Engineering formation of multiple recombinant Eut protein nanocompartments in E. coli

Compartmentalization of designed metabolic pathways within protein based nanocompartments has the potential to increase reaction efficiency in multi-step biosynthetic reactions. We previously demonstrated proof-of-concept of this aim by targeting a functional enzyme to single cellular protein nanocompartments, which were formed upon recombinant expression of the Salmonella enterica LT2 ethanolamine utilization bacterial microcompartment shell proteins EutS or EutSMNLK in Escherichia coli. To optimize this system, increasing overall encapsulated enzyme reaction efficiency, factor(s) required for the production of more than one nanocompartment per cell must be identified. In this work we report that the cupin domain protein EutQ is required for assembly of more than one nanocompartment per cell. Overexpression of EutQ results in multiple nanocompartment assembly in our recombinant system. EutQ specifically interacts with the shell protein EutM in vitro via electrostatic interactions with the putative cytosolic face of EutM. These findings lead to the theory that EutQ could facilitate multiple nanocompartment biogenesis by serving as an assembly hub for shell proteins. This work offers insights into the biogenesis of Eut bacterial microcompartments, and also provides an improved platform for the production of protein based nanocompartments for targeted encapsulation of enzyme pathways.

Towards this aim, we recently developed a system for targeting cargo proteins to the interior of recombinant protein nanocompartments in the heterologous host E. coli C2566, using the Eut BMC shell proteins from S. enterica LT2. By expressing either the five shell proteins EutSMNLK, or the single shell protein EutS, under the control of our in-house pUCBB BioBrick ™ modified lac promoter plasmid system which allows constitutive expression 67 , a single stationary nanocompartment was formed per E. coli cell. Cargo proteins, including the fluorescent protein EGFP and a functional enzyme β -galactosidase, could be targeted to, and encapsulated within, the nanocompartments by tagging with a short N-terminal peptide sequence (EutC [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18][19] ) that we identified on one of the native encapsulated Eut metabolic enzymes (ethanolamine ammonia lyase complex β -subunit EutC) 39 . This represented a straightforward model system for the recombinant production of tailor designed nanobioreactors for in vivo biocatalysis applications.
One aspect of our model system that could be optimized to provide a more versatile platform for efficient multi-step catalysis is the number of recombinant compartments formed per cell in a heterologous host. The native organism, S. enterica LT2, produces multiple 68 , and sometimes unconstrained Eut BMCs per cell when grown on ethanolamine to induce BMC formation 39 . This is in contrast to the single immobile nanocompartment that is formed upon recombinant expression of Eut shell proteins in the heterologous host E. coli 39 . We hypothesized that another gene encoded within the eut gene cluster of S. enterica LT2 must be involved in the biogenesis of multiple unconstrained BMCs, which was not present in our heterologous system. The goal of this work was therefore to identify which genetic elements of the S. enterica LT2 eut gene cluster are required for the biogenesis of multiple nanocompartments in our recombinant expression system, and to control the formation of these multiple compartments in a heterologous host such as E. coli which, unlike the human pathogen associated S. enterica (strain LT2 is highly attenuated for virulence 69 and is therefore not recognized as a GRAS organism), is a widely used and accepted bacterial expression system for industrial biotechnology and biocatalysis. The results of this work offer valuable insights into the formation of Eut BMCs, as well as provide an optimized configuration for in vivo nanobioreactor assembly, which will serve as a basis for future biotechnological applications.

Results
Search for eut genes that may play a role in BMC biogenesis. Recombinant expression of the Eut shell proteins (EutS or EutSMNLK) in the heterologous host E. coli is not sufficient for multiple nanocompartment formation. To identify the genetic element(s) that may be required for the multiple 68 , unconstrained 39 Eut BMC phenotype that has been observed in S. enterica LT2, we carried out bioinformatics searches. The genes encoding enzymes necessary for ethanolamine utilization by S. enterica LT2 are encoded on the >15-kb eut operon. This gene cluster is composed of 17 open reading frames encoding regulatory elements; the metabolic enzymes necessary for breakdown of ethanolamine to acetyl CoA, acetyl phosphate and ethanol; and the five BMC shell proteins EutSMNLK (reviewed in: 14 ). Many of the genes in the eut cluster have been fully characterized by other groups, and have not been described as being required for multiple Eut BMC formation (see 14 for details), ruling out their potential involvement in multiple BMC biogenesis. However, the potential role of the proteins encoded by three of the genes in the eut operon in biogenesis remain unclear, EutJ, EutP and EutQ.
