The biosynthesis of dalbaheptides is generally divided into three distinct stages (Yim et al., 2014b, 2016; Yushchuk et al., 2020b), that is, 1) generation of non-proteinogenic aa pool, further utilized in 2) non-ribosomal biosynthesis of the oligopeptide aglycones (coupled with the oxidative cross-linking); fully cross-linked aglycones are further 3) modified in a variety of tailoring steps. All dalbaheptide BGCs encode GTs, tailoring enzymes significantly contributing to the structural variety of these antibiotics (Nicolaou et al., 1999). Non-glycosylated dalbaheptide A47934 (from Streptomyces toyocaensis NRRL 15009) is the only exception here; consistently, the corresponding BGC lacks GT genes (Pootoolal et al., 2002).

More in detail, different steps might be defined in the glycosylation process of dalbaheptides, layer by layer “wrapping” the aglycone. The first step includes the biosynthesis of non-conventional sugar donors for aglycone decoration. Indeed, while some dalbaheptides are decorated with sugars deriving from primary metabolism (e.g., α-d-mannose and N-acetylglucosamine (GlcNAc) in teicoplanin or A40926), aglycones of vancomycin, balhimycin, and chloroeremomycin are decorated with the non-conventional monosaccharides l-vancosamine, l-4-oxovancosamine, and l-epivancosamine, respectively. In a similar manner to the biosynthesis of non-proteinogenic aa, enzymes required for the biosynthesis of such non-conventional monosaccharides are encoded within dalbaheptide BGCs. The second step (often the last one) consists of O-glycosylation of the aromatic aa forming the aglycone. In the third one, the installed sugars might be further modified in minor or major ways (e.g., α-d-mannose O-acetylation and GlcNAc deacetylation in A40926 biosynthesis).

Conventional sugars in GPAs‐from primary metabolism‐are d-mannose, d-glucose, d-arabinose, GlcNAc, and l-rhamnose. Non-conventional sugar residues include l-vancosamine, l-epivancosamine, l-4-oxovancosamine, l-ristosamine, and l-actinosamine. In addition to the aforementioned examples, l-vancosamine is present in Substitute with UK-68,597, while l-ristosamine is characteristic for ristocetin, and l-actinosamine for keratinimicin (Xu et al., 2019).

Biosynthesis of l-epivancosamine was initially studied in chloroeremomycin producer (van Wageningen et al., 1998). In vitro experiments (Chen et al., 2000; Kirkpatrick et al., 2000) demonstrated how five enzymes encoded within cep‐namely, EvaA-E‐transformed dTDP-4-oxo-6-deoxy-d-glucose into l-epivancosamine (Figure 1). Biosynthetic routes for aminosugars decorating other dalbaheptides were deduced from this model pathway. Initially activated substrate dTDP-4-oxo-6-deoxy-d-glucose is commonly derived from d-glucose-1-phosphate by the action of non-BGC-encoded enzymes (Figure 1). One notable exception is the UK-68,597 biosynthesis, since auk contains a gene for glucose 1-phosphate thymidylylransferase‐auk7 (Yim et al., 2014a)‐required for the D-glucose-1-phosphate activation. The presence of Auk7 likely positively contributes to the pool of l-vancosamine precursors in UK-68,597 biosynthesis.

