Optimized De Novo Eriodictyol Biosynthesis in Streptomyces albidoflavus Using an Expansion of the Golden Standard Toolkit for Its Use in Actinomycetes

Eriodictyol is a hydroxylated flavonoid displaying multiple pharmaceutical activities, such as antitumoral, antiviral or neuroprotective. However, its industrial production is limited to extraction from plants due to its inherent limitations. Here, we present the generation of a Streptomyces albidoflavus bacterial factory edited at the genome level for an optimized de novo heterologous production of eriodictyol. For this purpose, an expansion of the Golden Standard toolkit (a Type IIS assembly method based on the Standard European Vector Architecture (SEVA)) has been created, encompassing a collection of synthetic biology modular vectors (adapted for their use in actinomycetes). These vectors have been designed for the assembly of transcriptional units and gene circuits in a plug-and-play manner, as well as for genome editing using CRISPR-Cas9-mediated genetic engineering. These vectors have been used for the optimization of the eriodictyol heterologous production levels in S. albidoflavus by enhancing the flavonoid-3′-hydroxylase (F3’H) activity (by means of a chimera design) and by replacing three native biosynthetic gene clusters in the bacterial chromosome with the plant genes matBC (involved in extracellular malonate uptake and its intracellular activation into malonyl-CoA), therefore allowing more malonyl-CoA to be devoted to the heterologous production of plant flavonoids in this bacterial factory. These experiments have allowed an increase in production of 1.8 times in the edited strain (where the three native biosynthetic gene clusters have been deleted) in comparison with the wild-type strain and a 13 times increase in eriodictyol overproduction in comparison with the non-chimaera version of the F3′H enzyme.

Eriodictyol is naturally found in citrus fruits, some vegetables and medicinal plants such as Eriodictyon californicum [19,20]. However, its extraction from plants presents some major drawbacks, such as its low concentration in plant tissues and the expensive purification and separation procedures from the complex plant extracts [21].
As a promising industrial alternative, the development of metabolic engineering and synthetic biology tools has facilitated the heterologous production of eriodictyol in microbial cell factories, such as Escherichia coli [21,22], yeast [23,24], Corynebacterium glutamicum [25] and Streptomyces albidoflavus [26]. For this purpose, the biosynthetic gene cluster (BGC) of the compound must be assembled using synthetic genes from the natural plant source, adapted after some modifications (such as codon optimization), which enable its expression in the selected microbial host.
In the case of eriodictyol, five genes are necessary for its biosynthesis: a tyrosine ammonia-lyase (TAL) converts L-tyrosine to p-coumaric acid (p-CA), which is subsequently activated into coumaroyl-CoA by a 4-coumarate-CoA ligase (4CL). Then, one molecule of coumaroyl-CoA is condensed with three molecules of malonyl-CoA by chalcone synthase (CHS) to give rise to naringenin chalcone, which is then isomerized into naringenin via chalcone isomerase (CHI). Finally, naringenin is converted to eriodictyol by a flavonoid 3hydroxylase (F3 H), which is a membrane-bound cytochrome P450 monooxygenase [25,26].
The major challenge in the microbial biosynthesis of eriodictyol and other flavonoids is their low production efficiency, normally due to the limited availability of precursors and cofactors [21]. In this respect, many strategies have been adopted in order to increase the malonyl-CoA supply, with numerous examples of metabolic engineering favoring the metabolic fluxes of central carbon metabolism pathways towards malonyl-CoA [21,27,28], including the heterologous expression of matBC from Rhizobium trifolii to increase cytosolic malonyl-CoA upon exogenous malonate feeding [29,30]. Apart from that, the biosynthesis of hydroxylated flavonoids such as eriodictyol involves another limitation: the low activity associated with plant P450-related enzymes expressed in bacterial hosts [26]. These monooxygenases are membrane-bound enzymes anchored [28] to the endoplasmic reticulum in plant cells, and usually they require an associated P450 reductase coworker. This issue has been addressed in E. coli by creating a chimeric protein containing both a P450 hydroxylase and a reductase without their respective membrane-binding regions [21].
Among all the aforementioned microorganisms used as microbial cell factories for these polyphenols, Streptomyces is the only organism, apart from plants and some fungi, known to be able to produce flavonoids naturally [31,32]. These Gram-positive bacteria are known for their ability to produce a plethora of secondary metabolites, including malonyl-CoA-derived compounds. Currently, there are many metabolic engineering strategies that have been successfully applied to Streptomyces to increase malonyl-CoA intracellular levels [33][34][35]. Of note is a strategy exclusive to this microbial host that consists of removing endogenous BGCs that use malonyl-CoA as a precursor as a way of channeling this cytosolic precursor toward the desired heterologously biosynthesized metabolite (e.g., a flavonoid) [36,37]. However, despite all these advantages, no reports have been found of metabolic engineering efforts made in Streptomyces in order to increase eriodictyol production.
In this work, a SEVA-based plasmid collection has been created for synthetic BGC assembly in Streptomyces and other actinomycetes using Golden Standard (GS) technology in a plug-and-play manner, as well as for host genomic editing using CRISPR-Cas9 techniques [38]. These tools have been applied for assembling the synthetic eriodictyol BGC into two individual integrative plasmids and for generating an edited version of the S. albidoflavus bacterial factory genome via CRISPR-Cas9. This genomic editing involved the replacement of three chromosomal BGCs (encoding malonyl-CoA-derived molecules) by P ermE* -matBC. Finally, eriodictyol de novo production has been tested in wild-type (WT) and edited strains, validating the new S. albidoflavus edited strain as an eriodictyol overproducer.