EutJ (NP_461397.1), which is annotated as a putative chaperone-like protein similar to DnaK or Hsp70 70 , is not essential for growth of S. enterica LT2 on ethanolamine 71 . The function of EutJ is unknown. EutP (NP_461404.1) belongs to the EutP/PduV family and shares sequence homology with other members of the RAS-like GTPase superfamily (PF00071); which is a diverse family of small GTPases that are involved in signal transduction 72 . The N-terminal region of the EutP homolog PduV has been reported to target proteins to the outer surface of PduABJKNU BMCs heterologously expressed in E. coli 62 . EutP may play a similar role in targeting proteins to Eut BMCs, however, this remains to be verified. Finally, it has been suggested that EutQ (NP_461403.1) may play an unidentified role in enclosing the dehydrogenase EutE within S. enterica LT2 BMCs 73 . The crystal structure of an N-terminally truncated EutQ homolog, CD1925 from Clostridium difficile 630, reveals that the protein has a cupin barrel fold (members of this superfamily have diverse functions as dioxygenases, decarboxylases, isomerases or in small molecule binding 74 ), but that it lacks the typical metal coordinating histidines at the putative ligand binding site 25 . Therefore from a structural point of view the function of EutQ remains uncertain. A study by Penrod and Roth 71 concluded that deletion of eutP or eutQ from S. enterica LT2 does not affect overall growth on ethanolamine, but does result in an increased acetaldehyde release during growth on this carbon source. This would suggest that these two proteins may be involved in the formation of fully functional Eut compartments in S. enterica LT2. Very recently, while this manuscript was under review, in vitro studies revealed that EutP and EutQ display novel acetate kinase activity, despite both proteins displaying significant differences in size and sequence to other previously characterized acetate kinases. It appears from this study that the acetate kinase activity of EutQ is not essential for growth of S. enterica Typhimurium on ethanolamine 75 .
To explore whether any of the proteins EutJ, EutP, or EutQ could be involved in multiple nanocompartment biogenesis, we cloned and expressed them as part of our recombinant system in E. coli.  Figure 1). Subsequently, we stacked EutJ, EutP or EutQ with the S. enterica LT2 shell proteins EutS or EutSMNLK on our pUCBB plasmid system (Supplementary Table 1). To observe the effect of EutJ, EutP and EutQ on EutS nanocompartment assembly in vivo, we coexpressed the constructs in E. coli C2566 (note that this lab strain contains a non-functional eut operon and is not capable of growth on ethanolamine (Supplementary Figure 2), and does not produce BMCs unless shell proteins are recombinantly overexpressed 39 ), with our previously characterized targeting sequence-fluorescent reporter construct, EutC 1-19 -EGFP (on our in-house low copy number constitutive plasmid pACBB) 39,67 . This system allows for the straightforward and rapid confirmation of nanocompartment assembly by visualization of discrete fluorescent puncta by fluorescent microscopy.
E. coli C2566 cells coexpressing the shell protein EutS and the EutC 1-19 -EGFP fusion formed nanocompartments, which we observed as a single GFP focus per cell, slightly off-center or close to the pole in 85% of cells ( Fig. 1, Supplementary Figure 3). This phenotype is in agreement with our previously published results 39 . A similar phenotype was apparent in construct EutJS, indicating that EutJ has no effect on EutS nanocompartment assembly (Fig. 1, Supplementary Figure 3). However, we noted a significant difference in the fluorescence signal of cells expressing EutPS. Aggregates of GFP formed at both poles in most of the cells coexpressing EutPS and EutC 1-19 -EGFP, with only a small number of cells (9%) forming nanocompartment-like foci as a third fluorescent punctate visible in the middle of the cell. The EutC 1-19 signal sequence was required for this phenotype to be observed (Fig. 1, Supplementary Figure 3). This suggests that EutS is no longer capable of correctly forming nanocompartments in the presence of EutP. Control experiments indicate that EutP alone does not interact with EGFP or EutC 1-19 -EGFP; all cells expressing these constructs had a diffuse GFP signal (Supplementary Figure 4), negating the possibility that EutP is interacting with the cargo protein non-specifically and preventing encapsulation. Finally, cells expressing EutQS also lost the typical nanocompartment phenotype; cells were small and rounded indicating that they were stressed, and each cell had a single large aggregate of GFP close to or located at the pole of the cell (Fig. 1, Supplementary Figure 3). The same phenotype was observed when no EutS shell protein was present or when no EutC 1-19 signal sequences was fused to EGFP, suggesting that expression of EutQ causes EGFP to aggregate (Supplementary Figure 4).
