Introduction

Tyrocidine is a ten amino acid long non-ribosomal peptide (NRP) that is produced by the Tyrocidine synthetase. This non-ribosomal synthetase (NRPS) is composed of ten modules with specificity for different proteinogenic and non-proteinogenic amino acids [1]. The Tyrocidine pathway can be found e.g. in Brevibacillus parabrevis. The gene cluster ( Fig. 1 ) consists of: tycA, tycB and tycC. Those genes encode for different numbers of modules - tycA is a single module, tycB is composed of three, tycC of six modules. Only the complex of modules is able to produce a NRP, whereas a single module is not completely functional independently [2].
Every module is again subdivided into domains. Among the most common domains are the Adenylation domain, the Condensation domain and the Thiolation domain. Beside, there are Epimerization domains, Thio-Esterase domains and Communication domains [3] [4] [5].

During the synthesis of non-ribosomal peptides, the growing peptide-chain is transferred from one module to the next. The domains within the modules fulfill distinct functions. An amino acid is first adenylated by the A domain and then bound to the T domain (also called Peptidyl-Carrier-Protein domain) via a thioester bond for subsequent reactions in the nascent NRP. The C domain then catalyzes the condensation of the already synthesized peptide chain (bound to the T domain) with the amino acid of the next module [7] [8]. The T domain itself does not exhibit any substrate specificity. Instead, it is merely a carrier domain to keep the peptide attached to the NRPS complex [4]. The core of every T domain is a conserved 4’-phosphopanthetheinylated (4’-PPT) Serine. The 4’-PPT residue is added by a 4’-PhosphoPanthetheinyl-Transferase (PPTase)activating the NRPS as a prosthetic group [9].

The remaining domains vary in their functions. Every NRPS is terminated by a TE domain that cleaves the thioester bond between the synthesized NRP and the last T domain [10]. E domains perform an epimerization reaction from the L- to the D-conformation or vice versa [11] [12]. Com domains are required for protein-protein interactions between subsequent modules that are not encoded on the same gene [13]. This is the case for the communication between the TycA and TycB1 module [14]. All of those six domain types mentioned above are present in the Tyrocidine synthetase of B. parabrevis [6].

In our experiments we also used the Indigoidine synthetase of Photorhabdus luminescens. Among others it is comprised of the A-Ox domain. In addition to the adenylation, this particular domain also oxidizes the amino acid [15] . Hence, the domain removes two hydrogen atoms such that a C-C double bond is formed. The oxidation leads in the case of glutamine, which is the substrate of the Indigoidine synthetase, to a cyclization.

Results

Claims

Here we show that modules of NRPSs can be interchanged resulting in engineered enzymes with novel functionalities. In the following, we have four claims to propose.

1. Modules of NRPSs can be interchanged.
2. The enzyme modularity allows the synthesis of custom peptides.
3. Customized NRPs can be detected and are functional.
4. Modules can be exchanged within one species and even between different species and/or pathways.

The establishment of a standard framework (RFC 100) for in-vivo synthesis of customized, novel NRPs by non-ribosomal peptide synthetases requires systematic investigation of NRPS modularity and compatibility. As proof of principle, we semi-rationally interchanged NRPS modules, a process we refer to as shuffling.

Shuffling modules of a single NRPS

Initially, we tried to interchange modules of the Tyrocidine cluster of B. parabrevis in E. coli. The following criteria were considered important to be investigated in this context:

1. The initiation module TycA can be located at any position within a NRPS, when placing it behind a suitable C domain.
2. Any other module can be used as an Initiation module, when removing the C domain.
3. Non-proteinogenic amino acids can be introduced at any position.
4. Com domain interactions can be replaced by suitable linkers.

Assembly of a synthetic NRPS

We started the amplification of the desired single modules for the assembly of various constructs required for proving the aforementioned criteria. Primers were annealed to the unconserved linker regions between the NRPS domains that were predicted by Pfam. We successfully validated the correct amplification of 12 single DNA fragments and corresponding pSB1C3 backbones by electrophoresis ( Fig. 2 ).

Their functionality as isolated modules cannot be shown, as they simply take up single amino acids without linking them to another monomer.
To show the compatibility of the Tyrocidine modules with one another, we put module genes into non-native order via Gibson Assembly. These constructs led to synthetic NRPSs and the production of five new peptides, i.e. one dipeptide, two tripeptides and two tetrapeptides.

