures. As discussed in the review by White et al. [12], most of them focus on physiological and morphological aspects. The generic term autolysis has been frequently used to summarize the involved processes. Hallmarks of autolysis are biomass decline, hyphal fragmentation, release of ammonia and increased extracellular hydrolase activity [12]. For different fungal species, the involvement of hydrolases, especially chitinases and glucanases but also proteases has been investigated in great detail [13][14][15]. An early and strong transcriptional induction in response to carbon starvation was shown in A. nidualns for the two hydrolases ChiB and NagA, which have been intensively studied because of their role in the degradation of the cell wall component chitin [16]. In addition to physiological and biochemical hallmarks of aging fungal cultures, several approaches have been developed to quantify the decreasing fraction of active hyphal compartments in aging mycelium by (semi-) automated image analysis [12].
An increasing number of publications highlights the importance of programmed cell death (PCD) in aging fungal cultures [12,[17][18][19]. PCD is generally classified into three types, which are referred to as apoptosis (type I), autophagy (type II) and necrosis (type III) [20]. Their physiological roles are very complex and their relationships are not completely understood. While apoptosis and necrosis are explicitly associated with cell death, autophagy is also a normal physiological process important for cellular homeostasis by lysosomal degradation and recycling. The cellular functions of autophagy have been proposed to exert roles that are both causative of and protective against cell death [20][21][22].
Improving our understanding of processes induced by carbon starvation and their dynamic interactions is important to further optimize industrial production processes. The aim of this study is to provide a systemwide description of the carbon starvation response of the filamentous fungus A. niger. Submerged carbon-limited bioreactor batch cultures were performed and maintained starving up to six days after carbon depletion. In addition to describing the physiology and morphology, we analyzed the secretome and established genome-wide transcriptional profiles for three distinct starvation phases. Besides specifically dissecting expression data for groups of selected genes including proteases, chitinases and glucanases, we performed enrichment analysis to dissect the complex transcriptional changes.
Our investigation shows that carbon starvation in submerged cultures caused complex morphological changes and cellular differentiation including emergence of empty hyphal ghosts, secondary growth of thin non-branching filaments on the expense of older hyphal compartments and formation of conidiating structures. Concomitantly, autophagy and conidiation pathway genes were clearly induced on the transcriptional level. We propose that metabolic adaptation to carbon starvation is mediated by autophagy and that cell death rather than hydrolytic weakening of the fungal cell wall can be considered a hallmark of aging carbon starved A. niger cultures.

Physiology of carbon starved cultures
The A. niger wild type strain N402 [23] was cultivated under controlled conditions in bioreactors to study its response to carbon starvation during prolonged submerged batch cultivation ( Figure 1A and 1B). The defined medium had a pH of 3 and was balanced such, that carbon (maltose) was the growth limiting nutrient. During exponential growth (μ max = 0.24h −1 ), pH 3 was maintained by alkaline addition ( Figure 1B), which linearly correlated with the biomass accumulation and was previously shown to reflect ammonium uptake during balanced growth on minimal medium [24]. The end of the exponential growth phase was detected by an increase of the dissolved oxygen signal ( Figure 1B) and depletion of the carbon source was confirmed by measurements of maltose and glucose concentrations (not shown). The corresponding time point (defined as t=0) was used to synchronize replicate cultures insuring that samples were taken from equivalent physiological phases. The biomass concentration peaked at 5 g · kg −1 culture broth ( Figure 1A).
After maltose was exhausted, pH 3 was maintained by acid addition ( Figure 1B). The metabolic activity of the culture decreased in response to the lack of an easily accessible carbon and energy source as indicated by the CO 2 production and O 2 consumption rates ( Figure 1B). Protease activity rapidly increased and was already detected within 3 hours after maltose depletion. http://www.biomedcentral.com/1471-2164/13/380 During the later starvation phase (up to 140 hours), the protease activity remained constant; however, extracellular protein levels doubled within 16 hours after carbon depletion and remained constant thereafter ( Figure 1B). Towards the end of the starvation phase, the cell mass decreased by nearly 60% ( Figure 1A). Importantly, CO 2 and O 2 levels in the exhaust gas indicated that the cultures were still metabolically active, even 140 hours after depletion of the carbon source ( Figure 1B).