vcm for vancomycin contains the orthologs of evaA-E genes, namely, vcaA-E (Xu et al., 2014) (Figure 1). The two sets of proteins are quite similar, sharing at least 75% of aa sequence identity (aa s.i.) in each pair (Xu et al., 2014). Nevertheless, minor differences in EvaE and VcaE seem to impact the function, making the first to convert dTDP-l-4-oxovancosamine into dTDP-l-epivancosamine, while the second yields dTDP-l-vancosamine (Figure 1). Instead, auk contains the orthologs of only evaA, evaB, evaC, and evaD (auk13, auk12, auk9, and auk8, respectively) (Yim et al., 2014a), lacking the ortholog of evaE: probably a functional homolog (or analog) of evaE resides outside the auk borders, finally contributing to l-vancosamine production. More understandable is the case of bal (Shawky et al., 2007), where evaA-E orthologs are named dvaA-E. Here, evaE ortholog dvaE is truncated, coding only the 99 aa C-terminal part of C4-ketoreductase (Donadio et al., 2005). Such truncated protein is non-functional, and consequently, the aminosugar biosynthesis terminates at the stage of dTDP-l-4-oxovancosamine (Figure 1). Biosynthetic pathways for l-ristosamine and l-actinosamine, coded within ris and ker, are more diverged. They both missed the EvaC ortholog (Spohn et al., 2014; Truman et al., 2014; Xu et al., 2019; Liu et al., 2021), resulting in the lack of a methyl group at the C3 position (Figure 1). Thus, biosynthesis of l-ristosamine might be attributed to the orthologs of EvaA, EvaB, EvaD, and EvaE‐Orf30, Orf32, Orf33, and Orf31, respectively (Truman et al., 2014). The same protein set is encoded within ker‐KraA-D (Xu et al., 2019). The major difference between l-actinosamine and l-ristosamine is O-methylation of the C4-position. It is still unknown how this methylation is achieved; however, the annotation of ker reveals the presence of a gene coding for an O-methyltransferase‐kerM (Xu et al., 2019). Since the aglycone of keratinimicin lacks any O-methylations, it seems plausible that KerM catalyzes the ultimate step of l-actinosamine biosynthesis (Figure 1).

In all cases described above the orthologs of evaA-E-B-D are most likely co-expressed, forming one operon, while the orthologs of evaC belong to a separate transcriptional unit (Shawky et al., 2007; Liu et al., 2021). auk is an exception, with auk8-9 (evaD-C) and auk12-13 (evaA-B) probably belonging to different operons (Yim et al., 2014a).

All GTs decorating aglycones of dalbaheptides belong to two families, according to the Carbohydrate-Active enZYmes Database (CAZy, http://www.cazy.org, Drula et al., 2022). GTs responsible for the installation of non-conventional aminosugars and conventional d-glucose, d-arabinose, GlcNAc, and l-rhamnose, belong to the GT1 family. These GTs require sugar substrates to be either dTDP- or UDP-activated, and share a unique two-domain structure (the so-called GT-B fold, Lairson et al., 2008), having C- and N-terminal Rossmann-like domains connected by a flexible linker region (Zhang et al., 2020). Recognition sites for the donor NDP-activated sugars are located at C-terminal domains (Chang et al., 2011), while N-terminal domains contain the acceptor binding site for dalbaheptide aglycone (Chang et al., 2011; Zhang et al., 2020). GTs of the second‐GT39‐family are responsible for the installation of d-mannose and require d-mannosyl-1-phosphoundecaprenol as a donor substrate. Large hydrophobic GT39-GTs are predicted as membrane-associated, having a GT-C fold with multiple transmembrane helices and intracellular active sites (Lairson et al., 2008).

In the aforementioned dalbaheptides, GT1-GTs attach sugar residues preferentially at AA-4 (4-hydroxyphenylglycine, Hpg) and AA-6 (β-hydroxytyrosine, Bht) of the aglycone or add additional monosaccharides to already existing mono/di/trisaccharides at AA-4. Fully cross-linked aglycones serve as acceptor substrates for GT1-GTs under physiological conditions, albeit some GT1-GTs were able to recognize partially cross-linked aglycones under certain experimental conditions, for instance, in mutasynthesis approaches (Weist et al., 2004; Butz et al., 2008). GT39-GTs attach mannose at AA-7 (3,5-dihydroxyphenylglycine, Dpg). The presence of multiple GTs within one pathway might result in the production of mixtures of related congeners, differing in glycosylation patterns. Dalbaheptides glycosylated at other AA positions were also described (Nicolaou et al., 1999), indicating that glycosylation might also occur at 1) AA-2 (Bht) and AA-1 (Hpg) in type II aglycone, and 2) AA-1 (Hpg) and AA-3 (Bht) in type III aglycone. Unfortunately, we currently lack any genomic information on the producers of these molecules, which would merit further investigations.