These results indicated that coexpression of EutP, and perhaps EutQ, with the single shell protein EutS was potentially disrupting or preventing correct nanocompartment assembly. To determine whether the four other shell proteins (EutMNLK) were required to facilitate correct formation of nanocompartments in the presence of EutP and EutQ, we stacked EutJ, EutP and EutQ with EutSMNLK (Supplementary Table 1). E. coli C2566 cells expressing EutSMNLK and EutC 1-19 -EGFP formed a single nanocompartment per cell (80% of cells), slightly off-center or close to the pole (Fig. 2, Supplementary Figure 5), as previously described 39 . EutJ had no effect on EutSMNLK nanocompartment assembly, 73% of the cells had a single nanocompartment (  Figure 5), indicating that EutP could be preventing or disrupting EutSMNLK nanocompartment assembly. This result corroborates our findings with EutPS where large polar GFP aggregates formed, also indicating a loss of proper compartment formation. Significantly, coexpression of EutQSMNLK with EutC 1-19 -EGFP resulted in multiple GFP foci distributed across the cytoplasm throughout the entire intracellular space, in 87% of the cells visualized (Fig. 2, Supplementary Figure 5). This was in contrast to just the one or two very small GFP spots localized to the cell wall in a small percentage of the EutPSMNLK cells. These fluorescent puncta in EutQSMNLK cells were only observed when EGFP was fused to the EutC 1-19 signal sequence, and were smaller than the single foci observed in the case of EutS or EutSMNLK without co-expression of EutQ (Figs 1 and 2). We also noted that some of the fluorescent puncta in some of the E. coli cells expressing EutQSMNLK appeared to be slightly unconstrained in cellular space (Supplementary Movie 1), which was in contrast to the single immobile GFP foci that we observed in E. coli cells expressing EutSMNLK (Supplementary Movie 2). These results pointed to EutQ potentially being responsible for multiple nanocompartment formation in a heterologous host, a finding which may have implications for understanding Eut BMC biogenesis in S. enterica LT2.

EutQ may be involved in Eut BMC biogenesis in the native host S. enterica LT2. To verify
whether EutQ is potentially involved in Eut BMC biogenesis and mobility, we explored the role of EutQ in the native host S. enterica LT2, by examining the formation of Eut BMCs in a EutQ deletion strain (strain TT24802 eutQ370∆ ::FRT (sw) containing a truncated version of eutQ with 40 bps of both the 5′ and 3′ ends of the eutQ Contrastingly, when the ∆ EutQ strain was grown on ethanolamine to induce Eut BMC formation, a single immobile GFP punctum was observed at or near the pole of the cells (Fig. 3B, Supplementary Movie 4). It is not entirely clear from the fluorescent images whether these puncta represent single BMCs or also show aggregated GFP, however all five Eut BMC shell proteins are produced under these growth conditions (Supplementary Figure 6), and thin cell sectioning and TEM (Fig. 3B) revealed a single Eut BMC-like structure per cell (highlighted with arrows). These electron dense zones surrounded by an electron diffuse zone seen in Fig. 3A,B are similar to those observed by others in S. enterica LT2 grown on Eut BMC inducing conditions 68 , and those that we previously observed in S. enterica LT2 and our heterologous expression system 39 . Together these findings suggest that in contrast to the wild-type S. enterica LT2 strain, the ∆ EutQ strain is producing a single BMC per cell, implying that EutQ could be required for the biogenesis of multiple, unconstrained BMCs in both the native host S. enterica LT2 and the heterologous host E. coli.

Cellular structures formed during recombinant expression of EutJ, EutP, and EutQ in E.
coli. We were surprised that EutS nanocompartments did not form correctly in the presence of EutQ, which induces multiple nanocompartment formation with EutSMNLK in our heterologous expression system. Furthermore, EutP appeared to be inhibiting correct nanocompartment formation (Fig. 2). Notably, Ras-like GTPases (like EutP) are sometimes associated with cytoskeletal elements and may form higher order structures in vivo 76 . Therefore, to explore whether EutQ, or either of the other proteins EutJ and EutP were forming large protein structures that could inhibit EutS nanocompartment formation, E. coli C2566 cells expressing all of the constructs were thin cell sectioned and visualized by TEM.
Cells overexpressing EutJ did not form any visible structures (Fig. 4, Supplementary Figure 7), indicating that it does not self-assemble as a large protein complex. These findings serve to support the fact that EutJ expression has Contrastingly, EutP self-assembled as elongated protein sheets (rectangular electron dense zones up to 500 nm in length running parallel to the axis of the cell) that also appeared folded up to give angular structures > 300 nm in diameter that were offset to the axis of the cell (in 68% of cell sections) (Fig. 4, Supplementary Figure 7). It could be hypothesized that these large assemblies may potentially inhibit correct nanocompartment formation, explaining the lack of nanocompartment-like fluorescent puncta in cells expressing EutPS and EutPSMNLK  Supplementary Figure 7), we examined cells overexpressing EutJ, EutP, and EutQ with the shell proteins EutS and EutSMNLK for higher order assemblies. BMC-like structures were observed in positive control cells overexpressing the shell proteins EutS or EutSMNLK alone (visible in 32% and 36%, respectively, of cell sections) (Fig. 4, Supplementary Figures 7 and 9). The polyhedral structures typically appear as angular electron dense regions off-center in the cellular cytosol, with a diameter of 100 nm in the case of EutS and 150 nm in the case of EutSMNLK. These structures are sometimes surrounded by an electron diffuse zone, similar to what we