SEVA-Based Plasmid Library Design and Construction
A collection of SEVA modular shuttle vectors for E. coli and Streptomyces has already been developed by our group [39]. These plasmids are composed of four interchangeable modules: (1) origins of replication for E. coli and Streptomyces, (2) cargo, (3) antibiotic resistance marker, (4) origins of transfer. Each module is flanked by unique restriction sites for unusual enzymes. Thus, each plasmid or its derived construct can be easily repurposed by exchanging the corresponding module, when necessary, by restriction ligation. However, the SEVA shuttle plasmids developed in this work contain some important extra modifications. The oriT module has been modified in order to include the traJ gene to increase conjugation efficiency, and UNS sequences [40] have been added flanking each module to enable plasmid generation by Gibson assembly (GA) of the interchangeable parts. The unique restriction sites between modules have been conserved. In addition, a UNS sequence and a NheI unique restriction site have been added between the origins of replication for E. coli and Streptomyces in order to separate them into two different modules. A complete list of the plasmids comprising the library as well as the plasmid derivatives used in this study can be found in Table 1.  The present collection of plasmids is organized into plasmids for Streptomyces genome engineering using CRISPR-Cas9 and plasmids for BGC assembly using Golden Standard. CRISPR-Cas9-based genome engineering relies on the ability of the endonuclease Cas9 to cleave a target double-stranded DNA (dsDNA) guided by a synthetic guide RNA (sgRNA). Precise genome editing can be achieved through the introduction of a user-designed editing template including homologous recombination arms [41]. The Golden Standard technique combines the sequential use of different type IIS restriction enzymes, ordered fusion sites, and antibiotic resistance markers in plasmids of different levels to allow the systematic and hierarchical assembly of complete transcription units (TUs) (Level 1) and of multigene constructs, such as BGCs (Level 2), from basic premade standardized modules, such as promoters, rbs, coding sequences, terminators and others (Level 0 parts) [42].
The plasmids for the CRISPR-Cas9-mediated genome edition possess five UNS-flanked modules. Module 1 is flanked by UNS2 (unique nucleotide sequence 2) and UNS3 sequences and harbors the pSG5 origin of replication in Streptomyces [43] for easy plasmid curing after the genome edition. Module 2 is flanked by UNS3 and UNS4 sequences and contains pUC as the origin of replication in E. coli and a PacI-SpeI-flanked multicloning site (MCS) for cloning of the DNA repair template. Module 3 is flanked by UNS4 and UNS5 sequences and contains the CRISPR-Cas9 machinery (cas9-sgRNA) [44,45]. Based on previous studies [46], P ermE* was chosen to lead Cas9 expression. Two versions of Cas9 were generated: the wild-type Cas9, which cuts both DNA strains, and a Cas9 nickase (Cas9D10A), which has one nuclease domain deactivated and therefore only cuts one DNA strand. Module 4 is flanked by UNS5 and UNS1 sequences and consists of the resistance marker, which can be apramycin, ampicillin-thiostrepton, ampicillin-kanamycin or hygromycin. A rational choice of the antibiotic selection marker allows us to carry out CRISPR-Cas9-mediated genome modifications in any strain with Golden Standard plasmids already integrated into their chromosome. Module 5 is flanked by UNS1 and UNS2 sequences and contains the oriT-traJ operon ( Figure 1).
Level 1 and Level 2 Golden Standard plasmids possess four UNS-flanked modules. Module 1 is flanked by UNS2 and UNS3 sequences, and it harbors an integrase and its corresponding attP site for stable integration of the plasmid in the Streptomyces chromosome. There are three optional integrases: φC31, φBT1 and pSAM2 [49][50][51][52]. Module 2 is flanked by UNS3 and UNS5 sequences and consists of the pUC origin of replication for E. coli and a PacI-SpeI-flanked Golden Standard cargo region. The configuration of each Golden Standard cargo region can be found in the literature [38]. Module 3 is flanked by UNS5 and UNS1 sequences and contains the antibiotic resistance marker. There are five different resistance marker cassettes: apramycin, ampicillin-thiostrepton, gentamicin-thiostrepton, apramycin-ampicillin and hygromycin. These resistance markers have been combined with integrases and Golden Standard cargo regions to enable the selection of assembled plasmids in Golden Standard reactions as well as their integration into the Streptomyces chromosome. Module 4 is flanked by UNS1 and UNS2 sequences and harbors oriT and traJ ( Figure 1). A set of Level 1 and Level 2 Golden Standard plasmids displaying resistance to apramycin or ampicillin and to be integrated into the φC31 site were developed, including pSEVAUO-M11101 (cargo 19 ). Then, a set of plasmids to be integrated into φBT1 site were generated, including Plasmids for Golden Standard assembly were designed in three levels. Level 0 plasmids harbor basic parts, such as promoters (SP43, SP25, SF14 and ermE* [45][46][47]), a genetic insulator (RBS-RiboJ-SR41 [47]), coding sequences (TAL, 4CL, CHS, CHI and F3 H-CPR) and a terminator (ttsbib [48]). These parts were assembled in TUs using Level 1 plasmids as destination vectors, and from these into basic circuits, such as BGCs, when assembled into Level 2 plasmids. Level 0 Golden Standard plasmids are ampicillin-resistant (normally pSEVA181 [48]).
Level 1 and Level 2 Golden Standard plasmids possess four UNS-flanked modules. Module 1 is flanked by UNS2 and UNS3 sequences, and it harbors an integrase and its corresponding attP site for stable integration of the plasmid in the Streptomyces chromosome. There are three optional integrases: ϕC31, ϕBT1 and pSAM2 [49][50][51][52]. Module 2 is flanked by UNS3 and UNS5 sequences and consists of the pUC origin of replication for E. coli and a PacI-SpeI-flanked Golden Standard cargo region. The configuration of each Golden Standard cargo region can be found in the literature [38]. Module 3 is flanked by UNS5 and UNS1 sequences and contains the antibiotic resistance marker. There are five different resistance marker cassettes: apramycin, ampicillin-thiostrepton, gentamicin-thiostrepton, apramycinampicillin and hygromycin. These resistance markers have been combined with integrases and Golden Standard cargo regions to enable the selection of assembled plasmids in Golden Standard reactions as well as their integration into the Streptomyces chromosome. Module 4 is flanked by UNS1 and UNS2 sequences and harbors oriT and traJ ( Figure 1) Level 1 ϕBT1-apramycin Golden Standard plasmids could not be integrated into Streptomyces chromosomes together with ϕC31-ampicillin-apramycin plasmids, as they also possess apramycin as a selector marker. Therefore, a series of Level 1 and Level 2 ϕBT1 plasmids with resistance to gentamicin-thiostrepton were built. Gentamicin works as a selector marker for both Level 1 and Level 2 Golden Standard reactions in E. coli, whilst selection in Streptomyces is carried out with thiostrepton, allowing the right integration even in the pres-ence of other Golden Standard integrative plasmids. Level 1 ϕBT1-gentamicin-thiostrepton plasmids, pSEVAUO-M21104 (cargo 1AI2), pSEVAUO-M21204 (cargo 2AI3), pSEVAUO-M21304 (cargo 3AI4) and pSEVAUO-M21404 (cargo 4AI5), were generated individually by 4-fragment GA but using fragments already containing the corresponding Golden Standard cargos. Level 2 ϕBT1-gentamicin-thiostrepton plasmids pSEVAUO-M21504 (cargo A13B), pSEVAUO-M21604 (cargo B14C) and pSEVAUO-M21704 (cargo C15D) were generated by restriction ligation of the corresponding cargos into pSEVAUO-M21404. A third series of plasmids containing the pSAM2 integrase and encoding resistance to hygromycin was developed. This antibiotic serves as a selector in both E. coli and Streptomyces. The first plasmid of this series to be generated was pSEVAUO-M31705, which was assembled by a 4-fragment Gibson assembly using a fragment already containing cargo C15D. Plasmids pSEVAUO-M31105 (cargo 1AI2), pSEVAUO-M31205 (cargo 2AI3), pSEVAUO-M31305 (cargo 3AI4), pSEVAUO-M31405 (cargo 4AI5), pSEVAUO-M31505 (cargo A13B) and pSEVAUO-M31605 (cargo B14C) were generated by cloning the corresponding cargos into pSEVAUO-M31705.