Morphological differentiation during carbon starvation
Throughout the entire cultivation, A. niger displayed a dispersed morphology. During exponential growth, the mycelium remained intact and no damaged or empty http://www.biomedcentral.com/1471-2164/13/380 hyphae were observed (Figure 2A). Early after depletion of maltose and onset of starvation, empty hyphal compartments emerged and the diameter of growing hyphae significantly decreased ( Figure 2B). Throughout prolonged starvation, the fraction of empty hyphal compartments increased, but the cell wall exoskeleton appeared to remain intact ( Figure 2B, 2C and 2D). Fragmented, broken hyphal ghosts were rarely observed. Outgrowing thin filaments emerged, which continued elongating in a non-branching manner. Towards the later starvation phases (60 hours post carbon depletion), morphologically crippled asexual reproductive structures appeared which resembled low-density conidiophores without clearly distinguishable phialides and metulae ( Figure 2C and 2D). Even 140 hours after exhaustion of the carbon source, surviving compartments were present, which often showed outgrowing hyphae bearing asexual reproductive structures ( Figure 2D). Secondary growth of thin hyphae was even observed within empty hyphal ghosts ( Figure 2C). Similar to our results, morphological data from A. oryzae [25] indicate a sharp transition between thick and thin compartments ( Figure 2B) in response to carbon starvation, suggesting that hyphal diameters can be used to distinguish populations of old and young hyphae formed during primary growth on the supplied carbon source and secondary growth fueled by carbon recycling, respectively. To visualize the transition dynamics from thick (old) to thin (young) hyphae in response to carbon starvation, an image analysis algorithm was developed to analyze hyphal diameter distributions of the cytoplasm filled mycelial fraction. Microscopic pictures from samples of various cultivation time points were analyzed and probability density curves were plotted for the distributions of hyphal diameters ( Figure 3). Diameters from exponentially growing hyphae resembled a normal distribution with a mean of approximately 3 μm. In response to carbon starvation, a second population of thinner hyphae with a mean diameter of approximately 1 μm emerged. Throughout the course of starvation, there was a gradual transition from thick (old) to thin (young) hyphae for the cytoplasm filled fraction, suggesting that compartments of older hyphae originating from the exponential growth phase gradually underwent cell death and became empty while a new population of thin hyphae started to grow on the expense of dying compartments.

