Efficient marker free CRISPR/Cas9 genome editing for functional analysis of gene families in filamentous fungi

Background CRISPR/Cas9 mediated genome editing has expedited the way of constructing multiple gene alterations in filamentous fungi, whereas traditional methods are time-consuming and can be of mutagenic nature. These developments allow the study of large gene families that contain putatively redundant genes, such as the seven-membered family of crh-genes encoding putative glucan–chitin crosslinking enzymes involved in cell wall biosynthesis. Results Here, we present a CRISPR/Cas9 system for Aspergillus niger using a non-integrative plasmid, containing a selection marker, a Cas9 and a sgRNA expression cassette. Combined with selection marker free knockout repair DNA fragments, a set of the seven single knockout strains was obtained through homology directed repair (HDR) with an average efficiency of 90%. Cas9–sgRNA plasmids could effectively be cured by removing selection pressure, allowing the use of the same selection marker in successive transformations. Moreover, we show that either two or even three separate Cas9–sgRNA plasmids combined with marker-free knockout repair DNA fragments can be used in a single transformation to obtain double or triple knockouts with 89% and 38% efficiency, respectively. By employing this technique, a seven-membered crh-gene family knockout strain was acquired in a few rounds of transformation; three times faster than integrative selection marker (pyrG) recycling transformations. An additional advantage of the use of marker-free gene editing is that negative effects of selection marker gene expression are evaded, as we observed in the case of disrupting virtually silent crh family members. Conclusions Our findings advocate the use of CRISPR/Cas9 to create multiple gene deletions in both a fast and reliable way, while simultaneously omitting possible locus-dependent-side-effects of poor auxotrophic marker expression.


Fungal Biology and Biotechnology
Page 2 of 13 van Leeuwe et al. Fungal Biol Biotechnol (2019) 6:13 between the outside world and the inside of the cell, providing structural integrity and protection from biotic and abiotic factors. Genome sequences of filamentous fungi, including Aspergillus niger, have shown that these cell wall related synthases and crosslinking enzymes often consist of large gene families. For example, the A. niger genome contains five Ags homologs, nine Chs homologs, seven Gas/GEL homologs and seven Crh-homologs [10]. The high number of genes in these families may correlate to the complexity of the multicellular, filamentous life style and offers the fungus to regulate the expression of these genes both in time and space during development or in response to stress. To perform functional analysis and determine possible redundancy of genes within a gene family, the construction of a single strain with multiple gene deletions is desirable.
Current methods to create multiple gene deletions in A. niger include the use of the pyrG or amdS-based transformation system combined with the subsequent recycling of pyrG or amdS via counter selection approach using 5-fluoroorotic acid (5-FOA) or 5-fluoro-acetamide (5-FAA) [11], respectively. Another approach is the use of multiple auxotrophic strains, but limits one to a total of four separate selection markers in A. niger (pyrG, nicB, argB, adeA) [12]. Dominant markers such as hygromycin or phleomycin resistance genes [13] can additionally be used. The recycling method is time consuming and the use of auxotrophic markers require supplementation when not all auxotrophic markers are used which can influence the growth phenotype of the strain, and both are therefore undesirable.
Alternatively, the CRISPR/Cas9 era has opened up the possibility to target and alter genes in an effective way with the potential to be selection marker free, hypothetically allowing limitless genetic alterations. For A. niger, several studies have been published to demonstrate the potential of CRISPR/Cas9, yet many still rely on the integration of selection marker in which the repair DNA fragment either remains integrated [14][15][16] or allows subsequent "pop-out" (pyrG) using 5-FOA counter selection [17,18]. Additionally, some studies have reported the use of marker-free deletion of single genes in one transformation, using repair DNA fragment(s) in combination with either integrative pUC-based plasmids [19] or plasmids with self-replicating extrachromosomal AMA1 elements [20,21]. As such, these developments have expedited the possibility to target multiple genes in a single transformation and thereby increase efficiency of strain construction. However, no studies have reported on the ability to multiplex knockouts of different genes in a single transformation without the use of integrative selection markers. To circumvent the time-consuming recycling of markers or the use of multiple auxotrophic strains to generate multiple gene deletion mutants, we demonstrate a marker-free CRISPR/Cas9-based transformation procedure. This procedure allows us to knock out multiple genes in one transformation in which the transformed strain can immediately be reused for subsequent transformations, if desired. Moreover, this marker-free approach excludes the possibility for "position effects" from genomic loci with low overall gene expression causing lack of marker gene expression [22][23][24][25]. Here, efficiency of multiplex knockout strain construction is demonstrated, using Cas9-sgRNA plasmids together with marker-free repair DNA fragments by knocking out the seven-membered cell wall chitin cross-linking gene family (crhA-G), without using an obvious selectable phenotype.