Eriodictyol Heterologous Biosynthesis in S. albidoflavus WT
De novo eriodictyol (ERI) biosynthesis by S. albidoflavus was already achieved by our group, but titers were low [26]. This was associated with poor activity of the enzyme F3 H (flavonoid 3'-hydroxylase), which catalyzes the hydroxylation of naringenin at ring B to give rise to eriodictyol. Heterologous expression of this enzyme in prokaryotic systems is normally challenging due to its poor solubility and cofactor incorporation [21]. Therefore, the F3 H sequence was redesigned in order to encode a chimeric protein comprising a truncated F3 H fused to a truncated cytochrome P450 reductase (CPR), both proteins from Arabidopsis thaliana.
First, the plasmid pSEVAUO-M11701-NAR-BGC was generated. This is a ϕC31 integrative Level 2 Golden Standard plasmid harboring the four genes necessary for naringenin biosynthesis (TAL, 4CL, CHS and CHI), each of them with their own promoter, RiboJ-RBS and terminator. Promoters of increasing strength were assigned to each CDS according to the order in which their encoded enzyme acted on the naringenin (NAR) biosynthesis pathway (Figure 2). This plasmid was integrated into the ϕC31 chromosomal site of S. albidoflavus WT by protoplast transformation, giving rise to the strain S. albidoflavus WT-NAR. Then, pSEVAUO-M21206-F3 H-CPR, a ϕBT1 integrative Level 1 Golden Standard plasmid was assembled. This harbors the gene encoding the F3 H-CPR chimaera under the control of the SF14 promoter ( Figure 2). Integration of this plasmid into the ϕBT1 chromosomal site of the strain S. albidoflavus WT-NAR gave rise to the strain S. albidoflavus WT-ERI.
The strains S. albidoflavus WT-NAR, WT-ERI and wild-type (as a negative control) were cultured in an R5A liquid medium. De novo production of naringenin by the strains S. albidoflavus WT-NAR and S. albidoflavus WT-ERI and eriodictyol production by the strain S. albidoflavus WT-ERI were validated by comparison of differential peaks with pure standards in HPLC-HRESIMS (Figure 3).

