Nucleic Acids Research, 2018 1
doi: 10.1093/nar/gky1183
Updating genome annotation for the microbial cell
factory
Aspergillus niger
using gene co-expression
networks
P. Sch¨
ape†,M.J.Kwon
†, B. Baumann, B. Gutschmann, S. Jung, S. Lenz , B. Nitsche,
N. Paege, T. Sch¨
utze, T.C. Cairns and V. Meyer *
Department of Applied and Molecular Microbiology, Institute of Biotechnology, Technische Universit¨
at Berlin,
Gustav-Meyer-Allee 25, 13355 Berlin, Germany
Received August 24, 2018; Revised October 31, 2018; Editorial Decision November 01, 2018; Accepted November 27, 2018
ABSTRACT
A significant challenge in our understanding of bio-
logical systems is the high number of genes with un-
known function in many genomes. The fungal genus
Aspergillus
contains important pathogens of hu-
mans, model organisms, and microbial cell factories.
Aspergillus niger
is used to produce organic acids,
proteins, and is a promising source of new bioactive
secondary metabolites. Out of the 14,165 open read-
ing frames predicted in the
A. niger
genome only 2%
have been experimentally verified and over 6,000 are
hypothetical. Here, we show that gene co-expression
network analysis can be used to overcome this limi-
tation. A meta-analysis of 155 transcriptomics exper-
iments generated co-expression networks for 9,579
genes (∼65%)ofthe
A. niger
genome. By populating
this dataset with over 1,200 gene functional exper-
iments from the genus
Aspergillus
and performing
gene ontology enrichment, we could infer biologi-
cal processes for 9,263 of
A. niger
genes, includ-
ing 2,970 hypothetical genes. Experimental valida-
tion of selected co-expression sub-networks uncov-
ered four transcription factors involved in secondary
metabolite synthesis, which were used to activate
production of multiple natural products. This study
constitutes a significant step towards systems-level
understanding of
A. niger
, and the datasets can be
used to fuel discoveries of model systems, fungal
pathogens, and biotechnology.
INTRODUCTION
The genus Aspergillus (phylum Ascomycota) is comprised
of nearly 350 species of saprophytic and ubiquitous fungi,
and includes important pathogens of humans (Aspergillus
fumigatus), model organisms (Aspergillus nidulans) and mi-
crobial cell factories (Aspergillus oryzae,Aspergillus niger).
Aspergillus niger has been exploited for over a century by
biotechnologists for the production of organic acids, pro-
teins and enzymes (1). It is the major worldwide producer
of citric acid with an estimated value of $2.6 billion in
2014, which is predicted to rise to $3.6 billion by 2020 (2).
As a prolific secretor of proteins, A. niger is used to pro-
duce various enzymes at a bulk scale (1). The first A. niger
genome was sequenced in 2007, which contained an esti-
mated 14,165 coding genes ∼6,000 of which were hypo-
thetical (3). More recent sequencing of additional A. niger
strains and other genomes from the genus Aspergillus (4–
9), combined with refinement of online genome analyses
portals (10–12), and comparative genomic studies amongst
the Aspergilli (5,13–15), have not significantly increased the
percentage of A. niger genes that have functional predic-
tions. While the exact number of ‘hypothetical’ genes varies
between databases and A. niger genomes, recent estimates
suggest that between 40 and 50% of the genes still remain
hypothetical (1,16). Furthermore, only 2% of its genes (n
=247) have a verified function in the Aspergillus Genome
Database (AspGD (17)). Even for the gold-standard model
organism Saccharomyces cerevisiae, 21% of its predicted
genes have dubious functional predictions (18), despite its
high genetic tractability and a research community with
>1,800 research labs worldwide. Indeed, gene functional
predictions for A. nidulans,A. fumigatus,A.oryze and other
Aspergilli typically cover 40–50% of the genome (16,17).
Such high frequency of unknown and hypothetical genes
severely limits the power of systems-level analyses. One ap-
proach to overcome this limitation involves the generation
and interrogation of gene expression networks based on
transcriptomic datasets (19–21). The hypothesis underlying
this approach is that genes which are robustly co-expressed
under diverse conditions are likely to function in the same
or closely related biological processes or pathways (22).
*To whom correspondence should be addressed. Tel: +49 30 314 72750; Fax: +49 30 314 72922; Email: vera.meyer@tu-berlin.de
†The authors wish it to be known that, in their opinion, the first two authors should be regarded as Joint First Authors.
C
The Author(s) 2018. Published by Oxford University Press on behalf of Nucleic Acids Research.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which
permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited.
Downloaded from https://academic.oup.com/nar/advance-article-abstract/doi/10.1093/nar/gky1183/5213907 by Technische Universitaet Berlin user on 18 January 2019
2Nucleic Acids Research, 2018
As one example, accurate delineation of fungal secondary
metabolite biosynthetic gene clusters has been achieved
by robustly defining contiguous gene co-expression dur-
ing both in vitro (23) and infectious (24) growth, with co-
expression analysis pipelines now publicly available to non-
coders (25).
In this study, we conducted a meta-analysis of 155 pub-
licly available transcriptomics analyses for A. niger,and
used these data to generate a genome-level co-expression
network and sub-networks for >9,500 genes. To aid user in-
terpretations of gene biological process, gene sub-networks
were analysed for enriched gene ontology (GO) terms,
and integrated with information gleaned from 1,200 val-
idated genes from the genus Aspergillus. Interrogation of
selected co-expression sub-networks for verified genes and
randomly selected hypothetical genes confirmed high qual-
ity datasets that enable rapid and facile predictions of bio-
logical processes. This co-expression resource has been in-
tegrated in the functional genomic database FungiDB (10)
for use by the research community.
In order to validate that novel predictions of gene bio-
logical function were accurate, we functionally character-
ized genes which we hypothesized played a role in A. niger
natural product synthesis based on co-expression datasets.
Experimental validation included generation of null and
overexpression mutants of transcription factors present in
these sub-networks, controlled bioreactor cultivations, and
activation of secondary metabolite gene expression and
metabolite biosynthesis. In addition to demonstrating that
this study enables novel predictions of A. niger gene func-
tion, these data suggest novel mechanisms for activating
cryptic secondary metabolism can be used in natural prod-
uct discovery programs, which is urgent due to the emer-
gence of multi-resistant bacteria and fungi (26,27). As A.
niger has been shown to be a superior expression host for
medicinal drugs in g/l scale (28), such discoveries have
significant translational potential. Taken together, the co-
expression resources and experimental validation devel-
oped in this study enable high quality gene functional pre-
dictions in A. niger.
MATERIALS AND METHODS
Strains and molecular techniques
Aspergillus niger strains used in this study are summarized
in Supplementary Table S1. Media compositions, transfor-
mation of A. niger, strain purification and fungal chromo-
somal DNA isolation were described earlier (29). Standard
PCR and cloning procedures were used for the generation of
all constructs (30) and all cloned fragments were confirmed
by DNA sequencing. Correct integrations of constructs in
A. niger were verified by Southern analysis (30). In the case
of overexpressing TF1, TF2 and HD, the respective open
reading frames were cloned into the Tet-on vector pVG2.2
(31) and the resulting plasmids integrated as single or mul-
tiple copies at the pyrG locus (for details see Supplementary
Table S1). Deletion constructs were made by PCR amplifi-
cation of the 5- and 3-flanks of the respective open reading
frames (at least 0.9 kb long). N402 genomic DNA served
as template DNA. The histidine selection marker (32)was
used for selecting single deletion strains, whereas the pyrG
marker was used for the establishment of the strain deleted
in both TF1 and TF2. Details on cloning protocols, primers
used and Southern blot results can be requested from the
authors.