Transcriptomic response to carbon starvation
To follow transcriptomic changes during carbon starvation, total RNA was extracted from biomass harvested at different time points during batch cultivation. Although difficulties to isolate intact RNA from aging cultures were reported for A. nidulans [26], we could isolate total RNA of high quality from samples up to 140 hours after depletion of the sole carbon source, as assessed by lab on chip  16 hours after carbon depletion empty hyphal compartments emerged (white triangles) and new hyphae with a significantly reduced average diameter of approximately 1 μm appeared (black triangles). (C) 60 hours after carbon depletion, the number of empty hyphal compartments increased and thin hyphae elongated in a non-branching manner. First reproductive structures emerged (white-edged triangles). Thin hyphae even grew cryptically inside empty hyphal ghosts (black-edged triangles). (D) Even 140 hours after carbon depletion, surviving compartments were present (black pentagon) often bearing morphologically reduced reproductive structures (white-edged triangle). The mycelial network consisted largely of empty hyphal ghosts but hyphal fragmentation was rarely observed. The scale bar refers to 5 μm. http://www.biomedcentral.com/1471-2164/13/380 Hyphal population dynamics. For six distinct time points, probability density curves of hyphal diameters are shown. 2 hours prior to carbon depletion, a single population of hyphae with a mean diameter of approximately 3 μm was detected. After carbon depletion, a second population with a significantly reduced mean diameter of approximately 1 μm started to emerge. Throughout the course of starvation, the ratio of thin/thick hyphae gradually increased, indicating secondary growth on the expense of dying compartments. quality control (data not shown) and Northern analysis ( Figure 1C). Transient expression levels of the gammaactin encoding gene actA (An15g00560), the glycosyl hydrolase nagA (An09g02240) and the regulator of asexual sporulation brlA (An01g10540) are exemplarily shown in Figure 1C. While nagA can be considered an early response gene whose expression peaked 16 hours after exponential growth, brlA expression was induced later and remained constant after reaching a plateau at 64 hours of carbon starvation. Expression levels of actA decreased considerably after exponential growth but remained constant during later cultivation phases.
RNA samples from four distinct cultivation phases were subjected to genome-wide transcriptional profiling: Exponential growth phase, 16 hours (day 1), 60 hours (day 3) and 140 hours (day 6) post carbon depletion. Differentially expressed genes were identified by a moderated t-test [27] applying a critical FDR q-value of 0.005. Compared to the exponential growth phase, 7,292 of totally 13,989 genes (52%) were identified as differentially expressed during at least one of the starvation time points (Additional file 1). 1,722 genes were conjointly upregulated, whereas 2,182 genes were conjointly downregulated during carbon starvation ( Figure 4). Enrichment analyses using Gene Ontology (GO) [28], Pfam domain [29] and Kyoto Encyclopedia of Genes and Genomes (KEGG) [30] pathway annotations were performed to uncover major transcriptional trends. For A. niger, all three annotations are based on computational inference. Among them, GO annotation can be considered to have the best quality because it was inferred from the computationally and manually curated GO annotation of the closely related species A. nidulans [31].
The GO enrichment results are summarized in Figure 5 (see Additional file 2 for complete GO enrichment results). They cover 20% (668) and 33% (1,334) of all upand downregulated genes, respectively. Among the genes induced under carbon starvation, common and timedependent overrepresentation of GO terms was observed. While GO terms related to e.g. catabolic (autophagy, cytoplasm to vacuole targeting (CVT) pathway, fatty acid oxidation and trehalose catabolism) and reproductive (conidiation and mitotic cell cycle) processes were generally enriched, other processes responded in a timedependent manner constituting early, intermediate or late responses. Among the transiently enriched processes were non-glycolytic fermentation and PCD (day 1), cell wall organization (day 3), regulation of transcription from RNA polymerase II promoter (day 3 and 6) as well as reactive oxygen metabolism (day 6). In contrast to the upregulated genes, the downregulated gene sets did not display any time-dependent differences with respect to the significantly overrepresented GO terms. The commonly downregulated processes included transcription http://www.biomedcentral.com/1471-2164/13/380 from RNA polymerase I promoter, ribosome biogenesis, translation, secretion and respiration.
Pfam domain and KEGG pathway enrichment results are summarized in the supplemental data (Additional files 3 and 4). Although the three annotations have different sources, structures and levels of complexity, the individual enrichment results confirm each other. Only in a few cases, Pfam domain and KEGG pathway enrichment analyses provided additional information beyond the GO enrichment results. For example, among the upregulated genes at day 1, 3 and 6, those having a putative sugar transporter domain (PF00083) were strongly enriched (49,40,39 predicted genes out of 110, respectively). In consideration of the severe carbon limitation, it can be assumed that these predicted sugar transporters comprise high-affinity sugar transporters. Indeed, mstA (An12g07450) and mstF (An02g00590) encoding two high-affinity sugar/H + symporters [32] were significantly upregulated at day 1 and 3 as well as day 1, 3 and 6, respectively. Furthermore, the cytochrome P450 domain (PF00067) was significantly enriched among genes upregulated at day 1. The biochemical roles of the majority of cytochromes P450 are unknown but many are expected to be involved e.g. in the formation of pigments, antioxidants and secondary metabolites [33]. Two of the 44 enriched cytochrome P450 domain proteins are physically associated with distinct (putative) secondary metabolite clusters, of which one is the fumonisin cluster [1]. Obviously, induction of the fumonisin cluster constitutes an early and orchestrated response to carbon starvation. Transcript levels for 11 of the 14 predicted open reading frames were exclusively elevated at day 1, including the putative transcription factor encoded by An01g06900 (Additional file 5).