Comparison of Conjugation and Genome Editing Efficiency between Wild-Type Cas9 and Cas9 Nickase
Cas9 expression has been described as toxic in Actinobacteria, but its toxic effects are circumvented by the proper selection of the promoter driving its expression [53]. Alternatively, a Cas9D10A nickase, which consists of a RuvC nuclease-defective Cas9, which introduces only a single-strand nick to the targeted DNA, has been adopted to overcome the same limitation in Clostridium cellulyticum [52]. Therefore, we have created a plasmid library with both the wild-type Cas9 and the Cas9 nickase and have tested their effectiveness in S. albidoflavus.
The conjugation and editing efficiency of both the wild-type Cas9 and the Cas9 nickase have been tested by activation of the S. albidoflavus endogenous BGC encoding for the blue compound indigoidine (IND). This BGC remains naturally silent in S. albidoflavus. However, when the constitutive promoter P ermE* is placed upstream of the first gene of the BGC, its expression is activated and colonies turn blue [53], enabling visual identification of recombinant strains. For this purpose, pSEVAUO-C41012-IND and pSEVAUO-C41022-IND have been constructed and separately introduced into S. albidoflavus WT by conjugation. The conjugation efficiency was measured as the number of colonies growing on a plate, and this was calculated as the mean of three independent experiments. The conjugation efficiency obtained with wild-type Cas9 was 195 ± 38 CFU, in contrast to 459 ± 78 CFU obtained with Cas9 nickase. The editing efficiency was calculated as the percentage of edited colonies out of 10 exconjugants selected from each conjugation and verified by PCR with the primers P ermE *-fw (annealing on P ermE* ) and P ermE* -IND-check-rev (annealing on the chromosome outside the fragment used as part of the DNA repair template) (Supplementary Table S3). In this case, the recombination efficiency was higher when the wild-type Cas9 was used (90%) compared to the Cas9 nickase (70%) (Figure 4). In light of these results, the plasmids bearing wild-type Cas9 were selected for subsequent experiments in S. albidoflavus. The strains S. albidoflavus WT-NAR, WT-ERI and wild-type (as a negative cont were cultured in an R5A liquid medium. De novo production of naringenin by the stra S. albidoflavus WT-NAR and S. albidoflavus WT-ERI and eriodictyol production by strain S. albidoflavus WT-ERI were validated by comparison of differential peaks with p standards in HPLC-HRESIMS (Figure 3).

Generation of the S. albidoflavus UO-FLAV-002 Edited Bacterial Factory
The CRISPR-Cas9-bearing plasmids were used to generate the S. albidoflavus UO-FLAV-002 bacterial factory in two steps: (1) the removal of chromosomal pseB4, a highly active pseudo-attB site for ϕC31 alternative integration; (2) the replacement of a series of native BGCs by the construction of P ermE* -matBC, in order to increase malonyl-CoA levels.
To direct the specific integration of ϕC31 integrative plasmids to only one site, the pseB4 pseudo-attB site was removed from the S. albidoflavus chromosome to ensure uniformity throughout the flavonoid production experiments. For this purpose, plasmid pSEVAUO-C41012-pSEB4 was introduced into the S. albidoflavus J1074 WT strain for pSEB4 replacement by a non-coding UNS8 sequence through CRISPR-Cas9-mediated recombination. The correct genome edition of the derived strain, S. albidoflavus UO-FLAV-001, was verified by PCR amplification (Supplementary Figure S1) and the sequencing of the obtained PCR products.
S. albidoflavus UO-FLAV-001 was further genetically modified to increase malonyl-CoA availability for flavonoid production. Two strategies were adopted for this purpose. Three biosynthetic gene clusters (BGC) encoding malonyl-CoA-derived metabolites, such as antimycins, candicidins and a predicted non-ribosomal peptide-polyketide, were removed from the S. albidoflavus chromosome and the matBC genes from R. trifolii were integrated into the chromosome under the control of the ermE* promoter. The operon matBC improves intracellular malonyl-CoA upon exogenous malonate addition to the culture, as MatC imports malonate into the cell and MatB converts this malonate into malonyl-CoA [30]. The final S. albidoflavus UO-FLAV-002 edited bacterial factory was therefore generated by a CRISPR-Cas9-mediated replacement of a 241,776 bp fragment (CP004370; DNA region from 6,576,725 to 6,818,501 bp) comprising the three mentioned BGCs (Cluster 22 henceforth) by P ermE* -matBC. To this end, plasmid pSEVAUO-C41012-C22-matBC was introduced into S. albidoflavus UO-FLAV-001 by conjugation. The correct genome edition was checked by PCR amplification (Supplementary Figure S2)