CRISPR/Cas9 plasmid design and proof of functionality
Our CRISPR/Cas9 procedure combines Cas9 expression driven by the constitutive Aspergillus nidulans tef1promoter from an autonomously replicating plasmid (pFC332; [32]) with expression of the sgRNA driven by the tRNA Pro1 RNAIII polymerase promoter [20]. The sgRNA expression cassette was amplified with extended primers to include PacI restriction sites on both ends of the cassette. A unique PacI restriction site in the pFC-series plasmids (pFC330-333) was used for ligation the sgRNA expression cassette into the vector (see "Methods").
To show functionality of the CRISPR/Cas9 system, we chose to target the brnA gene (NRRL3_01040; An14g05370) of A. niger which is homologous to Aspergillus fumigatus abr1 and encodes a multi copper oxidase that is involved in the synthesis of melanin [26,27]. A knockout of brnA results in the formation of browncolored spores, thus providing a direct read-out on transformation plates. A sgRNA target sequence for brnA was designed with CHOPCHOP (see "Methods"). Using fusion PCR, the brnA-target was cloned into a sgRNA expressing PCR fragment and ligated into pFC332 to yield pFC332_brnA-sgRNA (Table 1).
To test whether both cas9 and the brnA-sgRNA were expressed (and able to target the brnA locus), a series of transformations were performed in a non-homologousend-joining (NHEJ) deficient strain MA234.1 (∆kusA) ( Table 2). The ∆kusA (∆ku70 orthologue) mutant cannot repair a Cas9-induced double strand break (DSB) through NHEJ and thus relies on homology directed repair (HDR). As shown in Additional file 1: Figure S1A, the transformation of MA234.1 with pFC332_brnA-sgRNA did not yield in any viable transformants, whereas transformation with the control plasmid, lacking a sgRNA (pFC332, Additional file 1: Figure S1B), resulted in 100 viable, black transformants, indicating that the brnA-sgRNA generates a DSB which cannot be repaired leading to non-viable cells in a ∆kusA background. Cotransformation of homology-containing knockout-repair DNA fragment consisting of fused 5′-and 3′-flanks of the brnA gene, together with pFC332_brnA-sgRNA yielded 17/17 brown transformants on the transformation plates (Additional file 1: Figure S1C). Brown spores were visible across the entire colony for all transformants, suggesting that no sectors remained untransformed. Hence, no heterokaryotic colonies were observed on the initial transformation plate or found after single streaking. Single streaking of knockout-repair DNA fragment containing transformants on MM plates containing hygromycin showed brown coloration more clearly (Additional file 1: Figure S1E). In addition, a control co-transformation with plasmid pFC332 and knockout-repair DNA fragment did not yield brown colonies (0/63, Additional file 1: Figure  S1D), showing that repair DNA fragment does not integrate autonomously at the site of homology without the assistance of a Cas9-sgRNA targeting plasmid (Single streak shown in Additional file 1: Figure S1F for clarity of coloration). Both brown transformants (pFC332_brnA-sgRNA + knockout-repair DNA fragment) and black transformants (pFC332 + repair DNA fragment) were single streaked on MM without hygromycin as described in "Methods" to assess whether the transformants could lose the plasmid without selection pressure. About 80% of the transformants were confirmed to lose their plasmid (data not shown).
Genomic DNA (gDNA) of successfully plasmidcured transformants was isolated to genotype the brnA locus via a diagnostic PCR. A primer set was designed that prime outside of the repair DNA fragment (Additional file 2: Figure S2A). Amplification on gDNA of the wild type strain is expected to generate a PCR fragment of 4012 bp whereas the seamless ORF removal in the mutants is expected to generate a PCR fragment of 2411 bp. Additional file 2: Figure S2B shows the result of the diagnostic PCR. All brown transformants show a smaller PCR product at the expected size compared to both wild type strain and black transformants  (pFC332 + knockout-repair DNA). Therefore, we conclude that our combined system of the pFC332 plasmid and a Pro 1 -promoter expression driven sgRNA construct are efficient, and can be successfully used in future gene editing.