Type I dalbaheptides chloroeremomycin, balhimycin, and vancomycin served as the first models for experimental investigation of GT functions. Corresponding BGCs encode slightly different sets of GT1-GTs: GtfA, GtfB, and GtfC in cem; BgtfA, BgtfB, and BgtfC in bal; and GtfD and GtfE in vcm. Orthologous GTs GtfA and BgtfA install l-epivancosamine and l-4-oxovancosamine at AA-6 of chloroeremomycin and balhimycin, respectively. vcm does not encode GtfA ortholog, explaining why vancomycin is not glycosylated at AA-6. Peculiarly, this particular difference between vancomycin and chloroeremomycin seems to augment the antimicrobial activity of the latter, implying that l-epivancosamine at AA-6 facilitates cell wall binding (Nagarajan, 1993; Allen et al., 2002). GtfB, BgtfB, and GtfE are orthologs, glucosylating AA-4 in the biosynthesis of all three antibiotics (Figure 2A, Pelzer et al., 1999; Losey et al., 2001; Mulichak et al., 2001). Then, GtfC and GtfD attach l-epivancosamine or l-vancosamine to d-glucose at AA-4 in the biosynthesis of chloroeremomycin and vancomycin, respectively (Figure 2A, Losey et al., 2001, 2002; Mulichak et al., 2004). GtfC, GtfD, and BgtfC are orthologous proteins, but balhimycin lacks a disaccharide at AA-4, which could be found only in balhimycin V (a congener produced in residual amounts) (Pelzer et al., 1999; Stegmann et al., 2010). This might be due to the low affinity of BgtfC for the donor substrate‐l-4-oxovancosamine, produced as a consequence of DvaE mutation. More recently described type I dalbaheptide pekiskomycin is only glucosylated at AA-4, coherently with the only 1 GT encoded in pek: Pek28, which is a GtfB ortholog (Figure 2A) (Thaker et al., 2013).

So far, little is known about GTs decorating aglycones of type II dalbaheptides. ker (single type II BGC sequenced) carries three genes for GT1-GTs and one gene for GT39-GT (Xu et al., 2019). Among ker-encoded GT1-GTs, gtfA _ker , gtfB _ker , and gtfC _ker (ker was added to distinguish them from cep genes) are orthologs of gtfA, gtfB, and gtfC, respectively (Xu et al., 2019). Thus, GtfB_ker most likely installs d-glucose at AA-4, and l-rhamnose is then appended to d-glucose by GtfC_ker; this leaves GtfA_ker responsible for the attachment of l-actinosamine at AA-6 (Figure 2B). Single ker-encoded GT39-GT GtfD_ker most probably attaches d-mannose at AA-7 (Figure 2B).

More information is available on the GTs involved in the biosynthesis of type III dalbaheptides. Ristocetin BGC encodes 6 GTs, four of GT1 family and two belonging to the GT39 family (Truman et al., 2014). Phylogenetic reconstruction allowed to assign functions to 4 GTs, assuming that Orf16 attaches l-ristosamine at AA-6, Orf17‐d-glucose at AA-4, then appended with d-mannose by Orf34; finally, a second mannosyltransferase‐Orf22‐was expected to act at AA-7 (Figure 2C, Truman et al., 2014), as later confirmed by its heterologous expression in Streptomyces coelicolor carrying sta (Yim et al., 2016). Functions of Orf18 and Orf20 were not assigned, since these proteins were distantly related to known GTs (Truman et al., 2014). Another type III dalbaheptide‐UK-68,597‐is decorated with 2-L-vancosaminyl-d-glucose disaccharide at AA-4 (Yim et al., 2014a), while auk carries three genes for GT1-GTs. In vitro assay showed that Auk10 (GtfB ortholog) is responsible for glucosylation; in silico analysis then suggested l-vancosamine to be installed by Auk11 (GtfC ortholog, Figure 2C) (Yim et al., 2014a). Although the third GT, Auk14, is a GtfA ortholog, UK-68,597 lacks sugar residues at AA-6. In vitro assay suggested that Auk14 might be inactive or possess a very low affinity to substrates available in UK-68,597 biosynthesis, as observed with BgtfA (Yim et al., 2014a).