Comparison of Conjugation and Genome Editing Efficiency between Wild-Type Cas9 and Cas9 Nickase
Cas9 expression has been described as toxic in Actinobacteria, but its toxic effects are circumvented by the proper selection of the promoter driving its expression [53]. Alternatively, a Cas9D10A nickase, which consists of a RuvC nuclease-defective Cas9, which introduces only a single-strand nick to the targeted DNA, has been adopted to overcome of edited colonies out of 10 exconjugants selected from each conjugation and verified by PCR with the primers PermE*-fw (annealing on PermE*) and PermE*-IND-check-rev (annealing on the chromosome outside the fragment used as part of the DNA repair template) (Supplementary Table S3). In this case, the recombination efficiency was higher when the wildtype Cas9 was used (90%) compared to the Cas9 nickase (70%) (Figure 4). In light of these results, the plasmids bearing wild-type Cas9 were selected for subsequent experiments in S. albidoflavus.

Generation of the S. albidoflavus UO-FLAV-002 Edited Bacterial Factory
The CRISPR-Cas9-bearing plasmids were used to generate the S. albidoflavus UO-FLAV-002 bacterial factory in two steps: (1) the removal of chromosomal pseB4, a highly active pseudo-attB site for φC31 alternative integration; (2) the replacement of a series of native BGCs by the construction of PermE*-matBC, in order to increase malonyl-CoA levels.
To direct the specific integration of φC31 integrative plasmids to only one site, the pseB4 pseudo-attB site was removed from the S. albidoflavus chromosome to ensure uniformity throughout the flavonoid production experiments. For this purpose, plasmid pSE-VAUO-C41012-pSEB4 was introduced into the S. albidoflavus J1074 WT strain for pSEB4 replacement by a non-coding UNS8 sequence through CRISPR-Cas9-mediated recombination. The correct genome edition of the derived strain, S. albidoflavus UO-FLAV-001, was verified by PCR amplification (Supplementary Figure S1) and the sequencing of the obtained PCR products.
S. albidoflavus UO-FLAV-001 was further genetically modified to increase malonyl-CoA availability for flavonoid production. Two strategies were adopted for this purpose.

Eriodictyol Heterologous Biosynthesis in S. albidoflavus UO-FLAV-002
The performance of the S. albidoflavus UO-FLAV-002 edited strain as a flavonoidoverproducing bacterial factory was investigated by testing its eriodictyol heterologous production. First, the plasmid pSEVAUO-M11701-NAR-BGC was integrated into the ϕC31 site of S. albidoflavus UO-FLAV-002 by protoplast transformation to generate the S. albidoflavus UO-FLAV-002-NAR strain. This strain was further modified by the integration of the plasmid pSEVAUO-M21206-F3 H-CPR into its ϕBT1 chromosomal site, giving rise to S. albidoflavus UO-FLAV-002-ERI.
Then, the S. albidoflavus WT-ERI and S. albidoflavus UO-FLAV-002-ERI strains were cultured in R5A in order to test the impact on eriodictyol production after removing the mentioned three malonate-consuming BGCs. Additionally, the S. albidoflavus UO-FLAV-002-ERI strain was cultured in R5A with malonate at 20 mM in order to test the effect of the presence of MatBC on eriodictyol production.
Both parameters, the eriodictyol production and the biomass, were monitored every 24 h over a cultivation period of 120 h. Indeed, the S. albidoflavus UO-FLAV-002-ERI strain produced more eriodictyol than the WT strain, with a maximum of 0.06 mg/L yield 96 h after the inoculation. This constitutes a 3.44-fold compound increase over the maximum level produced by the WT strain (0.033 mg/L, 72 h after the inoculation) (Figure 5a). In terms of growth, the S. albidoflavus UO-FLAV-002-ERI strain grew slower than the WT strain at early time points but reached equal biomass at the end of the cultivation (Figure 5b). Both strains were also tested for naringenin production, and the S. albidoflavus UO-FLAV-002-ERI strain showed a 2.33-fold increase in its production (Figure 5c). Unfortunately, no major differences were observed between S. albidoflavus UO-FLAV-002-ERI cultured with and without malonate. This result is expected as malonyl-CoA was no longer a limiting factor to eriodictyol production in the given cultivation conditions.