We were able to successfully assemble all of our plasmids and continued our work with our synthetic Dipeptide NRPS and Tripeptide-I-NRPS ( Fig. 3 ) by transforming them into the E. coli strain BAPI.

Expression and detection of the NRPS and its products

The expression of the 212 kDa Dipeptide synthetase and the 380 kDa Tripeptide synthetase was shown by SDS-PAGE ( Fig. 4 ). Since SDS-PAGE is not sensitive enough to detect small peptides, we wanted to assess the presence of the newly synthesized peptides at different time points by the use of mass spectrometry. Expression of the NRPSs was induced with Isopropyl β-D-1-thiogalactopyranoside (IPTG). Samples were taken at different time points post induction. Since residual salts could potentially disturb the acquisition of small peptide abundances, we washed our LB-Cm culture in M9 minimal medium to minimize this effect and improve the detection of our short NRPs. Supernatant and the bacterial pellet were processed separately. Final sample preparation for tandem mass-spectrometry was conducted at the neonate screening facility of the university medical center. There the peptides were hydrolyzed during the butylation reaction. Finally the highly specific m/z profile allowed the identification of different amino acid abundances (Fig. 5) [16][17]. Since Ornithine is a non-proteinogenic amino acid that is incorporated in our Tripeptide, we mainly focused on detecting Ornithine levels. The abundance of Ornithine in the Tripeptide samples was strongly elevated with time compared to our Dipeptide (Fig. 6).

Due to variation of the general amino acid concentration in both used media and samples, we normalized the Ornithine values by the amount of all amino acids present in the respective medium ( Fig. 7 ). The Ornithine level in the supernatant of the Tripeptide samples peaked 21 hours upon induction. Afterwards concentrations returned to basal levels. Samples prepared from bacteria pellets showed minor increases in Ornithine levels in comparison to the pure medium and our negative control (untransformed BAPI). To aquire additional data on the existance of our short peptides we sent two samples to analysis via high-resolution electrospray ionization (HR-ESI) mass spectrometry at the mass spectrometry facility of the Institue for Chemistry, based on preliminary results from the Ornithine screening. However, we could not obtain conclusive data for our samples, because the background was too high (MS Results: Dipeptide and Fig. 8 ).

Fig. 5 Retraced spectrum of the amino acid screening for ornithine. A peak at 189 Da indicates enrichment of ornithine, which can be translated to quantitative measures by a conversion faktor deduced from a running standard.

Fig. 6 Levels of Ornithine measured at neonate screening via mass spectrometry. The supernatant containing the Tripeptide shows highly elevated levels of ornithine in comparison to the Dipeptide.

Fig. 7 Normalized levels of Ornithine measured at neonate screening via mass spectrometry. The supernatant containing the Tripeptide shows highly elevated levels of ornithine in comparison to the Dipeptide.

Fig. 8 Spectrum of the supernatant containing the Tripeptide Phe-Orn-Leu. A specific Peak at 392,49 Dalton has not been displayed due to insufficient purification.

In summary, we were able to amplify single modules from the Tyrocidine NRPS cluster, and we shuffled them via Gibson Assembly. Two constructs coding for two entirely new NRPSs were successfully transformed. The synthetases are both well expressed on pSB1C3 in BAPI and their products, short NRPs, can be detected through mass spectrometry finally confirming the functionality of the engineered NRPSs.

Showing inter-species module compatibility by fusion of Tyrocidine modules to the Indigoidine synthetase

We demonstrated interchangeability of Modules from NRPS within one species. To assess how preserved the functionality of different module combinations is, we investigated on the compatibility beyond species barriers. Therefore, the synthetase of the blue pigment Indigoidine from P. luminescens, which consists of only a single module, was fused to several modules of the Tyrocidine synthetase from B. Parabrevis to generate various new constructs.

Fusing single amino acids to Indigoidine

First, we combined single modules of the Tyrocidine synthetase with the indC synthetase resulting in three distinct NRPSs producing Asparagin, Valine or Phenylalanine respectively, all fused to Indigoidine ( Fig. 9 ).
To assure compatibility the constructs were designed in such a way that the C domain of the C2 module was always used, given its specificity for Glutamine required for the Indigoidine production. SDS-PAGE showed the expected bands for the expression of the NRP synthetases in the transformed BAPI.