Gene network analysis and quality control
As of 2 March 2016, 283 microarray data (platform:
GPL6758) of A. niger covering 155 different cultivation con-
ditions were publically available at the GEO database (33),
whose processing and normalization of the arrays have been
published (34). In brief, array data in the form of CEL-files
(35) were processed using the Affymetrix analysis package
(35) (version 1.42.1) from Bioconductor (36) and expression
data were calculated for genes under each condition with
an MAS5 background correction. Pairwise correlations of
gene expression between all A. niger genes were generated
by calculating the Spearman’s rank correlation coefficient
(37) using R. To assess a cut-off indicating biological rele-
vance, Spearman correlations were firstly calculated using
a pseudo random data set whereby normalized transcript
values for each individual gene were randomized amongst
the 283 arrays and 155 experimental conditions. Using this
pseudo random data set, the Spearman’s rank coefficient
was calculated pairwise for all predicted A. niger genes, giv-
ing a total of 104,958,315 comparisons/calculated Spear-
man’s rank coefficients, from which 52,476,536 were pos-
itively correlated. From these, only two were greater than
|0.4|and none above |0.5|. Subsequently |0.5|was taken as a
threshold for co-expression. Sub-networks were calculated
at an individual gene level using Python. All genes that were
co-expressed with individual query ORFs are reported at
FungiDB (10) and summarized in Supplementary File 1 us-
ing a |≥0.5|and |≥0.7|Spearman cut-off.
To expedite investigation of the sub-networks and their
common biological process, gene ontology enrichment
(GOE) was implemented using Python version 2.7.13. GO
terms and their hierarchical structure were downloaded
from AspGD (17). Enriched GO Biological Process terms
for all genes residing in a query sub-network were calculated
relative to the A. niger genome and statistical significance
was defined using the Fishers exact test (P-value <0.05).
For all datasets now available at FungiDB, GO enrichment
is conducted using Fisher’s exact test and Bonferroni cor-
rections (10). For informant ORFs, experimentally verified
A. niger genes were retrieved from AspGD (17). Addition-
ally, A. niger orthologs for any gene with wet lab verification
in A. fumigatus,A. nidulans or A. oryzae were identified us-
ing the ENSEMBL BLAST tool using default settings (12).
Finally, 81 secondary metabolite core enzymes (39,40)were
also defined as informant ORFs. We generated informant
ORF (‘prioritized ORF’) sub-networks, which report sig-
nificant co-expression of query genes exclusively with one
or more informant ORFs which are reported in Supplemen-
tary Table S2.
Reporter gene expression
Protocols for luciferase-based measurement of gene expres-
sion in microtiter format based on Tet-on (31)oranafp (34)
promoter systems have been published. In case of strain
Downloaded from https://academic.oup.com/nar/advance-article-abstract/doi/10.1093/nar/gky1183/5213907 by Technische Universitaet Berlin user on 18 January 2019
Nucleic Acids Research, 2018 3
BBA17.6, unable to form spores, the strain was inoculated
on complete medium and allowed to grow for 7 days at
30◦C. Biomass was harvested using physiological salt solu-
tion and used for inoculation. All data shown are derived
from biological duplicates each measured in technical qua-
druplicates if not otherwise indicated. Raw datasets can be
requested from the authors.
Bioreactor cultivation
Medium composition and the protocol for glucose-limited
batch cultivation of A. niger in 5 l bioreactors have been
described (31). In the case of strains overexpressing TF1
(strain MJK10.22) and TF2 (strain MJK11.17), the Tet-
on system was induced with a final concentration of 10
g/ml doxycycline when the culture reached 1 g/kg dry
biomass. Samples for transcriptional and metabolome pro-
filing were taken ∼6h(∼72 h) after induction, i.e. dur-
ing exponential (post-exponential) growth phase. For con-
trol (MJK17.25) and deletion strains (MJK14.7, MJK16.5,
MJK18.1), doxycycline was added twice; once after the cul-
ture reached 1 g/kg dry biomass and 24 h before samples
were taken from post-exponential growth phase for tran-
scriptomics and metabolomics analyses. To avoid modifica-
tions in gene expression due to degradation of doxycycline
in growth media, 10 g/ml doxycycline was added every
10–12 h (five times in total) during growth of conditional
expression mutants. Growth and physiology profiles for all
seven strains cultivated in biological duplicates are summa-
rized in Supplementary File 4.
Transcriptional profiling
Total RNA extraction, RNA quality control, and RNA
sequencing were performed at GenomeScan (Leiden, the
Netherlands). Quality analysis of raw data was done as
previously described (41). In brief, ∼13 million reads of
150 bp were obtained from paired-end mode for each
sample. Read data were trimmed and quality controlled
with FastQC (http://www.bioinformatics.babraham.ac.uk/
projects/fastqc/). STAR (42) was used to map the reads
to the A. niger CBS 513.88 genome (http://fungi.ensembl.
org/). On average, the unique alignment rate was ∼95%.
Data normalization was performed with DEseq2 (43). Dif-
ferential gene expression was evaluated with Wald test with
a threshold of the Benjamini and Hochberg False Discovery
Rate (FDR) of 0.05 (44) with DEseq2. Raw and processed
data are summarized in Supplementary Table S3 and have
been deposited at the GEO database (33) under the acces-
sion number GSE119311.
Metabolome profiling
Metabolites were extracted from biomass corresponding
to 2.5 mg biomass dry weight by Metabolomic Discov-
eries GmbH (Potsdam, Germany) and identified based
on Metabolomic Discoveries’ database entries of authen-
tic standards. Liquid chromatography (LC) separation was
performed using hydrophilic interaction chromatography
with a iHILIC®-Fusion, 150 ×2.1 mm, 5 m, 200 ˚
A5
m, 200 A column (HILICON, Ume˚
a Sweden), operated
by an Agilent 1290 UPLC system (Agilent, Santa Clara,
USA). The LC mobile phase was (i) 10 mM ammonium ac-
etate (Sigma-Aldrich, USA) in water (Thermo, USA) with
5% and 95% acetonitrile (Thermo, USA) (pH 6) and (ii)
acetonitrile with 5% 10 mM ammonium acetate in 95%
water. The LC mobile phase was a linear gradient from
95% to 65% acetonitrile over 8.5 min, followed by linear
gradient from 65% to 5% acetonitrile over 1 min, 2.5 min
wash with 5% and 3 min re-equilibration with 95% ace-
tonitrile. The flow rate was 400 l/min and injection vol-
ume was 1 l. Mass spectrometry was performed using a
high-resolution 6540 QTOF/MS Detector (Agilent, Santa
Clara, USA) with a mass accuracy of <2 ppm. Spectra
were recorded in a mass range from 50 m/zto 1700 m/z
at 2 GHz in extended dynamic range in both positive and
negative ionization mode. The measured metabolite con-
centrations were normalized to the internal standard. Sig-
nificant concentration changes of metabolites in different
samples were analyzed by appropriate statistical test proce-
dures (ANOVA, paired t-test) using R. When the adjusted
Pvalue based on Benjamini and Hochberg FDR (44)was
lower than 0.05 and the fold change (log2) higher than ±1,
expression of the metabolites was considered as significantly
different.