Discussion
As in many other actinomycete bacteria, the S. albidoflavus chromosome contains a plethora of BGCs, which drive processes associated with the biosynthesis of a variety of native secondary metabolites, such as polyketides, terpenoids, siderophores, bacteriocins, and non-ribosomal peptides (NRPs) [53]. In the case of S. albidoflavus, a peripheral region of its chromosome (CP004370; 6,841,649 bp) [54], located between 6,576,725 and 6,818,501 bp, contains the gene clusters involved in the biosynthesis of candicidins, antimycins and an unknown hybrid polyketide-NRP compound [53], as predicted by the ANTISMASH software [55,56]. The biosynthetic pathways associated with these secondary metabolites share the use of high quantities of malonyl-CoA as one of their building blocks. For example, candicidins are antifungal type I polyketides using 14 malonyl-CoA building blocks per molecule (rendering acetate moieties in the final polyene backbone) [57]. In addition, antimycins are antifungal, antitumor and piscicidal depsipeptides generated by a hybrid polyketide and NRPS complex enzymatic machinery that uses, as biosynthetic precursors, L-Trp (to generate the starter unit 3-aminosalicylate) and diverse amino acids and carboxylic acids as extension units [58]. These include pyruvate, a precursor of acetyl-CoA (via pyruvate dehydrogenase) and therefore of malonyl-CoA biosynthesis [59][60][61]. The structure of the secondary metabolite generated by the third gene cluster in this chromosomal region (coding for a hybrid polyketide-NRP compound) is unknown, but this gene cluster was deleted as it may also require the consumption of malonyl-CoA (as this is the most common building block in polyketide moieties).
To achieve higher eriodictyol production yields in this bacterial factory, a deletion (using CRISPR-Cas9) of the previously mentioned chromosomal region was carried out,

Discussion
As in many other actinomycete bacteria, the S. albidoflavus chromosome contains a plethora of BGCs, which drive processes associated with the biosynthesis of a variety of native secondary metabolites, such as polyketides, terpenoids, siderophores, bacteriocins, and non-ribosomal peptides (NRPs) [53]. In the case of S. albidoflavus, a peripheral region of its chromosome (CP004370; 6,841,649 bp) [54], located between 6,576,725 and 6,818,501 bp, contains the gene clusters involved in the biosynthesis of candicidins, antimycins and an unknown hybrid polyketide-NRP compound [53], as predicted by the ANTISMASH software [55,56]. The biosynthetic pathways associated with these secondary metabolites share the use of high quantities of malonyl-CoA as one of their building blocks. For example, candicidins are antifungal type I polyketides using 14 malonyl-CoA building blocks per molecule (rendering acetate moieties in the final polyene backbone) [57]. In addition, antimycins are antifungal, antitumor and piscicidal depsipeptides generated by a hybrid polyketide and NRPS complex enzymatic machinery that uses, as biosynthetic precursors, L-Trp (to generate the starter unit 3-aminosalicylate) and diverse amino acids and carboxylic acids as extension units [58]. These include pyruvate, a precursor of acetyl-CoA (via pyruvate dehydrogenase) and therefore of malonyl-CoA biosynthesis [59][60][61]. The structure of the secondary metabolite generated by the third gene cluster in this chromosomal region (coding for a hybrid polyketide-NRP compound) is unknown, but this gene cluster was deleted as it may also require the consumption of malonyl-CoA (as this is the most common building block in polyketide moieties).
To achieve higher eriodictyol production yields in this bacterial factory, a deletion (using CRISPR-Cas9) of the previously mentioned chromosomal region was carried out, comprising about 242 kb of chromosomal DNA. The objective here was to divert the unused cytoplasmic malonyl-CoA in the new edited strain (S. albidoflavus UO-FLAV002) towards the biosynthesis of this flavanone: each molecule of eriodictyol uses three malonyl-CoA building blocks and one p-coumaroyl-CoA as precursors. In addition, a further step towards the enhancement of malonyl-CoA cytoplasmic pools was the insertion of the plant matBC genes from R. trifolii (under a constitutive promoter) during this chromosomal editing event. These two genes code for a dicarboxylate membrane transporter in charge of introducing malonate from the extracellular medium (MatC) and a malonyl-CoA synthetase, which intracellularly converts the imported malonate into the malonyl-CoA building block [30]. As a result of this genetic replacement in the S. albidoflavus chromosome, the strain S. albidoflavus UO-FLAV-002 was generated and tested to produce antimycins and candicidins by HPLC-HRESIMS. Both secondary metabolites were absent in the edited strain extracts, but a large quantity of m/z peaks corresponding to the structural diversity of antimycins was easily observable in the WT strain extracts (Supplementary Figure S3).
Additionally, in this work, the integration systems from the temperate phages ϕC31 and ϕBT1 and from the conjugative element pSAM2 have been used to allow the stable integration of the Golden Standard plasmids developed here into specific loci at the Streptomyces chromosome. Although they have a preferred chromosomal site of integration, sometimes these integrases might carry out the recombination event at alternative pseudo-attB sites in the target chromosome, but at a far lower efficiency [62]. In this sense, a highly active pseudo-attB site for the ϕC31 integrase has been identified in the S. albidoflavus chromosome. This pseudo-attB site, named pseB4, comprises a 50 nt sequence (CP004370; DNA region from 3,168,410 to 3,168,459 bp) located in the intergenic region between the XNR_2791 and XNR_2792 genes, and it exhibits a similar integration efficiency as the native attB [63]. This pseudo-attB site was therefore deleted (using CRISPR-Cas9 edition) to maintain a unique integration site along the bacterial factory chromosome for the ϕC31derived integrative plasmid vectors created, avoiding transcriptional level differences for the eriodictyol biosynthetic genes used (eventually derived from different chromosomal integration sites with variable transcription level capabilities) [64], as well as preventing eventual duplication of the heterologous gene cluster in the chromosome.
With the pseudo-attB site deleted from the bacterial chromosome, as well as the three mentioned biosynthetic gene clusters, the introduction of all the synthetic genes for the heterologous production of the flavanone eriodictyol (and its precursor naringenin) was achieved both in the WT and in the UO-FLAV-002 edited bacterial factory. The four genes necessary for naringenin biosynthesis were integrated into the canonical ϕC31 attB site, and the gene coding for the F3 H-CPR chimaera was integrated into the ϕBT1 attB site. With the modifications carried out in the edited bacterial factory, total naringenin production titers achieved a 2.33 increase with respect to the WT strain, and total eriodictyol yields reached a 3.44 rise with respect to the WT conditions; therefore, validating the fact that saving biosynthetic building blocks (in this case malonyl-CoA), due to the deletion of these three native gene clusters, is a reliable approach for enhancing final flavonoids heterologous production titers in this actinomycete. However, the same production experiments for eriodictyol biosynthesis, but adding 20 mM malonate to the R5A culture media, did not render an increase in the final production yields, probably indicating that, after deleting these three native gene clusters in S. albidoflavus, the intracellular levels of malonyl-CoA are no longer a limiting factor for the heterologous production of these flavonoids. This also indicates that, in the WT strain (with full native machinery for the biosynthesis of secondary metabolites such as antimycins and others), the addition of malonate to the culture media does not generate a net increase in the heterologously produced flavonoids.