Multiplexing crh gene knockouts in successive transformation cycles
To assess the use of the multiplex CRISPR/Cas9 approach, we selected the seven-membered gene crh family. As shown in Table 3, this gene family consists of genes with a largely different expression levels, including both significantly expressed genes (e.g. crhB, crhC and crhD) and genes with very low expression levels (i.e. crhG) in different zones an A. niger colony under plate growth conditions grown on xylose [29]. Additionally, we looked into expression data of liquid fermentation conditions on both xylose and glucose to find that the expression of crh genes is independent of the carbon source. Also, under these growth conditions the expression of crhG was consistently low [30]. Expression of crhA and crhC also shows to be lower than during plate growth, whereas crhD is overall most highly expressed, followed by crhB and crhF (data not shown). To target the individual crh genes, we constructed seven Cas9-sgRNA plasmids (pFC332_crhA-G-sgRNA). Targets of each crh gene were designed using CHOP-CHOP and listed in Table 1. Crh targeting plasmids were co-transformed with PCR-amplified repair DNA fragments that consisted of a fused 5′ flank upstream of the crh ORF with a 3′ flank downstream of the crh ORF, similar as described for brnA. All transformations were performed in a ∆kusA background (MA234.1, Table 2).
Plasmid and repair DNA fragment combinations were transformed successfully to create single gene knockout mutants. Transformants were purified and cured of their plasmids through single streaks (see "Methods"). Knockouts were confirmed using diagnostic PCR, validating the functionality of all seven plasmids (Additional file 3: Figure S3). Out of thirty transformants tested, we observed twenty-seven with the correctly deleted gene; resulting in a frequency of repair DNA integration of 90% across seven independent transformations (Table 4). Transformation with plasmids pFC332_crhC-sgRNA, pFC332_crhD-sgRNA and pFC332_crhE-sgRNA and respective knockout-repair DNA fragments were both found to have a single tested transformant that did not display the deleted ORF (Table 4, Additional file 3: Figure S3). These transformants were likely able to evade the Cas9-sgRNA complex mediated DSBs, and were not investigated any further.
In addition, control sgRNA-plasmid transformations of the ∆kusA background without knockout-repair DNA fragments did not yield transformants for crhB, crhD, crhF and chrG. The crhA sgRNA-plasmid transformation showed 90% fewer transformants without knockout-repair DNA fragment compared to when the knockout-repair DNA fragment was added (data not shown). In only two cases (crhC and crhE), close to equal numbers (Table 4) of transformants were observed, either with or without knockout-repair DNA fragments, indicating that these guides were less efficient in creating DSBs. This type of control sgRNA-plasmid transformation without repair DNA fragment provides a good control for sgRNA functionality, and may help to predict the knockout efficiency among transformants.
In order to examine whether multiple genes could be deleted in a single transformation, we transformed the

Table 3 A. niger colony expression levels of crh genes in different zones grown on xylose
Seven-day old sandwiched colonies grown on xylose were used for RNA isolation and subsequent microarray analysis [29]. Distinct zones of the mycelium (zone 1, 3, and 5) harvested form the colony. Zone 1 represents the oldest or central part of the colony. Zone 5 represents the youngest or peripheral part of the colony, whereas zone 3 represents the intermediate zone.  cured ∆crhE strain (TLF61) with pFC332_crhD-sgRNA and pFC332_crhF-sgRNA (Table 1). Plasmids were transformed together with both knockout-repair DNA fragments for crhD and crhF and yielded eight colonies. Control transformation without any knockout-repair DNA fragment showed no colonies (data not shown). All eight transformants were plasmid-cured and were tested for the knockout of both crhD and crhF by diagnostic PCR. Six out of the eight transformants were found to have a double deletion of both crhD and crhF (Additional file 3: Figure S3). Subsequently, we continued using ∆crh-DEF (TLF65) for a successive round of transformation. Inspired by the successful transformation with two separate plasmids harboring the same hygromycin selection marker, we decided to perform both additional double and also triple gene targeting transformations. To construct quintuple mutants (∆crhABDEF) and (∆crhADEFG) or sextuple mutant (∆crhABDEFG), TLF65 (∆crhDEF) was co-transformed with respective pFC332_crh-sgRNA plasmids and corresponding knockout-repair DNA fragments. Simultaneous deletion of either crhA and crhB or crhA and crhG were both found to be 100% efficient in the tested transformants (crhA and crhB (7/7) and crhA and crhG (4/4), Additional file 3: Figure S3). Concurrent deletion of crhA, crhB and crhG resulted in eight transformants that were all tested for successful knockouts of all three genes. Three out of eight transformants were found to be ΔcrhA, ΔcrhB and ΔcrhG (∆crhABDEFG). In one transformant both ΔcrhB and ΔcrhG were deleted (∆crhBDEFG), whereas the four other transformants were deleted for ΔchrA (Additional file 3: Figure S3). The observed reduction in knockout efficiency among transformants marks a tipping point at which the attempt can be considered worthwhile. Note however that, despite a significantly lower efficiency score compared to double knockouts, the set-up with three individual gene targeting plasmids-each with identical selection markers and separate knockout-repair DNA fragment-remains within practically manageable numbers of strains to be analyzed.