RESULTS
The transcriptomic landscape of A. niger inferred from a gene
expression meta-analysis
We normalized and interrogated gene expression across 155
published transcriptomic analyses for the A. niger labora-
tory wildtype strain N402 (ATCC 64974) and its descen-
dants, comprising of 283 Affymetrix microarray experi-
ments in total (34). Experimental parameters include a di-
verse range of cultivation conditions (agar plate, bioreac-
tor, shake flask), developmental and morphological stages
(germination, mycelial growth, sporulation), deletion and
disruption mutants, stress conditions (antifungals, secretion
stress, pH), different carbon and nitrogen sources, starva-
tion, and co-cultivation with bacteria. These experimental
conditions represent diverse niches inhabited by A. niger
as well as industrial cultivation conditions, in addition to
(a)biotic and genetic perturbations that result in global
changes in gene expression.
In order to demonstrate that accurate values of transcript
abundance were derived from this meta-analysis, we plotted
average gene expression values for each gene throughout the
155 conditions as a function of chromosomal locus (Fig-
ure 1A). From these data, we categorized low, medium, and
highly expressed loci. Subsequently, we generated a DNA
cassette expressing a luciferase reporter gene under con-
trol of the inducible Tet-on promoter (31) and targeted it
to the 5-upstream region of two low and one high expres-
sion locus (Figure 1B). The pyrG locus present on chro-
mosome III, routinely used for gene-targeted integration in
A. niger, served as locus control for Tet-on driven medium
expression of luciferase (31). Luciferase levels measured at
these loci were confirmed to be low, medium and high in
relative terms in microtiter cultivations of the different A.
niger strains (Figure 1C) and in controlled batch cultiva-
tion at bioreactor scale (Figure 1D). These data demon-
Downloaded from https://academic.oup.com/nar/advance-article-abstract/doi/10.1093/nar/gky1183/5213907 by Technische Universitaet Berlin user on 18 January 2019
4Nucleic Acids Research, 2018
Figure 1. The transcriptomic landscape of A. niger.(A) Chromosomal expression values expressed as log2 mean from 155 cultivation conditions. Blue
arrows indicated centromere position on each chromosome (3). (B) Selected chromosomal positions in chromosome III and VII for in vivo verification.
Red arrow indicates position of chosen integration site at high level expression locus, and grey arrows indicate positions of chosen integration sites at low
level expression loci. (C) Data for luciferase activity measured by luminescence in counts per second (cps) over time after induction in hours in microtiter
plates. Data for strains expressing luciferase at the locus An11g08480 (‘low gene expression locus’; strain BG1.1), An11g07580 (‘low gene expression locus’;
strain BG3.1), An12g03570 (‘medium gene expression locus’, pyrG, strain VG8.27), and An11g11310 (‘high gene expression locus’; strain BG6.14) are given.
Doxycycline was added 16 h after inoculation. (D) Cultivation of strain BG1.1 and BG6.14 in 5L bioreactors. For each strain, two batch cultivations were
performed; one was induced with 5 g/ml doxycycline at a dry biomass concentration of 2 g/kg. Cultivation in the presence of doxycycline mediated gene
expression as demonstrated by increased luminescence after induction measured in technical quintuplicate.
strate that low, medium, or high expression at these loci are
applicable for both high-throughput assays (microtiter) and
more labour and resource intensive bioreactors cultivations.
We conclude that the microarray data accurately reflect A.
niger gene expression values. Note that this transcriptomic
landscape is a significant addition to the A. niger molecu-
lar toolkit, as it facilitates rational control of gene dosage
(time of induction and absolute expression level) by targeted
locus-specific integration of a gene of interest.
Construction of high quality co-expression networks for A.
niger
Experimentally validated gene expression data from the A.
niger transcriptional meta-analysis was utilized to generate
a gene co-expression network based on Spearman’s rank
correlation coefficient (37). In order to define a minimum
Spearman’s rank correlation coefficient () for which we
could be confident in extracting biologically meaningful co-
expression, we conducted a preliminary quality control ex-
periment, where transcript values for each individual gene
were randomized amongst the 155 experimental conditions.
This gave a dataset with identically distributed but random-
ized expression patterns. Next, we calculated every possible
transcriptional correlation between genes on the A. niger
genome, resulting in over 100 million -values. This identi-
fied 52 million positive and 48 million negative correlations
(Figure 2). From this dataset, only two -values were above
|0.4|, and none were above |0.5|. Consequently, we took ≥
|0.5|as a minimum cut-off for biologically meaningful co-
expression relationships. Calculations of Spearman corre-
lations using the non-randomized microarray data resulted
in over 4.5 million correlations which passed the minimum
≥|0.5|cut-off. From these datasets, co-expression sub-
networks for every gene in the global network were gen-
erated for both positively and negatively correlated genes
(Figure 2, Supplementary File 1). We classified them into
two groups: ‘stringent’ (≥|0.5|, 9,579 gene networks) and
‘highly stringent’ (≥|0.7|, 6,305 gene networks) and calcu-
lated enriched GO terms for each gene sub-network relative
to the A. niger genome.
Integration of co-expression networks with community-wide
experimental evidence of gene function
The Aspergillus community has functionally characterized
over a thousand genes in different species of the genus As-
pergillus, which we reasoned can be used to aid apriori
Downloaded from https://academic.oup.com/nar/advance-article-abstract/doi/10.1093/nar/gky1183/5213907 by Technische Universitaet Berlin user on 18 January 2019
Nucleic Acids Research, 2018 5
Figure 2. Workflow to generate gene co-expression resources. (A) Various transcriptional signatures across 155 cultivation conditions of A. niger obtained
from 283 microarrays are schematically represented. Spearman’s rank correlation coefficients were calculated pairwise between all predicted genesinthe
A. niger genome. Over 96 million correlations between gene pairs were not defined as co-expressed based on the |0.5|cut-off (black and orange lines),
whereas 2.7 million were positively correlated (black and green lines) and 1.8 million were negatively correlated (black and red lines). These >4.5 million
significant co-expression relationships are visualized as a network in (B), with genes shown as grey squares, and correlations as lines. The length of each
line is proportional to the Spearman correlation between gene pairs. Genes predicted to encode transcription factors (3) are highlighted in blue. For each
individual gene in the network (13,975 in total; this number deviates from 14,165 reported in 2007 (3), because we have omitted truncated ORFs reported
there from our analysis), sub-networks were calculated which report all significant correlations with the query gene of ≥|0.5|(stringent dataset) and of
≥|0.7|(very stringent dataset). Both sub-networks were analyzed for significantly enriched GO terms (biological processes). Additionally, we highlight
co-expression between query genes and ‘informant ORFs’, which have either been experimentally validated in the Aspergilli, or are predicted to encode
key secondary metabolite biosynthetic enzymes (C).
predictions of hypothetical genes or not yet verified genes
in A. niger. In order to integrate such experimental data
with the co-expression network, we mined the Aspergillus
genome database AspGD (17) to generate a near-complete
list of ORFs that have been functionally characterized in
Aspergilli. All experimentally validated ORFs for A. niger
(n=247), A. nidulans (n =639), A. fumigatus (n =218) and
A. oryzae (n =81) were included in this dataset. Given the
strong potential of A. niger as a platform for discovery and
production of new bioactive molecules, we also included 81
putative polyketide synthase (PKS) or nonribosomal pep-
tide synthetase (NRPS) encoding genes of A. niger that re-
side in 78 predicted secondary metabolite clusters (39,40),
giving in total 1,266 prioritized ORFs. For every gene in the
A. niger genome, we calculated co-expression interactions
specifically with these 1,266 rationally prioritized ORFs. A
total of 9,263 (≥|0.5|) and 5,178 (≥|0.7|) candidate
genes had one or more correlations with prioritized ORFs
(Supplementary Table S2). These datasets thus constitute
the most comprehensive co-expression resource for a fila-
mentous fungus and are accessible at FungiDB (10).