As implied by the enrichment results for both GO and KEGG pathway annotations, carbon starvation coordinately induced the expression of genes involved in autophagic processes. To date, more than 30 autophagy (atg) genes have been identified for Saccharomyces cerevisiae and other fungi [34,35], 23 of which have a predicted orthologue in A. niger. All except one were detected as significantly upregulated during at least one of the starvation time points (see Table 1). The expression level of atg8 (An07g10020), encoding a lipid-conjugated ubiquitin-like protein that controls the expansion of preautophagosomes [36], was the highest among all atg genes. At day 3 it reached 75% of the actin expression level during exponential growth. Despite this concerted induction during carbon starvation, it is clearly evident from the expression data that autophagic processes also play an important role during exponential growth, because atg gene expression levels ranged from 0.6% (atg12: An11g06920) to 24% (atg8: An07g10020) when compared with the actin gene expression level ( Table 1).
The induction of hydrolases, including proteases and glycosyl hydrolases, has been proposed as a key event in aging fungal cultures [12]. During carbon starvation, glycosyl hydrolases are involved in both the liberation of carbon from fungal cell wall polymers and cell wall remodeling. We identified those upregulated genes that putatively encode glycosyl hydrolases active on fungal cell wall polymers such as chitin, glucan and mannan ( Table 2) by mining publicly accessible data [1,37]. The expression profiles allow a general classification into early and late response genes. In agreement with literature [16], the chitinolytic genes chiB http://www.biomedcentral.com/1471-2164/13/380 (An02g07020) and nagA (An09g02240) were among the highest induced early response genes. The rapid transient induction of nagA as shown by Northern analysis ( Figure 1C) exemplarily corroborates the microarray data. In addition to the chitinolytic hydrolases, the group of intensely induced early response hydrolases includes the α-glucanase agnB (An07g08640), multiple β-glucanases and one mannanase. Besides a number of glycosyl hydrolases that were only marginally induced during the later time points, the chitinases cfcI (An02g13580) and ctcB (An09g05920) showed strong specific induction during the two later time points. It is thus tempting to speculate that cfcI and ctcB are rather involved in cell wall remodeling during asexual development than liberation of carbon from cell wall polymers.
The second group of hydrolases, namely proteases, fulfills diverse physiological functions ranging from signaling to nutrient recycling. In accordance to the rapidly increasing extracellular protease activity after carbon depletion ( Figure 1A), an early transcriptional induction of extracellular proteases was observed (Table 3). Compared to exponential growth, the expression levels of the two major secreted proteases pepA and pepB [38] were increased by more than 130 fold at day 1. Additionally, roughly 20 further predicted secreted proteases were induced during carbon starvation with transcript level changes ranging from 2 to 40. In agreement, expression of the main transcriptional regulator of proteases PrtT [39] was strongly upregulated. Furthermore, transcript levels of about 20 proteases lacking predicted signal peptide sequences were identified as significantly elevated (Table 3), suggesting considerable intracellular proteolytic activities during carbon starvation.
Northern ( Figure 1C), microscopic ( Figure 2C and 2D) and GO enrichment ( Figure 5) analyses clearly indicated that conidiation is one of the main responses provoked by carbon starvation. Transcriptomic data of a subset of genes predicted to be involved in asexual development in Aspergillus are shown in Table 4. Expression profiles of orthologous genes belonging to the two core regulatory pathways identified in A. nidulans, STUNTED (stuA → wetA) and BRISTLE (brlA → abaA → wetA) [40,41] suggest conservation of these regulatory pathways between the two Aspergilli. Whereas the first pathway is induced early upon achievement of asexual developmental competence (day 1), induction of the latter pathway is delayed (day 3). Among the fluffy genes flbA-E encoding upstream regulators of BrlA [8], only flbC and flbD were clearly induced. Remarkably, although only little asexual differentiation occurred, hydrophobins were among the most intensely induced genes (Table 4). In a global ranking based on highest expression levels at day 6, the three predicted hydrophobins encoded by An03g02400, An08g09880 and An03g02360 were at positions one, http://www.biomedcentral.com/1471-2164/13/380      five and six, respectively. In agreement, conidial pigmentation genes including olvA were strongly induced (Table 4).