have previously described 39 . We do not observe these structures in E. coli C2566 cells that are not overexpressing the shell proteins (Fig. 4, Supplementary Figure 7). Note that the percentage of cells with BMC-like structures observed by TEM is typically significantly lower than that observed by fluorescent microscope (Figs 1 and 2, Supplementary Figures 3 and 5) due to the nature of cell sectioning and the subsequent visualization procedure. Interestingly, cells overexpressing EutJS and EutJSMNLK had multiple BMC-like structures (visible in 35% and 85%, respectively, of sectioned cells) as well as other larger, rounded structures (>300 nm in diameter) and unordered electron dense regions towards the poles of cells (note the whorl-like patterns in the case of EutJSMNLK) (Fig. 4, Supplementary Figures 8 and 9), observations that contradict the fluorescence data (Figs 1 and 2, Supplementary Figures 3 and 5). The nature and role of these higher order structures is unknown. In the case of EutP, large protein sheets were no longer present upon overexpression of EutPS, and some BMC-like structures were observed (32% of sectioned cells). However, no obvious BMC-like structures were visible in cells overexpressing EutPSMNLK. Instead, large protein sheets (rectangular/angular electron dense zones that run parallel or offset to the axis of the cell) again formed (94% of sectioned cells) (Fig. 4, Supplementary Figures 8 and 9), which may explain the lack of nanocompartment formation with this construct (Figs 1 and 2, Supplementary Figures 3 and 5). Whether EutP requires the presence of EutS alone, or the other shell proteins EutMNLK to restrain uncontrolled self-assembly is not known; correct EutS nanocompartment assembly is disrupted in the presence of EutP (Fig. 1, Supplementary Figure 3). In the case of EutQ, cells overexpressing EutQS contained both BMC-like structures (35% of sectioned cells) as well as large protein tubules or filaments (long protein structures that have defined electron dense edges surrounding an electron diffuse core) (25% of sectioned cells), in some cases in the same cell. However, this phenotype was eliminated upon overexpression of EutQSMNLK. Instead, cells contained what appeared to be multiple BMC-like structures; however, these were less well defined and smaller than typical EutSMNLK BMCs (150 nm), with a diameter more similar to that of EutS BMCs (100 nm) (Fig. 4, Supplementary Figures 8 and 9). This supports our hypothesis that EutQ is involved in the formation of multiple EutSMNLK nanocompartments (Fig. 2, Supplementary Figure 5).

Isolated nanocompartments encapsulate EutC 1-19 -EGFP.
Given the propensity of EutQ to form protein tubules or filaments in the presence of EutS (Fig. 4, Supplementary Figure 8), as well as causing aggregates of the EGFP cargo protein (Supplementary Figure 4), we sought to confirm that the multiple fluorescent foci and BMC-like structure phenotype that we had observed in cells overexpressing EutQSMNLK (Figs 2 and 4, Supplementary Figures 5 and 9) were in fact related to the formation of multiple nanocompartments. Towards this aim we isolated native S. enterica LT2 BMCs and recombinant E. coli nanocompartments from cell lysates for comparative measures, and confirmed both the presence of shell proteins by mass spectrometry, as well as the structural integrity of purified compartments by negative stain TEM.
Previously, we had partially purified native BMCs from S. enterica LT2, and recombinant nanocompartments from E. coli, using an adapted sucrose gradient method 39,77 . We noted that BMCs appeared deflated when purified using this method 39 , which is time consuming and labor intensive (taking 2-3 days). A straightforward and rapid (taking 2-3 hours) centrifugation method for the enrichment of Pdu BMCs has recently been established by Bobik's group 78 . We therefore chose to use this improved method for the enrichment of Eut BMCs and nanocompartments.
Initial attempts at establishing and adapting the centrifugation based method 78 to isolate native Pdu and Eut BMCs from S. enterica LT2 grown on 1,2 propanediol or ethanolamine to induce BMC formation proved successful in our hands. SDS-PAGE analysis and subsequent trypsin digest and mass spectrometry ( Supplementary  Figures 10 and 11) of isolated BMCs confirmed the presence of metabolic enzymes and/or shell proteins associated with both types of BMC. However, it should be noted that native Eut BMCs pelleted at a lower salt concentration (250 mM KCL, as a clear glassy pellet) than that of native and heterologously expressed Pdu BMCs (500 mM KCL, as a pink glassy pellet) 41,78 , indicating a difference in the composition/physiochemical properties of the two different types of BMCs. The Pdu BMCs are composed of a different number of shell proteins (seven instead of the five Eut shell proteins), which may cause them to behave differently in vitro. Additionally, the purity of the Pdu BMCs was far superior to that of the Eut BMCs (Supplementary Figure 10), likely due to the different solubility qualities of the two protein assemblies. Furthermore, Eut BMCs isolated using this method were morphologically different to Pdu BMCs (Supplementary Figure 12). The Pdu BMCs that we isolated appear almost identical to those purified by others 78 ; consistently sharp-edged polyhedral bodies with a diameter of 100 nm were observed on multiple occasions by negative stain TEM (Supplementary Figure 12A). On the other hand, Eut BMCs were fewer in number (likely due to limitations in measuring and normalizing Pdu and Eut BMC protein concentrations owing to the presence of copelleting proteins in the Eut BMC sample), appeared more round-edged, variable in size (100-250 nm in diameter), and also had a deflated appearance (Supplementary Figure 12B), again highlighting the fact that native Pdu and Eut BMCs are apparently not identical.