Lastly, ∆crhABDEFG (TLF68) was transformed with pFC332_crhC-sgRNA and respective knockout-repair DNA fragment yielding a septuple ∆crhABCDEFG knockout (TLF39) of the whole crh gene family. A diagnostic PCR (primers used are listed in Additional file 4: Table S1) of the each crh gene shows a band size difference between wild type strain and TLF39 (Fig. 1a). Results of the diagnostic PCR of both the mutant (TLF39) and wild type (MA234.1) strains are shown in Fig. 1b, where all ORFs of the seven crh genes are removed resulting in a size difference.
All single, intermediate triple, quintuple and final septuple CRISPR/Cas9 derived crh knockouts were spotted on MM 1% glucose. As is evident from Fig. 2, there is no difference in growth between the different knockouts. However, there is a subtle difference in compactness of the colony morphology between the wild type and the septuple crh-knockout (Fig. 2a, b).

Crh gene knockouts via split marker transformations and effects of the integrated AOpyrG marker
At the time of setting up CRISPR/Cas9 based gene editing in A. niger, we used traditional split marker (i.e. bipartite) transformation to assess the effects of single gene deletions of the crh gene family. All seven single knockouts of crhA-G were obtained using the auxotrophic Aspergillus oryzae pyrG (AOpyrG) selection marker in a ∆kusA, pyrG-background (MA169.4, Table 2). To allow possible future transformation of single gene knockouts we opted to use direct-repeat (DR) split marker flanks required for looping-out the marker via counter selection (see "Methods"). However, synthesis of DR split marker flanks proved to be problematic for crhG and forced us to refrain from creating a ΔcrhG::DR-AOpyrG-DR type construct (data not shown) and to use a ΔcrhG::AOpyrG deletion construct instead. Transformant colonies were single streaked for purification, and genotyping through diagnostic PCR showed correct replacement of the ORF by AOpyrG for all seven genes (data not shown). During single streaking for strain purification it became evident that ΔcrhG::AOpyrG transformants already showed poor growth on MM (Additional file 5: Figure S4), whereas CRISPR/Cas9 constructed ∆crhG was not found to display a growth phenotype (Fig. 2b). Supplementation of the medium with 10 mM uridine (auxotrophic pyrG-supplementation) abolished the poor growth phenotype for ∆crhG::AOpyrG compared to wild type strain MA234.1 (Additional file 5: Figure S4), indicating that poor growth was due to insufficient levels of endogenously synthesized uridine/uracil.
All remaining single knockouts (∆crhA-F) were cultured to obtain fresh spore solutions and were spotted in equal spore numbers on MM with cell wall disturbing compounds Calcofluor White (CFW) and Congo Red (CR), to assess phenotypic effects (MM + CFW and MM + CR). Wild type MA234.1 and mutants were cultured under equal conditions (Additional file 6: Figure S5). No growth effects were found for ∆crhA::DR-AOpyrG-DR, ∆crhB::DR-AOpyrG-DR, ∆crhC::DR-AOpyrG-DR, ∆crhD::DR-AOpyrG-DR and ∆crhF::DR-AOpyrG-DR on both MM, MM + CFW and MM + CR. However, ∆crhE::DR-AOpyrG-DR only displayed a growth phenotype on MM containing either CFW or CR, but not on MM alone (Additional file 6: Figure S5). Similar to the situation of ∆crhG::AOpyrG, uridine supplementation to MM + CFW and MM + CR plates abolished the growth phenotype for ∆crhE::DR-AOpyrG-DR (Additional file 7: Figure S6). Set aside, our data show that single knockouts of crh genes do not affect growth or morphology per se, analogous to CRISPR/ Cas9 obtained single mutants.