The E. coli strain BAPI was used for expression of the NRPS fusions because it carries the required PPTase endogenously. All three of the fusion variants turned the colonies blue even before expression induction with IPTG. The blue pigment thus served a first indicator that peptide synthesis was successful. To further verify the existence of the fusion peptide, we ran comparative thin-layer chromatography (TLC). The peptides were detected under visible and UV light due to Indigoidine’s properties as a dye.

We delievered a sample of our purified Val-Ind NRP and purified Indigoidine to the mass spectrometry facility at the Institute for Chemistry handling these samples the same way as the samples sent from the Module Shuffling experiments. (Val-Ind MS and Negative Control Ind)

Apparently none of the spectra showed peaks at the desired positions ( Fig. 10 ). Although the sample showed a strong blue color, even after the prementioned purification steps, neither a peak for pure Indigoidine, nor for the target fusion products could be obtained. A slightly adopted work-up procedure for an Indigoidine-fused asparagine sample analyzed at the centre for molecular biology Heidelberg (ZMBH) did not lead to success either. Resulting in unspecific peaks as the val-ind sample mentioned before, the purification procedure did not seem to be optimized enough.

Experimental validation of software predictions

The NRPS Designer predicts module boundaries and linker regions based on Pfam. We used these predictions for our module shuffling experiments, which all worked well in our constructs. However we feel that the borders and linker regions predicted are rather vague and want to evaluate the predictions to contribute more data to the NRPS Designer. Therefore we systematically varied the module/ boundaries of the A domain and C domain of C2 within the Val-Ind-construct in relation to the Pfam prediction. Based on this altered domain positioning, eight constructs were designed combining narrow, broad and original Pfam module regions ( Fig. 11 ). Their functionality was always compared to the original construct based on the boundaries obtained from Pfam. We could induce the production of Val-Ind in two of our constructs: pLV03 (narrow and pfam border)and pLV08(broad and narrow border).

To summarize, with our inner- and interspecies module shuffling, we successfully validated the concept of modularity. We have integrated our experimental data and fuelled it into our software to improve its prediction accuracy. To make NRPSs and their custom design more accessible to the community, we have submitted standardized versions of the modules we used for our shuffling experiments, to the parts registry (see the parts registry entries).

Conclusion

We can prove all mentioned claims since we have:

1. Exchanged modules of of the Tyrocidine synthetase,
2. Constructed a custom dipeptide- and a custom tripeptide-NRPS,
3. Detected both the engineered NRPSs and the synthetic peptides via SDS-PAGE and mass spectrometry,
4. Demonstrated compatibility of modules from the Tyrocidine synthetase of B. parabrevis with modules from the Indigoidine synthetase of P. luminescens and

The achievements were enabled through and fed back to the NRPS Designer. This software tool incorporates a database utilizing our experimental on NRPS modules for guiding through user-oriented design of NRPSs.

Discussion

Module compatibility is the vital basis for any standardized work with NRPSs. Hence, the major objective of this project was to investigate flexible interchangeability of modules, which allows for customized synthesis of short peptides via NRPSs, as we propose in our standard (RFC 100). Tyrocidine served as a paradigm for semi-rational rearrangements in the modular structure of NRPSs, a process, which we refer to as shuffling.
Here, we present a clear line of evidence stating that it is possible to shuffle modules to produce functional NRPs. Modules of NRPSs can be interchanged by creating two different novel peptides through rearrangement of the respective modules that were amplified from B. parabrevis. The detection of these peptides was eased by the use of Ornithine, which is a non-proteinogenic amino acid and thus a proper marker for the synthesis of the desired peptide. Comparing the normalized levels of Ornithine in the different samples, we could conclude that the synthetic NRPS is in fact functional enabling the creation of customized peptides that can be detected via mass spectrometry.
Other approaches describe an extensive purification protocol [18], which requires large amounts of product, as high salt concentrations are an important interference factor. Small peptides are hardly specifically separable from salt-ions, hence high amounts could have bad influence on the signal at mass spectrometry. The method we describe above offers an useful and efficient way of detecting synthesis of the NRP.

Before our experiments, there was no evidence whether the synthetic peptides would be released to the medium or remained in the pellet. Showing that the resuspension of lyophilized supernatant in ethanol obtains a higher yield in ornithine content compared to the pellet samples, we can conclude that the small peptides are emitted into the medium. As salt concentrations did not seem to interfere with mass spectrometry measurements, the whole work up process has been successful. The final output of the tandem mass spectrometry demonstrated a highly elevated concentration of synthetic peptides in the medium compared to the cell interior, which leads to an improved protocol for standardized purification of short non-ribosomal peptides (RFC 100).