Using the same method we isolated recombinant EutSMNLK nanocompartments coexpressed with EutC 1-19 -EGFP and with or without EutQ, from E. coli C2566. Isolated heterologously expressed nanocompartments were similar in morphology to natively expressed Eut BMCs (Fig. 5A-C), with the same rounded-edge and deflated appearance, although they were slightly smaller in size than native Eut BMCs (100-150 nm in diameter). Notably, when EutQ alone was coexpressed with EutC 1-19 -EGFP, which appears to cause aggregation of the cargo protein (Supplementary Figure 4), no BMC-like structures, filaments or large aggregates were observed by negative stain TEM (Fig. 5D), further supporting the conclusion that the polyhedral structures observed in the case of the EutSMNLK and EutQSMNLK samples (Fig. 5B+ C) were nanocompartments and not aberrant aggregates or membrane fractions. The identity of the EutSMNLK shell proteins as well as EutQ in isolated nanocompartment samples was also confirmed by SDS-PAGE analysis and mass spectrometry (Fig. 5B-D).
To confirm that the polyhedral structures observed by negative stain TEM (Fig. 5B+ C) were isolated nanocompartments encapsulating EutC 1-19 -EGFP cargo protein, we conducted anti-GFP western blot analyses. Isolated EutSMNLK and EutQSMNLK nanocompartments harboring EutC 1-19 -EGFP were broken by sonication, and intact and broken shells were separated by native PAGE electrophoresis followed by exposure to anti-GFP antibody, as described previously 39 . GFP was detected when antibody was incubated with broken EutSMNLK and EutQSMNLK nanocompartments (Fig. 6A), but not when incubated with intact nanocompartments, confirming that the intact polyhedral structures (Fig. 5B+ C) sequestered EutC 1-19 -EGFP within the lumen of the shell, similar to what we have previously observed 39 . GFP appeared to migrate as two separate bands on a native-PAGE gel (Fig. 6A), indicating that the cargo protein may behave as a multimeric species. When the EutSMNLK and EutQSMNLK nanocompartments encapsulating EutC 1-19 -EGFP were separated under denaturing conditions, GFP was also detected as a multimeric species (Fig. 6B), suggesting that the signal sequence-cargo protein is capable of forming stable complexes. Note, this behavior appears to relate to the presence of the EutC 1-19 signal sequence and/or shell proteins; using the same anti-GFP western blot conditions with a EGFP standard results in the detection of a single band on native PAGE gels.
These in vitro data -showing that intact shells sequestering EutC 1-19 -EGFP can be isolated -unequivocally confirm that our interpretation that the fluorescent puncta we observed in cells expressing EutSMNLK and EutQSMNLK (Fig. 2, Supplementary Figure 5) is correct, they are indeed intact nanocompartments containing encapsulated EutC 1-19 -EGFP.

Interactions between EutQ and recombinant nanocompartment shell proteins. Our results
indicated that EutSMNLK nanocompartments could form correctly in the presence of EutQ (Figs 2 and 4, Supplementary Figures 5 and 9) whereas EutS nanocompartment formation was partially disrupted in the presence of EutQ (Figs 1 and 4, Supplementary Figures 3 and 8). It appeared that the presence of one or more of the shell proteins EutMNLK was required for the multiple nanocompartment phenotype. This may indicate that a Scientific RepoRts | 6:24359 | DOI: 10.1038/srep24359  shell protein/EutQ interaction is required to mediate correct nanocompartment formation, opening up the possibility that EutQ serves as an anchor point for nanocompartment biogenesis. To determine whether EutQ did interact with any of the shell proteins, we carried out in vitro protein pull downs.
Using N-and C-terminally His 6 tagged EutQ bound to a Co 2+ -affinity resin as bait, we passed E. coli C2566 cell lysate containing pUCBB-overexpressed untagged individual shell proteins EutS, EutM, EutN, EutL or EutK (Supplementary Table 1) over the resin as prey, similar to methods used for the Pdu shell proteins 62 . None of the shell proteins interacted with N-terminally His 6 tagged EutQ (Fig. 7A); however, a sufficiently tight interaction between EutM and C-terminally His 6 tagged EutQ in vitro resulted in co-elution of the two proteins from the resin (Fig. 7B). This may indicate that binding of the N-terminus of EutQ to the resin rendered the protein in a configuration that was not amenable to protein:protein interactions with EutM, or perhaps an interaction between the two proteins occurs close to, or at the N-terminus of EutQ. We tested the hypothesis that the N-terminus of EutQ harbored a putative interaction site for EutM by creating an N-terminally truncated version of EutQ lacking the first 100 amino acids (Supplementary Table 1), in a similar fashion to the construct design of EutQ 100-229 from S. enterica LT2, the crystal structure of which has been deposited in the Protein DataBank (PDB ID: 2PYT). Truncating the N-terminus of EutQ appeared to abolish an interaction with EutM in vitro, with ∆ EutQ1-100 remaining in the flow-through during protein pulldown experiments (Fig. 7C, Supplementary Figure 13). These findings indicate that the N-terminus of EutQ may be involved in protein interactions with EutM. It is not known what residues of EutQ mediate an interaction with EutM; no obvious surface charge distributions or hydrophobic patches that could act as potential protein interaction sites have been identified in the structurally characterized EutQ homolog CD1925 from C. difficile 630 25 which lacks 27 amino acids at the N-terminus.