Effects on cell wall sensitivity in crh-knockouts derived from CRISPR/Cas9 mediated transformation
To follow up on the phenotypic analysis of crh knockouts in bipartite-obtained transformants, CRISPR/ Cas9-derived transformants were also tested for cell wall sensitivity in accordance with yeast literature on CRH functionality [5], using MM with CFW or MM with CR. In addition to previously reported expression data on plates (Table 3), we checked expression of crh genes in the ΔugmA (UDP-galactopyranose mutase A) strain [31]. The ΔugmA strain is known to exhibit a constitutive state of cell wall stress and may therefore resemble conditions similar to cell wall stress induced by either CFW or CR. Expression data of the ΔugmA strain indicates a 2.91 log2 fold-change of crhE under cell wall stress conditions, whereas additional positive differential expression of both crhA and crhF and negative differential expression of crhB are noticeable, but not scored as significant (data not shown). Consequently, we postulate that crh genes are required in the cell wall stress response and that deletion mutants phenotypes reflect this in a cell wall stress assay with CFW or CR. Phenotypes on MM with either CFW or MM with CR of CRISPR/Cas9 derived single knockouts and the septuple knockout of crhA-G are shown in Fig. 2c-f. It is clear from the assay that both single knockouts and the septuple knockout of crhA-G do not show a disturbed growth phenotype on MM, and show no sensitivity towards either of the tested compounds. These data suggest that the knockoutsboth in singular cases and in septuple combination-do not affect the ability of A. niger to grow. Moreover, CrhA-G do not appear to be required for the survival when the cell wall is stressed with either CFW or CR.

Discussion
This study demonstrates an efficient CRISPR/Cas9 based gene-editing procedure for A. niger using a combination of the previously described methods [20,32]. Our  Table S1); PCR samples were loaded on 1% agarose gels approach combines the use of the pFC332 Cas9-AMA1 vector with the Pro 1 -sgRNA expression cassette. Due to a unique PacI site located on the pFC332 plasmid, newly synthesized sgRNA expression cassettes could easily be ligated to generate pFC332_sgRNA plasmids both quickly and effectively. These plasmids have been shown to generate either single, double or triple knockouts in combination with homologous knockout-repair DNA fragments in high frequencies, allowing fast and efficient construction of multiple gene knockouts in a markerfree way. Such gene alterations can be performed in any strain devoid of a hygromycin selection marker, and circumvents the reliance on auxotrophic mutant strains (e.g. AOpyrG-). Moreover, the curability of these hygromycin containing AMA1-vectors in the absence of selection pressure as an intrinsic property warrants the recyclability of this system for putative recurrent transformations, without the need to use of possibly mutagenic counter selection strategies. An often overlooked limitation of CRISPR/Cas9 based gene editing is that Cas9-sgRNA functionality is difficult to predict. In our approach, the use of the ΔkusA strain allows for sufficient assessment of this functionality; in the absence of a repair DNA fragment, an efficient sgRNA does not result in any transformants on the transformation plates due to the inability to repair a Cas9 induced DSB. This was observed for six out of the eight sgRNAs presented here. In two instances, transformants were observed when no knockout-repair DNA fragment was supplied, which may suggest less efficient generation of DSBs. However, in these two occasions the knockout efficiency was still 80-83% through integration of the knockout-repair DNA fragment (Table 4). These data show that sgRNAs are effectively expressed using our procedure.
As an alternative strategy to homology directed DNA repair editing, gene disruptions can be made in a kusA+ wild type strain without the addition of repair DNA fragment. Selection for indels is enforced by recurrent recognition and cutting by the Cas9-gRNA complex of the targeted DNA. As such, this would provide a more simplistic model for gene disruption. Song et al. [20] have previously shown the occurrence of differently sized indels in either fwnA (NRRL3_00462, An09g05730) or olvA (NRRL3_01039, An14g05350) in A. niger NRRL3, ranging from 88 bp insertions up to 1096 bp deletions across 30 individual mutants. In this study, while targeting the brnA gene, we observed even larger deletions that potentially range over 2 kb (data not shown). The DSB in brnA in the wild type strain N402 caused by pFC332_brnA-sgRNA lies 2.0 kb upstream from the olvA start codon. OlvA is epistatic over brnA in the melanin synthesis pathway [26,27], and a knockout of olvA becomes olive-colored. N402 transformed with pFC332_brnA-sgRNA occasionally (2.8%, n = 580) showed olive-colored transformants (data not shown), suggesting that deletions extend over previously reported 1 kb. As a result, we avoid this type of gene disruption due to the unpredictable outcome, and instead, highly recommend to perform gene edits with targeted repair DNA fragments in ΔkusA background. In addition to creating single knockouts with Cas9-sgRNA plasmids, we have shown to efficiently multiplex two or three knockouts using identical vectors (pFC332-Cas9 backbone), harboring separate gRNAs, in a single transformation. These results suggest that multiple, different AMA1-vector copies with the same selection marker can successfully create knockout combinations in A. niger. This positive result contradicts findings in A. oryzae where multiplexing with two different sgRNA on separate AMA1-vectors only resulted in single knockouts [33]. Only when both sgRNA constructs were cloned into a single vector, similar double knockout efficiencies observed as described here. It remains to be tested to what extent this number of simultaneous gene edits can be achieved in A. niger, but may ultimately be limited to the maximum copy number of approximately 10-20 AMA1 plasmids per nucleus [34].