Reagents and Biochemicals
All solvents used for solid phase extraction and HPLC-HRESIMS analysis were LC-MS grade from either Sigma-Aldrich (Madrid, Spain) or VWR (Barcelona, Spain).
Authentic standards of naringenin and eriodictyol for HPLC-HRESIMS quantification and molecule identification were provided by Extrasynthese (Genay, France).

Synthetic DNA and Enzymes
Recombinant DNA techniques were performed following standard protocols [67]. Restriction enzymes and T4 DNA ligase were purchased from Thermo Scientific (Madrid, Spain), NEBuilder®HiFi DNA Assembly Master Mix from New England BioLabs (Thermo Scientific, Madrid, Spain), Terra PCR Direct polymerase from Takara (DISMED, Gijón, Spain), and Herculase II Fusion DNA Polymerase from Agilent (Madrid, Spain). Synthetic genes outsourced from Explora Biotech (Venice, Italy) were received and cloned into PacI-SpeI sites of pSEVA181. All gene sequences were modified to include silent point mutations to remove restriction sites incompatible with Golden Standard and SEVA architecture and to meet the criteria to allow chemical synthesis when necessary [68]. For the antibiotic resistance cassettes, SanDI-UNS5-SwaI and PshAI-UNS1 sequences were added as flanking regions: pSEVA181Tsr contains the thiostrepton resistance cassette from pGM1190; pSEVA181Hyg contains the hygromycin resistance cassette from pOSV805. For the site-specific integrases, UNS3-NheI and FseI-UNS2 sequences were added as flanking regions: pSEVA181BT1int contains the ϕBT1 integrase and its attP site from pOSV805; pSEVA181pSAM2 contains the pSAM2 integrase and its attP site from pOSV807. For the Level 0 Golden Standard plasmids harboring promoter regions, BsaI-GGAG (Fusion site A) and TACT (Fusion site B)-BsaI sequences were added as flanking regions: pSEVA181P ermE *, pSEVA181SF14, pSEVA181SP25 and pSEVA181SP43 contain the P ermE* , SF14, SP25 and SP43 promoter sequences, respectively [69]. Plasmid pSEVA181RiboJ-RBS contains the insulator RiboJ [70] followed by the RBS sequence SR41 [49,69] flanked by BsaI-TACT (Fusion site B) and AATG (Fusion site D)-BsaI. For the Level 0 Golden Standard plasmids harboring coding sequences, BsaI-AATG (Fusion site D) and GCTT (Fusion site G)-BsaI were added as flanking regions: pSEVA181CHS contains a codon-optimized CHS gene from Glycine max (GenBank accession no. LT629807.1); pSEVA181CHI contains a codon-optimized CHI gene from Glycine max (GenBank accession no. LT629808.1); and pSEVA181F3H-CPR contains a codon-optimized F3 H-CPR chimaera from Arabidopsis thaliana (accession number: OQ674225). This fusion protein was designed as described elsewhere [21]: the sequence of a previously codon-optimized F3 H (GenBank accession no. LT629809.1) was modified in order to remove the first 21 amino acids corresponding to the transmembrane region. An ATG codon was added at 5 and a linker Gly-Ser-Thr at 3 . Consecutively, the nucleotide sequence encoding a truncated NADPH-cytochrome P450 reductase (CPR) was added. This sequence was obtained from the protein sequence of isoform 1 of gen ATR2 from A. thaliana (UniProt accession no. Q9SUM3-1) and, after codon optimization, nucleotides encoding the first 72 amino acids corresponding to the transmembrane region were removed. Plasmid pIDTSMARTttsbib was outsourced from IDT Technologies (Leuven, Belgium) and contains the terminator ttsbib [71] flanked by BsaI-GCTT (Fusion site G) and CGCT (Fusion site I)-BsaI. Plasmid pmatBC was de novo designed and contains codon-optimized matBC genes from R. trifolii, each preceded by an RBS sequence (Accession number: OQ659402). A 499 bp gBlock for cloning two protospacer sequences targeting Cluster 1 was designed following the corresponding protocol [44] and outsourced from IDT Technologies (Leuven, Belgium).