We therefore turned our attention to identifying potential interaction site(s) for EutQ on the surface of EutM. Considering that EutQ appears to mediate multiple nanocompartment biogenesis (rather than be encapsulated within the lumen of the nanocompartment), we assumed that EutQ would interact with the cytosol-facing surface of EutM. A short helix has been identified on the BMC lumen-facing side of the EutM homolog PduA, which is the interaction site for the N-terminal 18 amino acid portion of the propionaldehyde dehydrogenase PduP, resulting in its encapsulation within Pdu BMCs [40][41][42] . We therefore chose to focus our search for potential interacting residues on the opposite face of EutM to that of the PduA encapsulation helix (which would correspond to α -helix 3, residues G80-F86 on EutM), using the crystal structure of EutM from E. coli (PDB ID: 3MPW 28 ) as a guide. The putative outer face of EutM is characterized by the presence of exposed α -helix 2 (V49-Q64) on each of the EutM monomers (Fig. 8); we theorized that this would be the most suitable site for any potential protein:protein interaction. Considering that hydrophobic amino acids play a major role in interactions between PduA and PduP 40-42 , we mutated all hydrophobic residues along exposed α -helix 2 (V49-Q64) to polar serine (Fig. 8). EutQ co-elution with EutM was not prohibited, implying that EutQ:EutM interactions are not mediated by hydrophobic residues. Next, we explored whether electrostatic interactions were involved, by mutating all charged residues to either remove the charge or to swap the charge of the side chain. Mutants K53D, D57A, D57K and Q64K on EutM abolished EutQ:EutM interactions in vitro, showing that a specific electrostatic surface on the likely cytosol-exposed helix of EutM is required for EutQ binding in vitro (Fig. 8). Interestingly, residues K53 and D57 are well conserved across 138 EutM homologs (Supplementary Figure 14); however, not all eut gene clusters across different types of bacteria contain a eutQ gene 45,51 . It remains to be explored whether the interaction between EutM and EutQ (and presumably EutQ-mediated biogenesis of multiple Eut BMCs) is specific to the Eut BMC proteins from S. enterica LT2, or whether this is a widely conserved mechanism adopted by other bacterial species. Nonetheless, we suggest that in S. enterica LT2 Eut BMC and recombinant nanocompartment biogenesis, EutQ may serve as a nucleation point for EutM, potentially organizing BMCs within the cell similar to ParA organization of the carboxysome 37,79 , although this hypothesis remains to be verified. If this were the case, it could be envisioned that the EutQ-nucleated EutM serves as a scaffolding site for the initiation of BMC biogenesis, similar to the biogenesis of carboxysomes which requires the scaffolding protein CcmM and shell-protein recruitment protein CcmN 36,38 .
Finally, in an attempt to determine whether the EutQ:EutM interactions that we had observed in vitro were physiologically relevant, providing some support for the theories presented above, we sought to create the K53D, D57K and Q64K EutM mutants in the context of EutQSMNLK and visualize the formation (or loss) of multiple nanocompartments in E. coli C2566. Unfortunately, after exhaustive attempts (including the use of a tightly controlled and inducible promoter 80 for shell protein expression) to obtain reproducible and stable phenotypes for cells expressing the EutM mutants in the context of the other four shell proteins and EutQ, we were forced to conclude that expression of these mutants in this genetic context causes severe stress for E. coli cells. In fact, the recombinant E. coli cells appeared small with large polar GFP aggregates and exhibited attenuated growth. Furthermore, sequencing of plasmids isolated from these cells showed that the mutated EutM genes were not stable, frequently accumulating additional, presumably inactivating mutations. At this point, we do not know why mutations in this surface helix of EutM cause these severe phenotypes. While these findings negate the possibility of properly testing EutQ:EutM interactions in vivo, they do provide us with valuable insights into the importance of levels of protein expression on heterologous nanocompartment biogenesis. Fine-tuning of individual shell protein, EutQ and cargo protein expression in E. coli remains to be engineered to further optimize the formation of nanocompartments for biotechnological applications.

Discussion
The goal of this work was to identify which protein(s) encoded by the S. enterica LT2 eut operon could be involved in multiple Eut BMC biogenesis, and to express this as part of our recombinant system for the production of multiple nanocompartments in the heterologous host E. coli C2566. We show that overexpression of EutQ, a putative cupin domain protein of unknown function, with the shell proteins EutSMNLK can increase the number of recombinant nanocompartments produced per cell. Additionally, the nanocompartments appear to gain slight mobility upon overexpression of EutQ. While the role of EutQ in facilitating this multiple unconstrained nanocompartment phenotype is not fully understood, we hypothesize that EutQ could serve as a nucleation point for BMC biogenesis, mediated by interactions between EutQ and the shell protein EutM.