The CRISPR/Cas9 procedure presented here provides evidence that multiplexing up to three targets does not require expression of three different sgRNAs on a single plasmid. High knockout efficiencies can be acquired by combining individual gene targeting plasmids with respective knockout-repair DNA fragments. This allows to combine different gene targeting plasmids as desired. Hence, existing plasmids can be used in various combinations without the need for reconstruction of complex cassettes, such as described by Nødvig et al. [21]. In addition, we show that multiple genetic alterations, such described for multiple gene knock-ins [17], do not need integrative selection markers to ensure efficient genome editing. This work has addressed the current limitations of creating marker-free multiplex knockouts of separate genes in A. niger, without the reliance on either integrative plasmids [14,19] or integrative (albeit recyclable) selection markers [15][16][17][18], and can directly be re-transformed for additional gene editing if desired. Therefore, it can be concluded that the use of singular sgRNA expression cassettes provides a flexible system in which both single, double and triple marker-free knockouts can efficiently be made.
CRISPR/Cas9 based genome editing is now a relevant technique in addition to conventional split marker transformations. The benefit of using CRISPR/Cas9 is multiplex ability, independence of integrative selection markers and the option to reuse the identical selection marker in follow-up gene-editing steps. Currently, recyclable systems do exist for pyrG and amdS in both CRISPR/Cas9 gene editing [17,18] and conventional split marker transformation [11]. This process can be induced with the counter-selectable compounds, either 5-FOA or 5-FAA, for pyrG and amdS markers respectively. Exposure to 5-FOA/FAA selects for either loop-out in case of direct repeat elements or loss (of function) of pyrG/ amdS, as the active gene metabolizes 5-FOA/FAA to a toxic compound. Consequently, the pyrG/amdS gene can be recycled to target a different GOI. In this way, any gene can be systematically knocked out in a relatively time consuming single-step fashion. In addition to being more labor intensive, successive use of 5-FOA/FAA may be undesirable due to their mutagenic nature. To put this in perspective, we estimate that construction of a septuple knockout strain can be obtained 3 times faster with CRISPR/Cas9 gene editing than split marker pyrG recycling due to the high efficiency of multiplexing and the omission of the counter selection procedure.
In addition to CRISPR/Cas9 obtained crh knockouts, single knockouts were constructed using the recyclable pyrG marker to test Crh functionality. We observed an effect from the locus of integration for the AOpyrG selection marker in case of crhG, which is most likely related to low baseline expression ( Table 3). Deletion of crhG via AOpyrG integration resulted in a poor growth phenotype on MM, and was shown to be complemented with the addition of uridine (Additional file 5: Figure S4). Additionally, unlike crhG, ∆crhE::DR-AOpyrG-DR shows normal growth on MM, but shows impaired growth on MM with either cell wall disturbing compounds CFW or CR. Cell wall stressing compounds CFW and CR are known disrupt fungal cell wall assembly [35], triggering the cell wall integrity (CWI) response [36]. In addition to chemical disruption, the CWI pathway is constitutively activated in the ΔugmA mutant, lacking cell wall galactofuranose [37]. This condition resembles exposure to CFW or CR, and transcriptomic analysis of the ΔugmA mutant showed upregulation of crhE, suggesting its importance during the CWI stress response. However, parallel phenotypic analysis of MA632.2 (∆crhE::DR-AOpyrG-DR) with the CRISPR/Cas9 derived TLF61 (∆crhE) showed no affected growth phenotype for TLF61 when exposed to either CFW or CR, whereas MA632.2 showed sensitivity toward both compounds. Only when the medium was supplemented with uridine, we observed the same wild type like phenotype as TLF61 (Additional file 7: Figure  S6). Despite the projected hypothesis on the requirement of crhE in the CWI response, these findings indicate an effect on the AOpyrG marker during the CWI response rather than the lack of CrhE. A possible explanation may be the locus activity of crhE (e.g. chromatin remodeling) during cell wall stress, affecting expression of ectopically integrated genes such as AOpyrG in MA632.2. Ectopic expression of auxotrophic markers have previously been described to be affected by the locus of integration. Specifically, the use of pyrG marker was reported to have a negative effect on either the ability to have a proper sexual cycle in A. nidulans [25] or during vegetative growth in Aspergillus flavus [38]. In case of A. nidulans, Robellet and colleagues clearly showed lower expression of pyrG at the alcS locus (required for ethanol utilization) compared to other ectopic integration sites of pyrG. This was tested in both alcS-locus inducing and non-inducing conditions, and may thus not only be related to an active or silent locus, but also where the position of the locus resides on the genome. These "position effects", where the expression of the selectable marker highly depends on the genetic elements at the locus of integration [39], have been also been reported for pyr4 (pyrG orthologue) in N. crassa, nadA and argB in A. nidulans [22][23][24] and URA3 (pyrG orthologue) in Candida albicans [40,41]. Taken together, these reports highlight that the expression of integrative selection markers, and specifically pyrG/pyr4/URA3, are sensitive to the locus of integration. In this study, we found that the auxotrophic AOpyrG selection marker is affected by the locus of integration. Therefore, caution must be taken in interpreting the phenotypic and/or pleiotropic effects that arise from this artifact in strains that harbor integrative selection markers. We propose to use either alternative, dominant selection markers such as hygromycin for construction of single knockouts or use non-integrative CRISPR/Cas9 selection procedures such as the one presented here.