Secretomic response to carbon starvation
To identify extracellular hydrolases secreted at various cultivation time points, mass spectrometric analyses of tryptically digested proteins precipitated from culture filtrates were performed. Neither chitin, α-glucan nor mannan active hydrolases were detected in the culture broth during exponential growth ( Table 2). In agreement to its high transcript levels during carbon starvation, NagA (An09g02240) was the most abundant extracellular hydrolase involved in chitin degradation at day 1, 3 and 6. However, the chitinase ChiB was, in contrast to its strong transcriptional upregulation, only marginally detected in filtrates at day 1. Both observations correspond well to the presence and absence of predicted signal peptide sequences for NagA and ChiB, respectively. Interestingly, ChiB of A. niger showed only low extracellular abundance, whereas the A. nidulans ChiB (AN4871) was identified as the major extracellular autolytic chitinase during carbon starvation [16]. The absence of ChiB in the culture broth of A. niger could explain why hyphal ghosts remained intact but were reported to fragment in aging cultures of A. nidulans [42]. In concordance with its expression profile, the α-glucanase AgnB (An07g08640) was detected extracellularly at day 1 and 3. While GelD (An09g00670) was the only reliably detected β-glucanase during exponential growth, various β-glucanases with predicted signal peptide sequences were detected at day 1, 3 and 6 of carbon starvation. Among the predicted mannanases, only An04g09650 was reliably detected in filtrates at later time points (day 3 and 6).
In agreement with increasing extracellular protease activity and expression profiles, a number of proteases with predicted signal peptide sequences were identified in culture filtrates of day 1, 3 and 6. Among them, PepA (An14g04710), the major extracellular protease [38], was most abundant. However, although PepB (An01g00530) has a predicted signal peptide sequence and showed strong transcriptional induction, it was not detected in culture filtrates. Transcriptionally induced proteases lacking predicted signal peptide sequences were not detected in culture filtrates, with the only exception of An01g00370. Similar results have been previously reported for A. niger by Braaksma et al. [43], who proposed that the high extracellular abundance of An01g00370 is likely a result of non-classical secretion rather than lysis.
The secretome during starvation conditions was clearly enriched by an additional group of proteins with strong similarity to phospholipases. Together the four putative phospholipases, An16g01880, An09g02180, An01g14940 and An02g13220 constituted on average about 7% of all detected extracellular proteins during day 1, 3 and 6. All except An02g13220 were transcriptionally induced during carbon starvation. This high abundance of predicted phospholipases during carbon starvation might be indicative for a role of membrane lipids as alternative carbon source during secondary growth. The complete list of identified extracellular proteins is given in Additional file 7.