Co-expression resources enable facile predictions of gene bi-
ological function
In order to test whether biologically meaningful interpreta-
tions of gene and network function can be extracted from
these resources, we interrogated both stringent and highly
stringent datasets for genes where the biological processes,
molecular function, and subcellular localization of encoded
proteins have been studied in fungi and which represent
the broad range of utilities and challenges posed by fungi.
From the perspective of industrial biotechnology, we inter-
rogated networks for the gene encoding the ATPase BipA,
which is required for high secretion yield of industrially use-
ful enzymes by acting as chaperone to mediate protein fold-
ing in the endoplasmic reticulum (45). With regards to po-
tential drug target discovery, we analyzed gene expression
networks for Erg11 (Cyp51), which is the molecular target
for azoles (46). For assessment of virulence in both plant
and human infecting fungi, we interrogated networks for
the NRPS SidD, which is necessary for the biosynthesis of
the siderophore triacetyl fusarinine C, and ultimately iron
acquisition during infection (47). Assessment of all con-
trol sub-networks at GO and individual gene-level revealed
striking co-expression of genes encoding proteins involved
in respective metabolic pathways, associated biological pro-
cesses, subcellular organelles, protein complexes, known
regulatory transcription factors/GTPases/chaperones, and
cognate transporters, amongst others (Figure 3). The lowest
Spearman correlation coefficient of |0.5|clearly results in bi-
ologically meaningful gene co-expression as exemplified by
the delineation of diverse yet related processes, including:
(i) orchestration of retrograde/anterograde vesicle traffick-
ing via COPI/COPII/secretion associated proteins (BipA)
(48); (ii) coordination of ergosterol biosynthesis by sterol
regulatory binding element regulators SrbA/SrbB and as-
sociation of this pathway with respiration at the mitochon-
drial membrane (Erg11) (49) and (iii) and the linking of
respective ergosterol and ornithine primary and secondary
metabolic pathways during siderophore biosynthesis via the
interdependent metabolite mevalonate (SidD) (50). With re-
gards to co-expressed genes as a function of chromosomal
location, a common feature of filamentous fungal genomes
is that genes necessary for the biosynthesis of secondary
metabolite products occur in physically linked contiguous
clusters. SidD resides in a six-gene cluster with SidJ, SidF,
SidH, SitT and MirD, all of which were represented in the
Downloaded from https://academic.oup.com/nar/advance-article-abstract/doi/10.1093/nar/gky1183/5213907 by Technische Universitaet Berlin user on 18 January 2019
6Nucleic Acids Research, 2018
Figure 3. Schematic representation of co-expression networks for BipA, SidD, and Erg11 encoding genes. Genes are represented by circles, with positive
and negative correlations depicted by grey and red lines, respectively. For simplicity, protein names are given in the centre of each circle. Where names
were not available in A. niger, we used the name for either A. nidulans or S. cerevisiae ortholog. Query genes encoding BipA, SidD, or Erg11 are given in
black diamond boxes. Co-expression sub-networks are given for gene pairs passing |0.7|(A) and |0.5|(B) Spearman correlation coefficient cut-offs. Both
sub-networks were assessed for enriched GO terms, and genes were manually filtered into functional categories based on these analyses and interrogation
of research literature (C). Note that in each instance, hypothetical proteins were co-expressed with each query gene, indicating the encoded products are
somehow associated with these biological processes.
high stringency network that contained a total of only 13
genes (Figure 3).
Based on enriched GO terms from gene sub-networks
and co-expression with experimentally verified ORFs, we
could further rapidly infer biological processes for a total
of 2,970 (≥|0.5|) and 1,016 (≥|0.7|) hypothetical genes
that were positively and/or negatively associated using this
analysis as exemplarily shown for eight hypothetical genes
in Supplementary Table S4. Additionally, we interrogated
entire families of functionally related genes that have been
well characterized in the Aspergilli, including phosphatases,
chromatin remodelers, and transcription factors and were
able to assign novel biological processes for all these pre-
dicted genes as exemplarily shown for nine genes in Supple-
mentary Table S5.
Additionally, in order to confirm that hypotheses re-
garding gene function and co-expression networks were
independent of the array platform and/or strain utilised,
we compared data derived from the microarray used in
our study (specifically the Affymetrix technology for A.
niger isolate CBS 513.88), with a separate tri-species plat-
form for strain ATCC 1015 (51). This latter microarray
has been used to concomitantly compare batch cultiva-
tions of A. niger,A. oryzae and A. nidulans on glucose
and xylose media, which identified 23 genes to be a con-
served response to xylose utilisation, including the xylose
transcriptional regulator XlnR (An15g05810, (51)). De-
spite differences in strain, microarray platform, and ex-
perimental design, interrogation of the XlnR sub-network
from our study revealed strong concordance with the con-
served xylose response genes reported by Andersen et
al.(51), including those encoding an l-arabitol dehydro-
genase (An01g10920), aldose 1-epimerase (An02g09090),
glycoside hydrolase (An12g01850), xylitol dehydrogenase
(An12g00030), sugar transporter (An03g01620) and short-
chain dehydrogenase (An04g03530). We therefore conclude
that sub-networks generated in our study can be used to
define A. niger co-expression relationships and infer gene
function across strain backgrounds.
Taken together, these quality control experiments
strongly suggest that the co-expression resources developed
in this study can be used for high confidence hypothesis
generation at a variety of conceptual levels, including
biological process, metabolic pathway, protein complexes,
and individual genes.
Co-expression resources accurately predict transcription fac-
tors of the ribosomally synthesized natural product AnAFP
of A. niger
In order to provide experimental confirmation in pre-
dictions of biological processes gleaned from this co-
expression resource, we interrogated all datasets associated
with the gene encoding the A. niger antifungal peptide
AnAFP. Ribosomally synthesized antifungal peptides of
the AFP family are promising molecules for use in medical
or agricultural applications to combat human- and plant-
pathogenic fungi (52). We and others could show that ex-
pression of their cognate genes are under tight temporal and
spatial regulation in their native hosts and precedes asexual
sporulation (34,53–55).
The gene encoding AnAFP (An07g01320), is co-
expressed with 986 genes (≥|0.5|; 605 positively cor-
related /381 negatively correlated (34)). GO enrichment
analyses of positively correlated sub-networks uncovered
that anafp gene expression parallels with fungal secondary
metabolism, carbon limitation and autophagy (34). In to-
tal, 23 predicted transcription factors are co-expressed at
a stringent level among which were the transcription fac-
tors VelC (An04g07320) and StuA (An05g00480; Figure 4),
both of which are key regulators of asexual development
and secondary metabolism in Aspergilli, whereby VelC is
Downloaded from https://academic.oup.com/nar/advance-article-abstract/doi/10.1093/nar/gky1183/5213907 by Technische Universitaet Berlin user on 18 January 2019
Nucleic Acids Research, 2018 7
Figure 4. In vivo proof of the predictive power of gene co-expression sub-networks. (A) Depicted are 23 predicted transcription factors in the anafp sub-
network which either positively (n =19) or negatively (n =4) correlate with the expression of the anafp gene. Two of these (velC and stuA) were selected
for in vivo verification by establishing velC and stuA deletion strains in an anafp::mluc reporter strain published earlier (strain PK2.9)(34). (B,C) Analysis
of anafp promoter activity in a strain deleted for velC (strain BBA17.9) and for stuA (strain BBA13.2), respectively, which was compared to the progenitor
strain PK2.9 abbreviated as ‘wt’. Note that the scale bar is different between B and C, and that StuA is thus likely a very strong repressor of anafp expression.