In Synechococcus, the cytoskeletal protein ParA self-polymerizes as filaments, and these higher order structures are required for organization of carboxysomes within the cell 79 . Biogenesis of multiple carboxysomes is initiated by aggregation of the scaffolding protein CcmM and the cargo RuBisCo along the cytoskeletal scaffold 36,37 followed by CcmN-mediated recruitment of shell proteins 38 . Furthermore, the Ras-like GTPase PduV has been suggested to play a role in the spatial distribution of recombinant Pdu BMCs 62 . Members of the GTPase superfamily are involved in vesicle trafficking along the cytoskeleton 76 and therefore it would seem logical that the PduV homolog EutP could play a similar role in Eut BMC mobility. Yet, this is not the case. Unexpectedly, overexpression of EutP disrupts nanocompartment assembly in a heterologous host.
Instead, we have shown that EutQ is involved in multiple native Eut BMC, and recombinant nanocompartment, assembly in S. enterica LT2 and E. coli C2566, respectively. How the cupin domain protein EutQ mediates this effect is not fully understood; the cupin superfamily is characterized by proteins sharing a conserved β -barrel fold, with widely diverse activities such as dioxygenase, decarboxylase, isomerase or small molecule binding 74 , but are not typically associated with cytoskeletal formation. Nonetheless, EutQ deletion from S. enterica LT2 results in an increased aldehyde release when cells are grown on ethanolamine 71 , and it has been suggested that EutQ is required for correct compartmentalization of EutE in Eut BMCs 73 . These findings are supported by our conclusion that EutQ is required for (multiple) Eut BMC assembly. Yet, EutQ does not appear to be absolutely essential for nanocompartment biogenesis in a heterologous host. The shell proteins EutS and EutSMNLK can self-assemble as a single nanocompartment when overexpressed in E. coli 39 , similar to the self-assembly of multiple compartments during overexpression of Pdu BMC 62 and carboxysome 66,81 shell proteins. However, it should be noted that neither the Pdu BMC nor the carboxysome operons appear to have a EutQ homolog. Therefore, EutQ-mediated formation of multiple BMCs may be limited to bacteria that have a EutQ-like protein; and it remains to be investigated if this is a S. enterica LT2 Eut BMC biogenesis specific phenomenon.
The interaction between EutQ and the shell protein EutM observed in vitro raises the possibility that EutQ regulates Eut BMC assembly by acting as (a) nucleation point(s) for initiating compartment growth, similar to the carboxysome biogenesis mechanism [36][37][38] . The EutM homologs PduA and CD1918 self-assemble as protein filaments when overexpressed in E. coli 23,25 and EutM also self-assembles as protein filaments, while EutS can form nanocompartments by itself 39 . Perhaps multiple EutQ:EutM aggregates can serve as shell assembly hubs, resulting in the growth of multiple nanocompartments; as opposed to a single nanocompartment whose growth could result from the propensity of EutS or EutM to self-assemble 39 .
Our long term goal is to optimize intracellular formation of multiple nanocompartments for targeted encapsulation of enzymes, with the aim of increasing multi-step pathway reaction efficiency in E. coli as a widely accepted recombinant host for applications biotechnology and biocatalysis. Here we have shown that it is possible to increase the number of nanocompartments produced per E. coli cell. Whether or not this improves overall reaction efficiency of metabolic pathways in E. coli remains to be explored. Nevertheless, formation of multiple compartments with encapsulated enzyme catalysts in E. coli should be beneficial for the production of compartments with encapsulated enzyme cargo as nanobioreactors for in vitro multi-enzyme biocatalytic applications.
Ultimately, targeting of multi-step pathways to (multiple) nanocompartments will likely require more than the one signal sequence that we have used to date (EutC [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18][19] 39 . Discovery and engineering of new signal sequences for targeting of multiple enzymes to our recombinant Eut nanocompartments is beyond the scope of this study, and will be addressed in future work e.g. by adopting strategies reported with Pdu BMCs 40,42,43,82 . Similarly, engineering of controlled, coordinated expression of both shell, accessory and cargo-proteins in non-native hosts such as E. coli will be required for the design of BMC nanocompartments as versatile industrially relevant platforms. Nonetheless, the work that we have presented here lays the foundation for an optimized configuration of nanocompartment assembly, which will eventually lead to the compartmentalization of pathways and biocatalytic multi-enzyme reaction cascades for the production of biotechnologically relevant compounds of choice.

Expression of nanocompartments in E. coli C2566. Recombinant nanocompartments were produced
in E. coli C2566 cells by the overexpression of EutS or EutSMNLK using our in-house high-copy BioBrick TM vector pUCBB, which contains a modified lac promoter allowing constitutive expression 67 . Localization of EGFP to recombinant nanocompartments in E. coli C2566 was achieved by coexpressing shell proteins with pACBB-EutC 1-19 -EGFP 39 . For the analysis of native Eut BMC formation in S. enterica LT2 wild type and ∆ EutQ, cells were transformed with pBBRBB-EutC 1-19 -EGFP, and BMC production was induced by growth on ethanolamine as described above.