Conclusions
We have demonstrated the efficiency of marker-free multiplex gene knockout construction, using Cas9-sgRNA plasmids with marker-free repair DNA fragments, in parallel to split marker fragment transformation to knockout the cell wall chitin cross-linking gene family (crhA-G). The use of multiple Cas9-sgRNA plasmids harboring the same selection marker can be achieved 3 times faster than AOpyrG recycling; showing that double or even triple knockouts are possible at relatively high efficiency using CRISPR/Cas9. Moreover, removal of selection medium allows the loss of Cas9-sgRNA plasmids while the gene knockout remains present. In turn, this grants the recurrent use of plasmids with the same selection marker in future transformations without prior need to recycle pyrG/amdS-type integrative selection markers. Additionally, it became evident that the expression of the integrated AOpyrG selection marker was affected in two of the seven crh knockout strains generated by a classical split marker approach compared to the same gene knockouts in CRISPR/Cas9 obtained mutants. Therefore, the marker-free CRISPR/Cas9 procedure presented here clearly favors over integrative selection markerbased transformations formultiplex knockout strain construction.
Strains were transformed after protoplastation as described previously [11]. We used 2 µg of Cas9-sgRNA plasmid with approximately 2 µg of repair DNA fragment (1.8-2.0 kbp) for each transformation. Transformation plates were incubated on MMS containing hygromycin (200 µg/mL) for 6 days at 30 °C. Transformed colonies were single streaked on MM containing hygromycin (100 µg/mL) to ensure nuclei of spores harbor the Cas9-sgRNA plasmid, thus are most likely to be transformed. Next, a single colony was picked and transferred to nonselective MM medium to allow loss of the Cas9-sgRNA plasmid. A third streak of a single colony on both MM containing hygromycin (100 µg/mL) and MM acts as a control for loss of plasmid. DNA from plasmid-cured strains was isolated as described by Arentshorst et al. [11], using mortar and pestle to grind the mycelium in liquid nitrogen.

CRISPR/Cas9 plasmid design
To express Cas9 and the guide RNA from the same autonomously replicating vector using hygromycin as a selection marker for fungal transformation, plasmid pFC332 was used [32]. pFC332 contains a unique PacI site which was used to insert a single guide RNA (sgRNA) expression cassette based on the native A. niger RNA polymerase type III Pro 1 -promoter and terminator [20]. The PCR strategy to generate sgRNA expression cassettes is schematically shown in Fig. 3. The strategy is based on designing overlapping PCR fragments containing the Pro 1 -promoter region followed by the reverse complemented sgRNA target, and a second PCR fragment containing the target followed by the sgRNA and terminator regions. Complementary target sequences of both PCR products allow fusion through PCR with outer primers (pTE1_for and pTE1_rev) to amplify the entire sgRNA expression cassette. These primers introduce a PacI site on either end of the construct for cloning into pFC332 (Fig. 3b).
The sgRNA targets were designed using the CHOP-CHOP web-tool [43]. Putative targets were obtained using the A. niger NRRL3 genomic sequence of the gene of interest (GOI) submitted in Fasta format with default CRISPR/Cas9 settings. The A. niger was used as reference genome for off-target matching. A selected, variable target sequence of 20 bp (without 5′ NGG 3′ PAM sequence) was added as overhang to target specific primer pTarget (Fig. 3a) in the identical forward orientation, whereas the reverse complement (RC) target sequence was added as overhang to target specific primer pRC-target (Fig. 3a).