Data are from liquid microtiter plate cultures, where a defined amount of biomass was used for inoculation and luciferase expression monitored online
during cultivation. Expression levels are depicted for one representative example from three independent experiments. Each experiment was performed in
quintuplicate.
known as an activator, and StuA is both activator or re-
pressor (56–58). In order to confirm a regulatory function
of these transcription factors on anafp expression, we used
a reporter strain in which the anafp ORF has been replaced
with a luciferase gene. Deletion of stuA or velC in this back-
ground revealed a strong increase or decreased/delayed ac-
tivation of the anafp promoter, respectively (Figure 4). In-
terestingly, the transcription factor binding site for VelC is
unknown, whereas the binding sequence for StuA is absent
from the predicted promoter region of anafp. These data
thus indicate that the resources generated in this study en-
able accurate predictions of (in)direct regulatory proteins
even in the absence of DNA binding sites.
Co-expression resources accurately predict transcription fac-
tors of non-ribosomally synthesized natural products of A.
niger
The transcriptional activation of secondary metabolite
(SM) gene clusters in different filamentous fungi is one cur-
rent focus of the fungal research community (16)as>60%
of currently approved clinical drugs are derived from nat-
ural products (59). A. niger stands out due to its excep-
tional high number of predicted SM gene clusters in its
genome (n =78), harboring 81 core enzymes in total, such
as NRPS and PKS (39). However, only a dozen of SMs
have been identified from A. niger so far (60). Our sur-
vey of the expression data of all gene clusters under the
155 cultivation conditions uncovered that the majority of
SM core genes (53) are expressed in at least one condi-
tion (Supplementary File 2). The majority of expressed
core genes are also co-expressed with their cluster mem-
bers (Supplementary File 3). Notably, not all cluster mem-
bers are co-expressed with contiguous transcription fac-
tors. Indeed, only ∼30% of the gene clusters display co-
expression with contiguous transcription factors (Supple-
mentary File 3). We thus questioned which transcription
factors are regulating these SM gene clusters, and used the
co-expression dataset to assign biological processes to genes
predicted to encode transcription factors. Given the impor-
tant role of chromatin remodelers in activation and silenc-
ing of secondary metabolite clusters, we also interrogated
genes predicted to encode histone deacetylases (61). This
identified two ORFs encoding putative transcription fac-
tors: An07g07370 (TF1) and An12g07690 (TF2), and a his-
tone deacetylase (An09g06520, HD) that are positively and
negatively (TF1, TF2) or only negatively (HD) co-expressed
with numerous core SM genes (Supplementary Table S5).
Notably, all three genes do not reside in contiguous SM gene
clusters but belong to a large SM sub-network consisting
of 152 genes including 26 SM core genes (Supplementary
Table S6), whereby gene expression of TF1 and TF2 cor-
relate very strongly (=0.87). Interrogation of enriched
GO terms for both TF1 and TF2 gene sub-networks re-
vealed enrichment of fatty acid metabolism, autophagy, mi-
tochondria degradation (positively correlated) and matura-
tion of rRNA and tRNA, ribosomal assembly and amino
acid metabolism (negatively correlated). This analysis thus
allowed us to select genes for in vivo functional studies based
on a non-intuitive selection procedure. Neither TF1, TF2
nor HD have been experimentally characterized in fungi so
far. In order to confirm a regulatory function of these puta-
tive regulators on SM core gene expression, we generated (i)
single deletion strains for TF1, TF2 and HD, respectively,
(ii) a double deletion strain for TF1 and TF2, and (iii) in-
dividual conditional overexpression mutants for TF1, TF2
and HD using the Tet-on system (31). The strongest effect
Downloaded from https://academic.oup.com/nar/advance-article-abstract/doi/10.1093/nar/gky1183/5213907 by Technische Universitaet Berlin user on 18 January 2019
8Nucleic Acids Research, 2018
Figure 5. Co-expression sub-networks uncover global regulators of secondary metabolism in A. niger. Controlled batch cultivations using 5 l bioreactor
systems were performed with the progenitor strain and strains deleted for HD, TF1 and/or TF2 (TF1, TF2, TF12, HD) or overexpressing either
transcription factors (TF1OE, TF2OE). Note that we excluded the strain overexpressing HD from this analysis because its gene expression is negatively
correlated with the mostly silent SM core genes. Growth and physiology profiles are given in Supplementary File 4. Samples were taken for untargeted
metabolome analyses (LC–QTOF/MS, Supplementary Table S7) and global RNA-Seq (Supplementary Table S3) during exponential and post-exponential
growth phase. Note that the maximum growth rate was for all strains 0.24 h−1except for TF1OE (0.19 h−1). (A) Differential gene expression for 81 predicted
secondary metabolite core genes in mutant isolates relative to the progenitor control as determined by RNA-Seq analysis. (B) Assessment of metabolite
abundance demonstrated differential abundance of hundreds of A. niger metabolites following deletion or over-expression of HD, TF1 and TF2. In total,
3,410 primary and secondary metabolites were detected in all runs, 1,260 of which were known compounds and 478 thereof were differentially expressed
upon deletion or overexpression of these putative regulators (Supplementary File 4). (C) Relative abundance of the SM aurasperone B is shown in progenitor
control and mutant isolates throughout bioreactor cultivation as an exemplar.
on the metabolome profile of A. niger was observed dur-
ing overexpression of TF1 and TF2 (Figure 5). SMs up-
regulated under these conditions included, but were not lim-
ited to aurasperones, citreoviridin D, terrein, aspernigrin A,
nigerazine A and B, pyranonigrin A and D, flavasperone,
fonsecin, O-demethylfonsecin, flaviolin, funalenone; among
the SMs down-regulated under these conditions were asper-
pyrone, L-agaridoxin and nummularine F (Figure 5, Sup-
plementary File 4, Supplementary Table S7). These physio-
logical changes were paralleled by up/down-regulation of
thousands of genes as determined by RNA-Seq analyses
(Supplementary Table S3), whereby controlled overexpres-
sion of TF1 (TF2) modulated expression of 45 (43) SM core
genes especially during post-exponential growth phase of
A. niger (Figure 5and Supplementary File 4). This strongly
suggests that both transcription factors are likely global reg-
Downloaded from https://academic.oup.com/nar/advance-article-abstract/doi/10.1093/nar/gky1183/5213907 by Technische Universitaet Berlin user on 18 January 2019
Nucleic Acids Research, 2018 9
ulators modulating gene expression dynamics during late
growth stages of A. niger either directly or indirectly.