Discussion
The present study is the first system-wide description of the carbon starvation response in a filamentous fungus. The application of bioreactor technology allowed highly reproducible culture conditions and physiological synchronization of replicate batch cultures. The use of minimal medium with maltose as the sole limiting nutrient, constant pH, sufficient aeration and homogeneously dispersed mycelial biomass reduced biological and technical variations to a minimum and allowed us to highlight those differences in gene expression, which were in direct relation to carbon starvation.
Submerged growth is fundamentally different from the natural fungal life style. Fungi experience spatio-temporal gradients of various ambient factors such as nutrients, temperature and pH in their natural habitats. These gradients lead to heterogenity within the fungal colony. Several studies have investigated this heterogeneity during growth on agar plates and have characterized differential concentric zones with respect to gene expression and protein secretion [44][45][46]. Recently, this heterogeneity has even been shown for microcolonies (pellets) in liquid shaken cultures of A. niger [47]. In an ideally mixed bioreactor, all dispersed hyphae experience identical environmental conditions and temporal profiles can be monitored and controlled by process parameters. Accordingly, many evolutionary acquired traits contributing to the natural fungal life style such as the formation of substrate exploring hyphae, secretion of certain hydrolases, cell death and conidiation are dispensable during industrial processes and might even negatively affect production yields.
In this study A. niger showed general hallmarks of autolysis [12] during prolonged carbon starvation. However, in contrast to A. nidulans [42], A. niger hyphae did not undergo substantial fragmentation. While an increasing number of hyphal compartments became empty after carbon depletion, microscopic analysis showed that hyphal cell wall skeletons remained mainly intact. Thus disintegration of aging mycelia appears rather to be initiated by intracellular activities such as cell death and/or endogenous recycling of neighboring compartments leading to empty hyphal ghosts than by extracellular hydrolysis of fungal cell walls ( Figure 6). This assumption is supported by studies in A. nidulans, where autolytic fragmentation http://www.biomedcentral.com/1471-2164/13/380  During the early phase of starvation, secondary growth fueled by carbon recycling is initiated as characterized by the formation of thin hyphae. Two mechanisms resulting in empty hyphal compartments are depicted. On the left side, apoptotic/necrotic signals lead to cell death of compartments. Cytoplasmic content leaks into the culture broth. Surviving compartments are protected by autophagic processes isolating/inactivating cell death signals. On the right side, endogenous recycling of neighboring compartments by autophagic processes leads to the formation of empty hyphal ghosts. Cytoplasmic content does not leak into the culture broth. During the intermediate phase, earlier processes continue and first reproductive structures emerge. Towards later phase, these processes proceed resulting in few surviving compartments often bearing reproductive structures and elongating thin hyphae. Depending on strain (e.g. creA) and cultivation conditions (e.g. elevated pH), a largely empty non fragmented mycelial network remains (left side) or fragmentation of empty hyphal ghosts occurs by hydrolytic weakening of cell walls (right side).
of hyphae and cell death were described as simultaneous but independently regulated processes [48]. While deletion of the major carbon catabolite repressor CreA in A. nidulans resulted in increased hydrolase activities and mycelial fragmentation during carbon starvation, the viability of A. nidulans was not affected [26]. Consistently, we observed hyphal fragmentation and enhanced biomass decline in bioreactor cultures during the starvation phase only when the pH control was switched-off leading to an elevated pH of approximately 5.8 towards the end of cultivation (Nitsche et al. unpublished data). We thus propose that hydrolytic weakening of the fungal cell wall and hyphal fragmentation is a secondary effect, which occurs after initial cell death events and only under favorable conditions ( Figure 6).
In flow chamber experiments with A. oryzae, Pollack et al. [25] followed single hyphae and studied their response to glucose depletion. Similar to our results, they observed secondary growth fueled by carbon recycling, which was morphologically characterized by the formation of hyphae with significantly reduced diameters. For A. niger and A. oryzae [25,49,50] hyphal diameters were shown to linearly correlate with the specific growth rate, hence the reduction of hyphal diameters reflects the slow rate of secondary growth during the starvation phase. Focusing on non-empty compartments, we analyzed hyphal population dynamics from statistically valid sample sizes for different cultivation time points (Figure 3). Our data showed that older hyphae with larger diameters grown during carbon-sufficient http://www.biomedcentral.com/1471-2164/13/380 conditions gradually became empty, giving rise to a new population of thinner hyphae. Carbon for this secondary phase of growth might have been liberated from extraand/or intracellular sources. In agreement with another study of A. niger [43], our secretomic data revealed that the relative contribution of lysis was very limited, even under starvation conditions (Table 3). Compared to exponential growth, no relative accumulation of proteins without predicted signal peptide sequences was observed in culture filtrates. However, because these results could also be explained by an equilibrium between proteolytic degradation and leakage of cytoplasmic proteins, it still remains to be shown whether intracellular resources are endogenously recycled by neighboring compartments or first leak into the culture broth where they are subsequently taken up by surviving compartments (Figure 6).
One process known to be important for endogenous recycling of cytoplasmic content in eukaryotes is macroautophagy. In filamentous fungi, it is thought to play an important role in nutrient trafficking along the hyphal network promoting foraging of substrate exploring hyphae and conidiation [51,52]. However, besides endogenous recycling of nutrients, autophagy in general is clearly associated with cell death and is discussed to have protective roles related to the degradation of e.g. damaged mitochondria or unfolded proteins [20,53,54] ( Figure 6). It is strongly evident from our transcriptomic data that the induction of autophagic processes is a hallmark of carbon-starved aging fungal cultures. To which extend autophagic processes play a role in the protection against apoptotic/necrotic PCD, endogenous recycling and autophagic PCD remains to be shown in future studies.
The GO enrichment showed a joint downregulation of general protein biosynthesis and secretion pathways during carbon starvation. However, the extracellular accumulation of certain proteins with predicted signal peptide sequences including proteases, glycosyl hydrolases and phospholipases indicates a specification of those pathways which might be related to the emergence of the second population of thin poorly branching hyphae. This phenomenum has been observed, for example, during nitrogen starving surface cultures of Phanerochaete chrysosporium for which thin hyphae rather than thick hyphae have been shown to secrete manganese peroxidase [55].
The liberation of carbon from polymers such as fungal cell wall carbohydrates and secreted proteins is indicated by increased expression of glycosyl hydrolases and proteases as well as by increased extracellular protease activity. Strikingly, the major secreted protease PepA [38] was the second most abundant extracellular protein during carbon starvation, which was only excelled by protein levels of the maltoseinduced alpha-glucosidae GlaA (An03g06550) secreted during exponential growth. Although transcripts of the ChiB/NagA chitinolytic system accumulated simultaneously during carbon starvation as described previously for A. nidulans [16], only NagA could be identified extracellularly in high relative abundances. While the low relative abundance of ChiB in filtrates from day 1 is in agreement with the absence of a predicted signal peptide sequence, it conflicts with results obtained in A. nidulans [16], where it was identified as the major extracellular autolytic chitinase. Interestingly, despite its extracellular abundance, also A. nidulans ChiB lacks a signal peptide prediction. Whether A. nidulans ChiB is released by non-classical secretion or lysis remains to be shown. It is tempting to speculate that cell wall degrading hydrolases lacking a signal peptide sequence are part of the fungal PCD program and accumulate intracellularly in dying compartments to be subsequently released upon cell death for recycling of the remaining hyphal ghost. In view of the natural emerse growth of fungi, this could be a successful strategy for survival -released hydrolases will remain localized to hyphal ghosts and not become diluted as under submerged conditions. Future studies will be necessary to elucidate whether intracellular localization, retention at the cell wall, protein instability or inefficient translation explain the low abundance of ChiB in filtrates of A. niger.
Carbon starvation provoked asexual reproduction of A. niger, which was clearly evident by the formation of condiospores ( Figure 2D) and by expression of respective conidiation-related genes (Table 4). This elaborate developmental program requires liberation and recycling of carbon to proceed in aging batch cultures ( Figure 6). Increased heterogeneity and compartmentalization of the hyphal network resulting in empty, cryptically growing and conidiating compartments implies an ordered form of fungal cell death ensuring selfpropagation to survive life-threatening starvation conditions. In A. nidulans it was shown that disruption of the flbA gene, encoding a regulator of G-protein signaling acting upstream of BrlA, resulted in an enhanced autolytic phenotype [8]. Hence, vegetative growth, autolysis and conidiation are closely interwoven processes and future factorial genome-wide transcriptomic studies of wild-type and developmental mutants will allow deconstruction of fungal cell death and its link to developmental processes.