Glycosylation events that take place in type IV dalbaheptide biosynthesis are well defined (Figure 2D). tei and dbv BGCs encode three and two GTs, respectively. Accordingly, both antibiotics carry a GlcNAc moiety at AA-4 and d-mannose at AA-7, while teicoplanin aglycone is also decorated with another GlcNAc at AA-6. GtfB ortholog‐Tei10*‐installs the GlcNAc moiety at AA-4, as demonstrated from multiple in vivo and in vitro experiments (Li et al., 2004; Howard-Jones et al., 2007; Truman et al., 2009; Yushchuk et al., 2016, 2020b). Consistently, Dbv9‐a Tei10* ortholog‐is supposed to play the same role in A40926 glycosylation. Tei1 was shown to attach GlcNAc to teicoplanin aglycone at AA-6 (Li et al., 2004), whereas dbv lacks any Tei1 ortholog, explaining why A40926 has no sugars at AA-6. Finally, Tei3* was shown to be responsible for the decoration of teicoplanin aglycone with d-mannose at AA-7 (Yushchuk et al., 2016), implying that Dbv20 (Tei3* ortholog) has the same function in A40926 biosynthesis (Figure 2D). Interestingly, ramoplanin BGC (ramo), recently shown to be genetically related to tei (Waglechner et al., 2019), encodes a homolog of Tei3*‐Ramo29 (45% aa s.i., Chen et al., 2013). Ramoplanin is a clinically relevant peptide antibiotic produced by Actinoplanes ramoplaninifer ATCC 33076 (Marcone et al., 2017). Unlike dalbaheptides, it carries a 4-d-mannosyl-d-mannose disaccharide, instead of a single d-mannose residue. Ramo29 was shown to install the first d-mannose residue (Chen et al., 2013), but the GT responsible for the second mannosylation remains unknown. This merits further investigation, since such GT looks like a promising tool to further modify mannosylated dalbaheptides.

Concluding this section, it is interesting to report that genes for GTs tend to form one operon in type I-III BGCs (Shawky et al., 2007; Liu et al., 2021), being more scattered in type IV BGCs (Alduina et al., 2007; Yushchuk et al., 2019). It is also notable that genes for mannosyltransferases are present in different types of BGCs coming from distant actinobacterial lineages. This might indicate d-mannose residues at AA-7 to be an ancestral feature for all dalbaheptides.

Some further modifications of attached sugars occur during the biosynthesis of dalbaheptides, although they are quite rare. The first example comes from pekiskomycin, having d-glucose methylated. pek encodes two methyltransferases. One of them‐Pek30‐was experimentally shown to methylate the N-terminus of A47934 aglycone (Yim et al., 2016), leaving the other‐Pek16‐possibly responsible for d-glucose methylation (Figure 2A). Another notable modification is the acylation of AA-4 GlcNAc in type IV dalbaheptides, such as teicoplanin and A40926. To achieve this modification, GlcNAc at AA-4 is first deacetylated with orthologous deacetylases Tei2*/Dbv21 (Ho et al., 2006; Truman et al., 2006), and N-glucosamine is then acylated with orthologous acyltransferases Tei11*/Dbv8 (Figure 2D) (Li et al., 2004; Kruger et al., 2005; Howard-Jones et al., 2007; Yushchuk et al., 2016). Peculiarly, orthologs of Tei2*/Dbv21 are present in many (if not in all) BGCs for non-acylated type I-III dalbaheptides. The one from cep‐CepI‐was studied in vitro and shown to be inactive due to a single aa substitution (Truman et al., 2008). The omnipresence of tei2* orthologs in type I-III BGCs induces to speculate that the N-acylglucosamine moiety at AA-4 is an ancestral feature, lost or modified in many evolutionary lineages of dalbaheptides.

Modifications of A40926 sugars do not end with acylation. N-acylglucosamine moiety is further oxidized to N-acylaminoglucuronic acid group by Dbv29 (Figure 2D) (Li et al., 2007). The biological role of such modification is unclear, although it seems to reduce the A40926 antimicrobial activity (Malabarba et al., 1995). Peculiarly, noc BGC in N. coxensis DSM 45129 lacks an ortholog for dbv29, coding the biosynthesis of non-oxidized A40926 analog‐dalbaheptide A50926 (Yushchuk et al., 2021). Finally, the d-mannose residue at AA-7 of A40926 is O-acetylated with Dbv23 (Figure 2D) (Sosio et al., 2010). This modification is unstable and fades away in the alkaline extraction of A40926 (Alt et al., 2019); once again, its biological role is unclear, although it might be important for the regulation of antibiotic export and self-resistance (Alduina et al., 2020).