DISCUSSION
In this study, we performed a transcriptomic meta-analysis
to generate a high-quality gene co-expression network, and
used this to predict biological processes for 9,579 (∼65%)
of all A. niger genes including 2,970 hypothetical genes (≥
|0.5|). The compendium of resources developed in this work
consists of (i) gene-specific sub-networks with two stringent
Spearman cut-offs that ensure high confidence in biolog-
ically meaningful interpretations; (ii) statistically enriched
GO terms for each co-expression network, and (iii) a re-
fined list of co-expression relationships which incorporate
over 1,200 experimentally characterized ORFs to aid pre-
dictions of gene biological process based on experimental
evidence.
In order to demonstrate the utility of these resources, we
firstly interrogated these datasets at a gene-level, demon-
strating that transcription factors VelC and StuA, which are
critical components of ascomycete development and sec-
ondary metabolism, also regulate expression of the A. niger
antifungal peptide AnAFP. Our data provide further evi-
dence of the coupling between development, biosynthesis
of secondary metabolites, and secreted antifungal peptides
(34). The datasets generated in this study can also be used to
identify global regulators at the level of biological processes
as demonstrated for fungal secondary metabolism and the
two transcription factors TF1 and TF2 (which we name
MjkA and MjkB, respectively). Previously, co-expression of
contiguous genes has been used to determine the bound-
aries of secondary metabolite clusters in A. nidulans and
other fungi (23–25). Our study can be viewed as comple-
mentary to such analyses, as MjkA and MjkB are phys-
ically located outside the boundaries of any putative sec-
ondary metabolite cluster, and as such would not be identi-
fied using previous approaches (23,25). Generation of loss-
and gain-of-function mutants demonstrated that they likely
(in)directly regulate dozens of secondary metabolite loci
at the transcript and metabolite level. Interestingly, MjkA
is a Myb-like transcription factor highly conserved in As-
pergilli with orthologues present in several plant genomes.
Myb transcription factors have recently been demonstrated
to regulate plant natural product biosynthesis (62), and
our co-expression data and wet lab experiments suggest
that titratable control of MjkA is a promising strategy for
the activation of ascomycete secondary metabolism during
drug discovery programs. With regards to the application
of our co-expression approach to predict gene biological
processes in other fungi, interrogation of the GEO database
(33) demonstrates that several hundred global gene expres-
sion experiments are available for industrial cell factories
(e.g. A. oryzae,Trichoderma reesei) and human or plant in-
fecting fungi (e.g. A. fumigatus,Cryptococcus neoformans,
Candida albicans,Magnaporthe oryzae), indicating that our
approach can be broadly applied for industrial, medically,
and agriculturally relevant fungi. As the financial costs for
gene expression profiling continues to decline, this study
paves the way for prediction of gene biological function us-
ing co-expression network analyses throughout the fungal
kingdom.
DATA AVAILABILITY
All datasets generated and/or analyzed during this study
are available at FungiDB (http://fungidb.org/fungidb/).
Spearman’s correlation coefficients for gene co-expression
sub-networks will also be made available by the correspond-
ing author upon request. RNA seq data have been deposited
at the Gene expression Omnibus under accession number
GSE119311.
SUPPLEMENTARY DATA
Supplementary Data are available at NAR Online.
ACKNOWLEDGEMENTS
The authors wish to acknowledge all members of the
FungiDB project for providing the bioinformatics infras-
tructure to integrate the omic’s datasets of this study. Carina
Feldle is acknowledged for her assistance during bioreactor
cultivations.
Authors’ contributions: P.S. performed the transcriptome
meta-analyses, constructed the co-expression network and
executed in silico quality analyses. M.J.K. analyzed sub-
networks related to secondary metabolism. M.J.K., S.J.
and N.P. generated deletion and overexpression strains
and characterized them. M.J.K. and T.S. performed biore-
actor cultivations. M.J.K. analyzed transcriptome and
metabolome data derived from deletion and overexpression
strains. B.B. and B.G. contributed to molecular analyses,
B.N. and S.L. contributed to bioinformatics analyses. T.C.
contributed to bioinformatics analyses and co-wrote the fi-
nal text. V.M. initiated this study, coordinated the project
and co-wrote the final text. All authors read and approved
the final manuscript.
FUNDING
European Commission (funding by the Marie Curie Inter-
national Training Network QuantFung, FP7-People-2013-
ITN) [607332]. Funding for open access charge: Technische
Universit¨
at Berlin.
Conflict of interest statement. None declared.
REFERENCES
1. Cairns,T.C., Nai,C. and Meyer,V. (2018) How a fungus shapes
biotechnology: 100 years of Aspergillus niger research. Fungal Biol.
Biotechnol.,5, 13.
2. Show,P.L., Oladele,K.O., Siew,Q.Y., Aziz Zakry,F.A., Lan,J.C.-W.
and Ling,T.C. (2015) Overview of citric acid production from
Aspergillus niger.Front. Life Sci.,8, 271–283.
3. Pel,H.J., Winde,J.H., Archer,D.B., Dyer,P.S., Hofmann,G.,
Schaap,P.J., Turner,G., de Vries,R.P., Albang,R., Albermann,K. et al.
(2007) Genome sequencing and analysis of the versatile cell factory
Aspergillus niger CBS 513.88. Nat. Biotechnol.,25, 221–231.
4. Gong,W., Cheng,Z., Zhang,H., Liu,L., Gao,P. and Wang,L. (2016)
Draft genome sequence of Aspergillus niger strain An76. Genome
Announc.,4, e01700–e01715.
Downloaded from https://academic.oup.com/nar/advance-article-abstract/doi/10.1093/nar/gky1183/5213907 by Technische Universitaet Berlin user on 18 January 2019
10 Nucleic Acids Research, 2018
5. de Vries,R.P., Riley,R., Wiebenga,A., Aguilar-Osorio,G., Amillis,S.,
Uchima,C.A., Anderluh,G., Asadollahi,M., Askin,M., Barry,K. et al.
(2017) Comparative genomics reveals high biological diversity and
specific adaptations in the industrially and medically important
fungal genus Aspergillus. Genome Biol.,18, 28.
6. Wang,B., Lv,Y., Li,X., Lin,Y., Deng,H. and Pan,L. (2018) Profiling of
secondary metabolite gene clusters regulated by LaeA in Aspergillus
niger FGSC A1279 based on genome sequencing and transcriptome
analysis. Res. Microbiol.,169, 67–77.
7. Singh,N.K., Blachowicz,A., Romsdahl,J., Wang,C., Torok,T. and
Venkateswaran,K. (2017) Draft genome sequences of several fungal
strains selected for exposure to microgravity at the international space
station. Genome Announc.,5, e01602–e01616.
8. Yin,C., Wang,B., He,P., Lin,Y. and Pan,L. (2014) Genomic analysis
of the aconidial and high-performance protein producer, industrially
relevant Aspergillus niger SH2 strain. Gene,541, 107–114.
9. Paul,S., Lude˜
na,Y., Villena,G.K., Yu,F., Sherman,D.H. and
Guti´
errez-Correa,M. (2017) High-quality draft genome sequence of a
biofilm forming lignocellulolytic Aspergillus niger strain ATCC
10864. Stand Genomic Sci.,12,9.
10. Stajich,J.E., Harris,T., Brunk,B.P., Brestelli,J., Fischer,S., Harb,O.S.,
Kissinger,J.C., Li,W., Nayak,V., Pinney,D.F. et al. (2012) FungiDB:
an integrated functional genomics database for fungi. Nucleic Acids
Res.40, D675–D681.