Conclusions
This study provides a comprehensive description of the carbon starvation response of the filamentous fungus A. niger during submerged cultivation. The impact of http://www.biomedcentral.com/1471-2164/13/380 secondary growth by carbon recycling was indicated by hyphal population dynamics illustrating a gradual transition from old to young hyphae. The induction of autophagic and reproductive processes was clearly evident by major genome-wide transcriptional trends. Hydrolases with strong transcriptional induction during carbon starvation include ChiB, NagA, AgnB, PepA and PepB. Importantly, fragmentation of empty hyphal ghosts was not observed, thus constituting direct evidence that autolysis in aging submerged cultures of A. niger is rather initiated by cell death than by hydrolytic weakening and fragmentation of cell walls.

Bioreactor cultivation Inoculation and culture conditions
Batch cultures were performed in 6.6 L BioFlo3000 bioreactors (New Brunswick Scientific) as previously described by Jørgensen et al. [11]. Briefly, autoclaved bioreactor vessels were filled with 5 L (kg) sterile MM. During cultivation at 30°C, the controller was set to maintain pH 3 by addition of titrants (2 M NaOH and 1 M HCl). Sterile air was supplied at a rate of 1 Lmin −1 . Prior to inoculation, 1.5 ml of 10% (w/v) filter-sterilized yeast extract was added to enhance conidial germination. Cultures were inoculated with freshly harvested spore suspensions to give 10 9 conidia per liter. To reduce the loss of hydrophobic conidia during germination, the stirrer speed was set to 250 rpm and the culture was aerated via the headspace during the first six hours after inoculation. Subsequently, the stirrer speed was increased to 750 rpm, 0.5 ml of polypropyleneglycol P2000 was added as antifoam agent and air was supplied via the sparger.

Online measurements
O 2 and CO 2 partial pressures of the exhaust gas were analyzed with a Xentra 4100C analyzer (Servomex BV, Netherlands). Dissolved oxygen tension (DOT) and pH were measured electrochemically with autoclavable sensors (Mettler Toledo).

Sample preparations
Cultures broth was harvested at regular intervals from batch cultures and mycelial biomass was retained by vacuum filtration using glass microfiber filters (Whatman). Both biomass and filtrate were quickly frozen in liquid nitrogen and subsequently stored at -80°C. Dry biomass concentrations were gravimetrically determined from lyophilized mycelium originating from a known mass of culture broth. Culture broth for microscopic analysis was quickly frozen in liquid nitrogen and stored at -80°C. For LC-MS/MS analysis, 1 ml of Sigmafast protease inhibitor cocktail (S8830, Sigma Aldrich) was added to 30 ml of culture filtrate and BSA was spiked as internal standard (1:10, BSA/total expected protein, w/w) before freezing in liquid nitrogen and storage at -80°C.

Protease activity assay
Extracellular protease activity measurements were performed similarly to a previously described method by Braaksma et al. [57] using N,N-dimethylated BSA as substrate. Measurements were performed in 96 well microtiter plates. 30 μl sample were incubated with 80 μl of 0.5% (w/v) N,N-dimethylated BSA in McIlvaine's citric acid-phosphate buffer, pH 3, for 30 min at 37°C. Reactions were stopped by addition of 190 μl fresh TNBSA borate buffer solution prepared by adding 50 μl of 5% 2,4,6,-trinitrobenzene sulfonic acid (TNBSA; Pierce) to 10 ml of borate buffer with 0.5 g l −1 Na 2 SO 3 , pH 9.3. TNBSA reacts with primary amines yielding a yellow chromophore that was measured at 405 nm after 10 min. Blank measurements for sample background correction were obtained by incubation of filtrates with citric acid buffer not containing N,N-dimethylated BSA. Non proteolytic release of amines from N,N-dimethylated BSA was assessed by incubation of N,N-dimethylated BSA without filtrate sample. 1 U of protease activity was defined as the activity, which within 1 min under the http://www.biomedcentral.com/1471-2164/13/380 described incubation conditions produces a hydrolysate with an absorption equal to that of 1 μmol glycine at 405 nm.

Extracellular protein quantification
Extracellular protein concentrations in culture filtrates were determined using the Quick Start Bradford Protein Assay (Bio-Rad) according to the manufacturer's instructions.

Microscopy and image analysis
Microscopic samples were slowly defrosted on ice. For differential interference contrast microscopy (DIC) an Axioplan 2 instrument (Zeiss) with a 100x oil immersion objective was used and micrographs were captured with an DKC-5000 digital camera (Sony). For the automated determination of hyphal diameters, samples were fixed and stained in a single step by mixing them at a 1:1 ratio with Lactophenolblue (Fluka). Sets of 40 micrographs were taken per sample with an DM IL LED (Leica) microscope using a 40x objective and an ICC50 camera (Leica). The microscope and camera settings were optimized to obtain micrographs with strong contrast. To measure hyphal diameters from micrographs of dispersed myclia in an automated manner, the following six-step image analysis algorithm was developed and implemented as a macro for the open source program ImageJ [58]: (1) Convert micrographs to binary images; (2) Copy binary images and outline all objects; (3) Copy binary images and skeletonize all objects; (4) Clean skeletons by removing all intersections; (5) Combine outline and skeleton images; (6) Fragment skeletons and orthogonally measure from the center of each skeleton fragment the distance to the outline.

RNA extraction, gene chip hybridization and Northern analysis
To minimize the chance of RNA degradation, frozen biomass was directly ground in liquid nitrogen and subsequently total RNA was isolated using the Trizol reagent (Invitrogen) according to the manufacturer's instructions. Prior to gene chip hybridization, samples were purified on NucleoSpin RNA II columns (Machery-Nagel) including a DNAse I treatment.