The sgRNA expression cassette was acquired through fusion of two PCR products, the 5′ flank and the 3′ flank. pTE1_for (5′-ccttaattaaACT CCG CCG AAC GTA CTG -3′) was used in combination with pRCtarget on template plasmid pTLL108.1 to produce a 264 bp PacI::Pro 1 -promoter::target sequence (5′ flank). Combined primer pair pTarget and pTE1_rev (5′-cctt aatt aaA A A AG C A A A A A A G GA AG G TAC A A A AAAGC-3′) was used on template plasmid pTLL109.2 to create the 133 bp target::trRNA::Terminator::PacI sequence (3′ flank). After gel purification, the 5′ flank was subsequently fused to the 3′ flank, facilitated by the unique target as complementary sequence between the two flanks. Amplification of both flanks in one reaction with pTE1_for and pTE1_rev (Fig. 3b) results in a 397 bp sgRNA construct PacI::Pro 1 -promoter::target::t rRNA::Terminator::PacI (Fig. 3b) 3 Schematic representation of the pFC332_Pro 1 -sgRNA plasmid construction. a Amplification of the two flanks that represent the Pro 1 -sgRNA expression cassette: pTE1_for and pRC-target are used to amplify the Pro 1 -tRNA promoter and target sequence, where pRC-target contains a variable 20 bp overhang (indicated by brown color) that represents the reverse complement target sequence of choice. In turn, pTarget and pTE1_rev are used to amplify the target-sgRNA-Pro 1 -tRNA terminator flank. Here, pTarget contains a variable overhang that contains the target sequence of choice. Separate flanks are joined together through fusion PCR by pTE1_for and pTE1_rev, where the overhang sequence (=target) facilitates the homologous region between both flanks. b Addition of PacI sites to either end of the fusion construct (part of pTE1_for and pTE1_rev sequence) allows ligation of the fusion construct into pFC332. Diagnostic restriction analysis of the cloned plasmid ought to be done with Cfr42i (SacII) and shows a fragment of either 497 bp or 500 bp in addition to 1 kb and 14.3 kb fragments, for forward or reverse orientation, respectively Plasmid construction and repair DNA fragment design sgRNA constructs that were obtained through PCR were column purified (GeneJET PCR Purification Kit, Thermo Scientific ™ ) and digested with PacI overnight at 37 °C for 16 h and inactivated for 20 min. at 80 °C in a thermocycler (set heated lid to 37 °C). Digested samples were column purified and stored on ice prior to ligation. Approximately 1 µg of pFC332 was digested with PacI at 37 °C for 1 h and heat inactivated at 80 °C for 20 min. in a water bath, prior loading the entire sample on gel. Excision of linearized pFC332 plasmid (15.6 kb) followed by gel purification (GeneJET Gel purification Kit, Thermo Scientific ™ ) was eluted in 30 µL MQ. Linearized plasmid (17 µL) was dephosphorylated in a total volume of 20 µL with FastAP (FastAP Thermosensitive Alkaline Phosphatase (1 U/µL), Thermo Scientific ™ ) for 10 min. at 37 °C, followed by 10 min. inactivation at 75 °C. Samples were put directly on ice after inactivation. Treated samples were not purified prior to ligation: 3 µL plasmid together with 12 µL sgRNA construct was used in a ligation reaction for 10 min. (Rapid Ligation Kit, Thermo Scientific ™ ). Entire ligation mix was transformed to competent E. coli DH5α cells via heat shock protocol, and cells were plated on LB containing 100 µg/ mL ampicillin. Plasmids were isolated from successful transformants according to Miniprep protocol (Gene-JET Plasmid Miniprep Kit, Thermo Scientific ™ ). Diagnostic digest with Cfr42i (SacII) results in three bands of approximately, 14.3 kbp, 1.1 kbp and 500 bp in case of sgRNA integration, whereas in case of a control digest of pFC332 (or empty transformed vector), the 500 bp band will be absent. Confirmed plasmids were sent for Sanger sequencing with pTE1_for (Macrogen Europe, Amsterdam, The Netherlands).
Repair DNA fragments to create gene knockouts were based on a similar concept as producing split marker flanks in bipartite transformation [11]. Flanking regions on both 5′ and 3′ end around the gene of interest (GOI) were selected to be approximately 800-1000 bp each. The forward primer on the 3′ repair flank (Primer 3, Fig. 4) was provided with an overhang sequence which is the reverse complement sequence of the reverse primer on the 5′ end (Primer 2, Fig. 4). This created a 20 bp overlap between 5′ and 3′ repair flanks, required for fusion PCR (Fig. 4).

Cell wall sensitivity assays
Cell wall disturbing compounds Calcofluor White (CFW) and Congo Red (CR) were added in respective concentrations of 400 µg/mL and 800 µg/mL to MM plates [44].