11. Grigoriev,I.V., Nikitin,R., Haridas,S., Kuo,A., Ohm,R., Otillar,R.,
Riley,R., Salamov,A., Zhao,X., Korzeniewski,F. et al. (2014)
MycoCosm portal: gearing up for 1000 fungal genomes. Nucleic
Acids Res.,42, D699–D704.
12. Zerbino,D.R., Achuthan,P., Akanni,W., Amode,M.R., Barrell,D.,
Bhai,J., Billis,K., Cummins,C., Gall,A., Gir´
on,C.G. et al. (2018)
Ensembl 2018. Nucleic Acids Res.,46, D754–D761.
13. Yin,X., Shin,H.D., Li,J., Du,G., Liu,L. and Chen,J. (2017)
Comparative genomics and transcriptome analysis of Aspergillus
niger and metabolic engineering for citrate production. Sci. Rep.,7,
41040.
14. Andersen,M.R., Salazar,M.P., Schaap,P.J., Van De
Vondervoort,P.J.I., Culley,D., Thykaer,J., Frisvad,J.C., Nielsen,K.F.,
Albang,R., Albermann,K. et al. (2011) Comparative genomics of
citric-acid-producing Aspergillus niger ATCC 1015 versus
enzyme-producing CBS 513.88. Genome Res.,21, 885–897.
15. Vesth,T.C., Nybo,J.L., Theobald,S., Frisvad,J.C., Larsen,T.O.,
Nielsen,K.F., Hoof,J.B., Brandl,J., Salamov,A., Riley,R. et al. (2018)
Investigation of inter- and intraspecies variation through genome
sequencing of Aspergillus section Nigri. Nat. Genet.,
doi:10.1038/s41588-018-0246-1.
16. Meyer,V., Andersen,M.R., Brakhage,A.A., Braus,G.H.,
Caddick,M.X., Cairns,C.T., de Vries,R.P., Haarmann,T., Hansen,K.,
Hertz-Fowler,K. et al. (2016) Current challenges of research on
filamentous fungi in relation to human welfare and a sustainable
bio-economy: a white paper. Fungal Biol. Biotechnol.,3, 1–17.
17. Cerqueira,G.C., Arnaud,M.B., Inglis,D.O., Skrzypek,M.S.,
Binkley,G., Simison,M., Miyasato,S.R., Binkley,J., Orvis,J., Shah,P.
et al. (2014) The Aspergillus Genome Database: Multispecies
curation and incorporation of RNA-Seq data to improve structural
gene annotations. Nucleic Acids Res.,42, D705–D710.
18. Cherry,J.M., Hong,E.L., Amundsen,C., Balakrishnan,R., Binkley,G.,
Chan,E.T., Christie,K.R., Costanzo,M.C., Dwight,S.S., Engel,S.R.
et al. (2012) Saccharomyces Genome Database: The genomics
resource of budding yeast. Nucleic Acids Res.,40, D700–D705.
19. Gov,E. and Arga,K.Y. (2017) Differential co-expression analysis
reveals a novel prognostic gene module in ovarian cancer. Sci. Rep.,7,
4996.
20. van Dam,S., V˜
osa,U., van der Graaf,A., Franke,L. and de
Magalh˜
aes,J.P. (2017) Gene co-expression analysis for functional
classification and gene–disease predictions. Brief. Bioinform.,19,
575–592.
21. Hsu,C.-L., Juan,H.-F. and Huang,H.-C. (2015) Functional analysis
and characterization of differential coexpression networks. Sci. Rep.,
5, 13295.
22. Dutkowski,J., Kramer,M., Surma,M.A., Balakrishnan,R.,
Cherry,J.M., Krogan,N.J. and Ideker,T. (2013) A gene ontology
inferred from molecular networks. Nat. Biotechnol.,31, 38–45.
23. Andersen,M.R., Nielsen,J.B., Klitgaard,A., Petersen,L.M.,
Zachariasen,M., Hansen,T.J., Blicher,L.H., Gotfredsen,C.H.,
Larsen,T.O., Nielsen,K.F. et al. (2013) Accurate prediction of
secondary metabolite gene clusters in filamentous fungi. Proc. Natl.
Acad. Sci. U.S.A.,110, E99–E107.
24. Cairns,T. and Meyer,V. (2017) In silico prediction and
characterization of secondary metabolite biosynthetic gene clusters in
the wheat pathogen Zymoseptoria tritici.BMC Genomics,18, 631.
25. Vesth,T.C., Brandl,J. and Andersen,M.R. (2016) FunGeneClusterS:
Predicting fungal gene clusters from genome and transcriptome data.
Synth. Syst. Biotechnol.,1, 122–129.
26. Cooper,M.A. and Shlaes,D. (2011) Fix the antibiotics pipeline.
Nature,472, 32.
27. Macheleidt,J., Mattern,D.J., Fischer,J., Netzker,T., Weber,J.,
Schroeckh,V., Valiante,V. and Brakhage,A.A. (2016) Regulation and
Role of Fungal Secondary Metabolites. Annu. Rev. Genet.,50,
371–392.
28. Boecker,S., Gr¨
atz,S., Kerwat,D., Adam,L., Schirmer,D., Richter,L.,
Petras,D., S¨ussmuth,R.D. and Meyer,V. (2018) Aspergillus niger is a
superior expression host for the production of bioactive fungal
cyclodepsipeptides. Fungal Biol. Biotechnol.,5,4.
29. Meyer,V., Ram,A.F.J. and Punt,P.J. (2010) Genetics, genetic
manipulation, and approaches to strain improvement of filamentous
fungi. In: Manual of Industrial Microbiology and Biotechnology.3rd
edn. Wiley, NY, pp. 318–329.
30. Green,M.R. and Sambrook,J. (2012) Molecular Cloning: A
Laboratory Manual. Cold Spring Harbor Laboratory Press, NY, Vol.
1-3, pp. 1–2028.
31. Meyer,V., Wanka,F., van Gent,J., Arentshorst,M., van den
Hondel,C.A. and Ram,A.F. (2011) Fungal gene expression on
demand: an inducible, tunable, and metabolism-independent
expression system for Aspergillus niger.Appl. Environ. Microbiol.,77,
2975–2983.
32. Fiedler,M.R.M., Gensheimer,T., Kubisch,C. and Meyer,V. (2017)
HisB as novel selection marker for gene targeting approaches in
Aspergillus niger.BMC Microbiol.,17, 57.
33. Barrett,T., Wilhite,S.E., Ledoux,P., Evangelista,C., Kim,I.F.,
Tomashevsky,M., Marshall,K.A., Phillippy,K.H., Sherman,P.M.,
Holko,M. et al. (2013) NCBI GEO: Archive for functional genomics
data sets - Update. Nucleic Acids Res.,41, D991–D995.
34. Paege,N., Jung,S., Sch¨
ape,P., M¨uller-Hagen,D., Ouedraogo,J.P.,
Heiderich,C., Jedamzick,J., Nitsche,B.M., van den Honde,C.A.,
Ram,A.F. et al. (2016) A transcriptome meta-analysis proposes novel
biological roles for the antifungal protein Anafp in Aspergillus niger.
PLoS One,11, e0165755.
35. Gautier,L., Cope,L., Bolstad,B.M. and Irizarry,R.A. (2004)
affy––analysis of Affymetrix GeneChip data at the probe level.
Bioinformatics,20, 307–315.