Transcriptome data analysis
RNA samples from four cultivation phases were subjected to genome-wide transcriptional profiling: Exponential growth phase, 16 hours (day 1), 60 hours (day 3) and 140 hours (day 6) post carbon depletion. While the expression data for the exponential growth phase was derived from triplicate cultures, expression data for the three post-exponential time points was obtained from duplicate cultures. Transcriptomic data were analyzed with the statistical programming language R [60]. The following packages of the open source and open development project Bioconductor [61] were used: affy [62], affycoretools [63], affyPLM [64] and limma [27]. Affymetrix probe level data was imported from .CEL files and preprocessed with the Robust Multi-array Average (RMA) [65] algorithm as implemented in the affy package. To improve background correction and data normalization, six additional .CEL files corresponding to day 2 and day 8 of carbon limited retentostat cultivations of A. niger [11], available at the Gene Expression Omnibus (GEO) database [66] (http://www.ncbi.nlm.nih.gov/geo/) under accession number: GSE21752, were included in the RMA preprocessing step. Prior to the computation of differentially expressed genes, 65 Affymetrix control probes and 204 probes targeting genetic elements were removed from the expression matrix. For 277 transcripts targeted by multiple probes, mean expression values were calculated from the RMA expression data of all associated probes. Subsequently, RMA expression data for the 13,989 transcripts were analyzed with the limma package comparing day 1, 3 and 6 of carbon starvation with the exponential growth phase. The Benjamini & Hochberg False Discovery Rate (FDR) [67] was controlled at 0.005. Because fold changes are not necessarily related to biological relevance [68,69], a minimal fold change criterion was not applied.

Annotation enrichment analyses
Enrichment analysis of Gene Ontology (GO) [28] terms was performed using the Fishers exact test Gene Ontology annotation tool (FetGOat: http://www.broadinstitute. org/fetgoat/index.html) [31] applying a critical Benjamini & Hochberg FDR q-value of 0.05. In order to compare and summarize enriched GO terms, we aimed to identify common most-specific GO terms for the sets of up-and downregulated genes, and thus implemented http://www.biomedcentral.com/1471-2164/13/380 the following algorithm in the Perl programming language: (1) Combine all enriched GO terms from the input sets; (2) Reduce redundancy from higher hierarchy terms by keeping only the most-specific (most-distant) GO terms; (3) For each remaining most-specific GO term, check all parental GO terms, sorted by increasing distance from the corresponding child term, for the presence in the input sets; (4) If a parental GO terms is present in all input sets denote it as common mostspecific, if any other equally distant parental terms are present in all input sets, denote them as common mostspecific as well and continue with the next most-specific GO term; (5) If none of the parental GO terms are present in all input sets, denote the corresponding mostspecific GO term as non-common most-specific; (6) After completing the analysis of all most-specific GO terms, reduce redundancy from the set of common most-specific terms by removing all their parental terms; (7) Output the sets of common most-specific and non-common GO terms. Fishers' exact test based enrichment analysis of Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway [30] and Pfam domain [29] annotations were performed using in-house developed Perl scripts. The Benjamini & Hochberg FDR was controlled at 0.05. KEGG pathway annotation (file: ang pathway.list, version from 22.06.2011) for A. niger was downloaded from the KEGG homepage (http://www.genome.jp/kegg/). Pfam domain annotation for A. niger was generated by analyzing the predicted proteome of A. niger strain CBS 513.88 [1] with the PfamScan Perl script (ftp://ftp.sanger.ac.uk/pub/ databases/Pfam/Tools/PfamScan.tar.gz).

Secretome analysis
Sample pre-treatment 8 ml of 100% (w/v) trichloracetic acid was added to frozen filtrate samples, which were subsequently completely defrosted by shaking at 4°C. Precipitated proteins were spun down and washed twice with ice-cold acetone. Precipitated protein pellets were air-dried, solubilized in 8 M urea (50 μl) and diluted 10x with 100 mM NH 4 HCO 3 . Reduction, alkylation of cysteines and digestion with trypsin were performed according to Thakur et al. [70]. Another aliquot of trypsine (2.5 μl 0.25 mg ml −1 pH 3) was added after overnight digestion followed by incubation for 3 hours at 37°C to ensure complete digestion. Samples were acidified to 1% formic acid.

Data analysis
The peptide data sets were searched against the A. niger database, which was manually modified to contain the sequences of trypsin and BSA, using the Sorcerer 2 Sequest search engine (SageN, San Diego, CA, USA). The search opted for carbamidomethylation (C), oxidation (M) and deamidation (NQ) as variable modifications. The Sequest results were processed using the APEX program [71] according to the author's description in order to obtain estimates of the protein quantities. Proteins identified with protein probability > 0.9 were considered as significant.

Microarray data accession number
Microarray data for the post-exponential time points have been made available at the GEO database [66] (http:// www.ncbi.nlm.nih.gov/geo/) under accession number GSE39559. Microarray data from the exponential growth phase were previously made available at the GEO database under accession number GSE21752. http://www.biomedcentral.com/1471-2164/13/380