Construction of Plasmids
A detailed description of each plasmid construction can be found in the Supplementary Material (Methods for Plasmids Construction). Sequence of all primers used can be found in Supplementary Table S3. Gibson Isothermal Assembly (GA) was performed with NEBuilder®HiFi DNA Assembly Master Mix (New England BioLabs) (Thermo Scientific, Madrid, Spain), following the manufacturer's instructions.Golden Standard reactions were set up according to CIDAR MoClo [44], followed by blue-white screening by adding X-Gal to the medium. Protospacer design, annealing and cloning were performed as described for pCRISPomyces-2 [44]. Correct protospacer insertion was verified by sequencing with primer "Prot seq rev". Correct assembly of all constructs was first checked by restriction digestion and then confirmed by sequencing. The GenBank accession numbers for these plasmids are OQ696801 to OQ696836. Supplementary Table S1 contains information on all the plasmids used in this work. All these vectors' data will be available from the Golden Standard database after manuscript publication and can be requested directly from the SEVA siblings repository (http://sysbiol.cnb.csic.es/GoldenStandard/database.php, accessed on 1 May 2023). Figure 6 shows a basic schema with an example of the construction of a level 2 vector from the different parts depicted in four level 1 vectors (constructed on the basis of the sixteen possible units of level 0 vectors) ( Figure 6).

Strain Generation
Exogenous DNA was introduced into S. albidoflavus by either protoplast transformation or intergeneric conjugation [67]. Description of each strain generated by CRISPR-Cas9 and further verification can be found in Supplementary Figures S1 and S2, as well as information about the primers used for PCR amplification (Supplementary Table S3). Strains containing integrative plasmids were selected based on their resistance to the corresponding antibiotics. For the strains generated by CRISPR-Cas9 edition, colonies showing resistance to the corresponding antibiotics were re-streaked for colony isolation. Then, spores from an isolated colony were subjected to PCR verification with Terra polymerase to check both genome edition and contamination with DNA from the non-edited strain. When the positive recombinant colonies showed a mixture with non-edited DNA, the spores were filtered and spread on a plate to obtain isolated colonies with unique genotypes and checked again by PCR. Plasmid curing was achieved with consecutive passages without antibiotics and phenotypically checked by their loss of resistance to that antibiotic. Supplementary  Table S2 contains information on all the strains used in this work.

Flavonoid Extraction and HPLC-HRESIMS Analyses
The recovery of flavonoids from the recombinant strains developed in this study was achieved by organic extraction with acetone (cellular pellet) and ethyl acetate (culture supernatant). Briefly, 1 mL of culture medium was centrifuged at 10,000 rpm for 2 min, and the biomass as well as the supernatant were extracted separately. First, an equal volume of acetone was added to the pellet to facilitate the breaking of the mycelium, and then a second extraction was performed with an equal volume of ethyl acetate. For the supernatant, two extractions were performed with the same volume of ethyl acetate. In all cases, the extractions were subjected to vortex cycles to ensure contact with the solvent. Finally, the organic phases were dried sequentially under vacuum, and the residues were resuspended in 100 µL DMSO/MeOH 1:1 (v/v) for further analysis.
Analysis of flavonoids was performed using HPLC-HRESIMS. Separation was performed in a UPLC system (Dionex Ultimate 3000, ThermoScientific, Madrid, Spain) equipped with an analytical RP-18 HPLC column (50 × 2.1 mm, Zorbax ® Eclipse Plus, 1.8 µm, Agilent Technologies, Madrid, Spain) as previously described [71]. The obtained base peak chromatograms (BPCs) were extracted for the deprotonated ions of a set of flavonoids with a mass error range of 0.005 milli mass units (mmu) and the obtained extracted ion chromatograms (EICs) were compared with authentic commercial standards. When needed, flavonoids were quantified by comparing the peak area with that of a known amount of an authentic compound through a calibration curve. The production titers are given in mg/L, and the mean value was calculated from three biological replicates.