36. Gentleman,R.C., Carey,V.J., Bates,D.M., Bolstad,B., Dettling,M.,
Dudoit,S., Ellis,B., Gautier,L., Ge,Y., Gentry,J. et al. (2004)
Bioconductor: open software development for computational biology
and bioinformatics. Genome Biol.,5, R80–R80.
37. Spearman,C. (1987) The proof and measurement of association
between two things. By C. Spearman, 1904. Am. J. Psychol.,100,
441–471.
39. Inglis,D.O., Binkley,J., Skrzypek,M.S., Arnaud,M.B.,
Cerqueira,G.C., Shah,P., Wymore,F., Wortman,J.R. and Sherlock,G.
(2013) Comprehensive annotation of secondary metabolite
biosynthetic genes and gene clusters of Aspergillusnidulans,A.
fumigatus,A. niger and A. oryzae.BMC Microbiol.,13, 91.
40. Sanchez,J.F. and Wang,C.C.C. (2012) The chemical identification and
analysis of Aspergillus nidulans secondary metabolites. In: Fungal
Secondary Metabolism - Methods and Protocols, pp. 97–109.
41. Park,J., Hulsman,M., Arentshorst,M., Breeman,M., Alazi,E.,
Lagendijk,E.L., Rocha,M.C., Malavazi,I., Nitsche,B.M., van den
Hondel,C.A. et al. (2016) Transcriptomic and molecular genetic
analysis of the cell wall salvage response of Aspergillus niger to the
absence of galactofuranose synthesis. Cell Microbiol.,18, 1268–1284.
42. Dobin,A., Davis,C.A., Schlesinger,F., Drenkow,J., Zaleski,C., Jha,S.,
Batut,P., Chaisson,M. and Gingeras,T.R. (2013) STAR: Ultrafast
universal RNA-seq aligner. Bioinformatics,29, 15–21.
43. Love,M.I., Huber,W. and Anders,S. (2014) Moderated estimation of
fold change and dispersion for RNA-seq data with DESeq2. Genome
Biol.,15, 550.
Downloaded from https://academic.oup.com/nar/advance-article-abstract/doi/10.1093/nar/gky1183/5213907 by Technische Universitaet Berlin user on 18 January 2019
Nucleic Acids Research, 2018 11
44. Benjamini,Y. and Hochberg,Y. (1995) Controlling the false discovery
rate: a practical and powerful approach to multiple testing. J. R. Stat.
Soc. Ser. B,57, 289–300.
45. Punt,P.J., van Gemeren,I.A., Drint-Kuijvenhoven,J., Hessing,J.G.,
van Muijlwijk-Harteveld,G.M., Beijersbergen,A., Verrips,C.T. and
van den Hondel,C.A. (1998) Analysis of the role of the gene bipA,
encoding the major endoplasmic reticulum chaperone protein in the
secretion of homologous and heterologous proteins in black
Aspergilli.Appl. Microbiol. Biotechnol.,50, 447–454.
46. Hargrove,T.Y., Wawrzak,Z., Lamb,D.C., Guengerich,F.P. and
Lepesheva,G.I. (2015) Structure-functional characterization of
cytochrome P450 Sterol 14-alpha-Demethylase (CYP51B) from
Aspergillus fumigatus and molecular basis for the development of
antifungal drugs. J. Biol. Chem.,290, 23916–23934.
47. Scharf,D.H., Heinekamp,T. and Brakhage,A.A. (2014) Human and
plant fungal Pathogens: The role of secondary metabolites. PLOS
Pathog,10, e1003859.
48. Hoang,H.D., Maruyama,J.I. and Kitamoto,K. (2015) Modulating
endoplasmic reticulum-Golgi cargo receptors for improving secretion
of carrier-fused heterologous proteins in the filamentous fungus
Aspergillus oryzae.Appl. Environ. Microbiol.,81, 533–543.
49. Dhingra,S. and Cramer,R.A. (2017) Regulation of sterol biosynthesis
in the human fungal pathogen Aspergillus fumigatus: Opportunities
for therapeutic development. Front. Microbiol.,8, 92.
50. Yasmin,S., Alcazar-Fuoli,L., Grundlinger,M., Puempel,T., Cairns,T.,
Blatzer,M., Lopez,J.F., Grimalt,J.O., Bignell,E., Haas,H. et al. (2012)
Mevalonate governs interdependency of ergosterol and siderophore
biosyntheses in the fungal pathogen Aspergillus fumigatus.Proc. Natl.
Acad. Sci. U.S.A.,109, E497–E504.
51. Andersen,M.R., Vongsangnak,W., Panagiotou,G., Salazar,M.P.,
Lehmann,L. and Nielsen,J. (2008) A trispecies Aspergillus
microarray: comparative transcriptomics of three Aspergillus species.
Proc. Natl. Acad. Sci.,105, 4387–4392.
52. Meyer,V. (2008) A small protein that fights fungi: AFP as a new
promising antifungal agent of biotechnological value. App. Microbiol.
Biotechnol.,78, 17–28.
53. Meyer,V., Wedde,M. and Stahl,U. (2001) Transcriptional regulation
of the antifungal protein in Aspergillus giganteus.Mol. Genet.
Genomics,266, 747–757.
54. Heged¨us,N., Sigl,C., Zadra,I., P´
ocsi,I. and Marx,F. (2011) The paf
gene product modulates asexual development in Penicillium
chrysogenum.J. Basic Microbiol.,51, 253–262.
55. Meyer,V. and Jung,S. (2018) Antifungal peptides of the AFP family
revisited: are these cannibal toxins? Microorganisms,6, E50.
56. Sheppard,D.C. (2005) The Aspergillus fumigatus StuA protein
governs the Up-Regulation of a discrete transcriptional program
during the acquisition of developmental competence. Mol. Biol. Cell,
16, 5866–5879.
57. Hu,P., Wang,Y., Zhou,J., Pan,Y. and Liu,G. (2015) AcstuA, which
encodes an APSES transcription regulator, is involved in conidiation,
cephalosporin biosynthesis and cell wall integrity of Acremonium
chrysogenum.Fungal Genet. Biol.,83, 26–40.
58. Clutterbuck,A.J. (1969) A mutational analysis of conidial
development in Aspergillus nidulans.Genetics,63, 317–327.
59. Newman,D.J. and Cragg,G.M. (2012) Natural products as sources of
new drugs over the 30 years from 1981 to 2010. J. Nat. Products,75,
311–335.
60. Nielsen,K.F., Mogensen,J.M., Johansen,M., Larsen,T.O. and
Frisvad,J.C. (2009) Review of secondary metabolites and mycotoxins
from the Aspergillus niger group. Anal. Bioanalyt. Chem.,395,
1225–1242.
61. Bayram, ¨
O., Krappmann,S., Ni,M., Jin,W.B., Helmstaedt,K.,
Valerius,O., Braus-Stromeyer,S., Kwon,N.J., Keller,N.P., Yu,J.H. et al.
(2008) VelB/VeA/LaeA complex coordinates light signal with fungal
development and secondary metabolism. Science,320, 1504–1506.
62. Liu,J., Osbourn,A. and Ma,P. (2015) MYB transcription factors as
regulators of phenylpropanoid metabolism in plants. Mol. Plant,8,
689–708.
Downloaded from https://academic.oup.com/nar/advance-article-abstract/doi/10.1093/nar/gky1183/5213907 by Technische Universitaet Berlin user on 18 January 2019