Furthermore, we found that there was no need to provide additional Ornithine in the media since it was incorporated from E. coli as an intermediate product from L-glutamate. Moreover, this assay accounted for the functional incorporation of non-proteinogenic amino acids into artificial non-ribosomal peptides. A high variety of non-proteinogenic amino acids as constituents has already been described in literature [19].

Interestingly, the Ornithine concentration in the samples peaked at the first day upon induction ( Fig. 5 ) but dropped rapidly during the second day to basal level. The transient enrichment of Ornithine could reflect the stability of the NRPS or the synthesized Tripeptide. Most likely Ornithine is cleaved by the endogenous enzyme ornithine decarboxylase encoded by the speC gene to form CO2 and Putrescine [20]. As ornithine abundance and acidic conditions serve as activators for the expression of the Ornithine decarboxylase respectively , this would explain the time-shift due to increased expression. Although this decarboxylase cannot cleave Ornithine incorporated in peptides, it could contribute to the decay process of free ornithine. A standardized test (MIO test) for the presence of Ornithine decarboxylase could help to determine the decay mechanism. In addition, the products could be toxic for the cells and are therefore degraded faster. The kinetics of Ornithine as a proxy for the production of the Tripeptide synthesis would thus imply that harvest of the peptide after one day would give optimal yield. Computational metabolic analysis could provide more insights into processes leading to peptide decay and instability. Metabolic reconstruction can give new insights to correlations at gene regulation and therefore provide new insights integrated in a biological context [21].

In the future, synthesized non-ribosomal peptides could be elongated even further. Therefore, a stepwise-assembly strategy should be considered since the complexity of cloning will increase with the number and size of DNA fragments as well. Of course, additional modules for shuffling will increase the amount of opportunities to create comprehensive constructs. Such an approach is guided by the NRPS Designer, which frames our proposed standard (RFC 100).
As we have seen, different modules of the Tyrocidine cluster are interchangeable with one another from the same synthetase. However, these findings only prove the compatibility of modules within a single pathway and a single species. To offer a general standard for short oligopeptide synthesis, we needed to establish synthesis of oligopeptides by NRPSs that were composed of modules originating from various pathways and organisms.

As we have seen, different modules of the Tyrocidine cluster are interchangeable with one another from the same synthetase. However, these findings only prove the compatibility of modules within a single pathway and a single species. To offer a general standard for short oligopeptide synthesis, we needed to establish synthesis of oligopeptides by NRPSs that were composed of modules originating from various pathways and organisms.

The synthetic composition of modules deriving from different species did result in a proper generation of a fusion peptide. The strong blue color of the inter-species fusion, a SDS-PAGES to detect the synthetases and a comparative TLC clearly account for a properly generated synthetase complex, which is able to assemble the target peptide correctly. Thus, we were able to reproduce the work of Marahiel et al. [22] and could open up the diversity of NRPS to the iGEM community. Proving intra-species and inter-species compatibility of different modules, we were able to confirm all four claims mentioned above.

Mass spectrometry often requires several tries and iterations of sample work up procedures to gain good results. Since the synthetic peptides assembled by our artificial nonribosomal peptide synthetases were rather small, their properties would not differ significantly from other salts in the solution. As we were able to cut salt concentrations significantly with our purification techniques, no problems of signal interference by the salty background occured at the tandem mass spectrometry of the neonate screening facility. The butylation reaction allows a highly specific compound detection making this method less prone to high backgrounds. This technique could be the reason for the proper detection compared to the other convenient mass sepctrometry measurements we performed. Therefore, an improved purification procedure could lead to the spectra we expected. Apparently, no peak for the specific mass of Indigoidine occured in the positive controle. As this is the same case for the inter-species combinations, all showing a strong blue color, other reasons could be considered, as well. In addition to contaminants disturbing a clear signal, a non-ionizability of our target compounds, several other adducts resulting from interactions with other remaining molecules or charge states, we did not anticipate could contribute to the inconclusive results. The fragmentation of molecules at the thought target mass did not display significant decay products. The results of the differential approaches were measured at two different facilities minimizing the risk of the mass spectrometre itself as an source of errors.

We ourselves thought of using methods such as Click-Chemistry for the purification of the short, synthetic peptides. Increased masses and easy cleavage enable high purities and can be determined via HPLC with UV and ESI-MS detection [23]. This approach, however, does not offer any opportunity to evaluate expression in vivo’ [24].