Light microscopy and time-lapse imaging. Static images of S. enterica LT2 and E. coli C2566 cells were acquired using a Nikon Eclipse E800 or Nikon Eclipse 90i microscope equipped with bright field, DIC, phase and fluorescence optics including a 120 W X-Cite epi-fluorescence illuminator with blue (excitation filter 470-490 nm, barrier 520-580 nm) filter set and a 100X, 1.4 n.a. plan apo, oil immersion objective. For time-lapse imaging, 1 μL cell sample was immobilized with 10 μL of BacLight mounting oil (Life Technologies TM ). Time-lapsed images were collected using a Nikon TiE equipped with a Lumencor SpectraX light source at 470 nm using a Plan Apo 100x, 1.45 n.a. objective with a Andor Zyla camera controlled via the Nikon Elements 4.6 software. Images were processed using 4-5 iterations of the 2D deconvolution Landweber method within Nikon Elements (module from AutoQuant ® ) to remove out-of-field fluorescence. Post-capture image analyses and cropping was conducted in Nikon NIS Elements Viewer 4.6 and GIMP 2.
Transmission electron microscopy. S. enterica LT2 (grown on supplemented E medium) and E. coli C2566 cells (grown on LB medium) were fixed, embedded, sectioned and stained prior to TEM analysis as described previously 39 . Samples were visualized and photographed using a Philips CM12 transmission electron microscope. Post-capture alignment and cropping was conducted in GIMP 2.

Isolation of native BMCs and recombinant nanocompartments. Enrichment of BMCs from cell
pellets was conducted using methods adapted from previously published protocols 78 . Overnight cultures were harvested by centrifugation (3000 g, 30 mins) and the cell pellet was washed in Buffer A (50 mM Tris pH 8.0, 500 mM KCl, 12.5 mM MgCl 2 , 1.5% 1,2 propanediol or ethanolamine-HCl). Pellets (1 g wet weight) were resuspended in 10 mL Buffer A and cells were lysed by addition of B-PER ® II Bacterial Protein Extraction Reagent (60% v/v) (ThermoScientific), supplemented with lysozyme (0.25 mg mL −1 ), DTT (1 mM), PMSF (0.1 mM) and DnaseI (2 U). After 30 minutes incubation at room temperature, cell debris and supernatant were separated twice by centrifugation (9700 g, 5 mins). BMCs were pelleted from the resulting supernatant by centrifugation (43000 g, 30 mins). The small, glassy pellet was resuspended in 500-1000 μL Buffer B (50 mM Tris pH 8.0, 50 mM KCl, 5 mM MgCl 2 , 1% 1,2 propanediol or ethanolamine-HCl). Purity of BMCs was assessed by separation on a SDS-PAGE gel followed by silver stain or Coomassie stain to detect proteins, and overall protein content of the sample was determined using Pierce BCA Protein Assay Kit (Life Technologies). For TEM analysis, isolated BMC samples were diluted to 0.1 mg mL −1 in Buffer B. BMCs were negatively stained by applying 10 μL protein to a formvar-coated copper 200 mesh grid, and fixing in an equal volume of Trump's fixative. Excess liquid was wicked off using filter paper, and excess salts were removed using 10 μL water. Proteins were stained in 10 μL uranyl acetate (2%) and grids were air dried prior to analysis using a Philips CM12 transmission electron microscope.

Anti-GFP western blot analyses of encapsulated EutC 1-19 -EGFP. Isolated EutSMNLK and
EutQSMNLK nanocompartments harboring EutC 1-19 -EGFP were broken by sonication. Broken and intact nanocompartments were loaded in separate lanes of a 10% native polyacrylamide gel. Following electrophoresis under non-denaturing conditions, proteins were transferred to a PVDF membrane (Roche), and GFP was detected using a HRP conjugated monoclonal primary anti-GFP antibody (Thermo Scientific). Development was conducted using the Pierce ® Fast Western Blot Kit with ECL substrate (Thermo Scientific). The membrane was then placed in a film cassette and was exposed to film (Bioexpress) for 2 minutes, and the film was subsequently developed.
Trypsin digestion and peptide sequencing. Protein bands were excised from a SDS-PAGE gel and silver stain was removed from the gel slices in a 1:1 solution of potassium ferricyanide (30 mM) and sodium thiosulphate (100 mM). Proteolytic digestion with trypsin was conducted in 50 mM NH 4 HCO 3 , 5 mM CaCl 2 , 12.5 ng μL −1 trypsin. Peptides were extracted in acetonitrile and formic acid and were desalted prior to analysis on a LTQ mass spectrometer (ThermoScientific). Mass spectra were analyzed using PEAKS ® 7 or Scaffold version 4.2.1, with expected thresholds of 90.0% probability and at least 2 identified peptides.