Introduction

Bacteria employ different types of contractile injection systems (CIS) to mediate cell-cell interactions or to mediate cell death (1). CIS are evolutionarily and structurally related to contractile tails of bacteriophages and are comprised of core modules including a baseplate, a contractile sheath, and an inner tube (2,3). CIS firing is triggered by a conformational change within the baseplate (4,5) and results in the contraction of the sheath, which in turn propels the inner tube along with the spike tip. Tube expulsion facilitates the release of associated effector proteins into target cells or the extracellular space (6,7).

Bioinformatic analyses have shown that CISs are conserved across diverse microbial phyla, including Gram-negative and Gram-positive bacteria, as well as archaea (8,9). Based on their distinct modes of action, CIS can be categorized into two main groups: intracellular type VI secretion systems (T6SS) and extracellular CIS (eCIS). T6SS are abundant among Gram-negative bacteria and in some archaea, and they are anchored to the cytoplasmic membrane during assembly and firing (10–12). A crucial component of the T6SS is the baseplate, which acts as a nucleus for T6SS assembly and mediates the binding of the T6SS particles to the cytoplasmic membrane until contraction. Upon contraction, the spike and inner tube are propelled out of the cell into an adjacent cell or the medium. eCIS, on the other hand, are assembled as free-floating particles in the cytoplasm and are subsequently released into the extracellular space upon lysis of the producer cell (13–16). Following the release, eCIS attach to the surface of target cells via their tail fibers, followed by contraction and puncturing of the target cell envelope.

Besides T6SS and eCIS, there is accumulating evidence of additional CIS assemblies with distinct modes of action in bacteria. For example, in multicellular cyanobacteria, CIS are anchored to the intracellular thylakoid membrane stacks via an extension of the baseplate components and have been proposed to function in cell lysis and the formation of “ghost cells” in response to stress conditions (17). Furthermore, several recent studies reported the production of free-floating cytoplasmic CIS particles in the multicellular Gram-positive Streptomyces bacteria, which have been shown to modulate cellular development through presumably slightly different mechanisms (18–22).

Streptomyces are best known for producing a plethora of medically and industrially important secondary metabolites, including molecules with antimicrobial, antifungal, anticancer or immunosuppressive properties. The production of these molecules is tightly coordinated with the Streptomyces developmental life cycle that encompasses two filamentous cell types: vegetative hyphae that grow by tip extension and branching to scavenge for nutrients, and reproductive (aerial) hyphae that eventually differentiate into chains of spores, which can then be dispersed to restart the life cycle (23,24). Using Streptomyces coelicolor as a model organism, we and others previously demonstrated that S. coelicolor CIS (CIS^Sc) mediate a form of regulated cell death, which influences the onset of sporulation and secondary metabolite production in response to exogenous stress and unknown cellular signals (Figure 1a) (18,19).

Figure 1:

It remains unclear whether CIS^Sc contraction occurs while the CIS^Sc particles are free-floating in the cytoplasm or whether this depends on an interaction with the membrane. Moreover, CIS encoded in Streptomyces genomes lack clear homologs for tail fiber components or a T6SS trans-envelope complex, raising the question of how CIS^Sc particles could potentially interact with the cytoplasmic membrane. We and others recently noted an uncharacterized protein (18,19), SCO4242, which is conserved in the majority of Streptomyces species and other Actinomycete species that carry a type IId eCIS locus (8,9) (Supplementary Fig. 1 and Supplementary Data 1 and 2). Interestingly, SCO4242 is encoded just downstream of the baseplate components in the CIS^Sc gene cluster, a location that usually encodes tail fibers from conventional eCIS (13,16,25). We will, therefore, refer to the gene product of SCO4242 as CisA for “CIS^Sc-associated protein A.” Here, we set out to assess the importance of CisA for CIS^Sc function and its possible role as a membrane anchor.

Results

CisA is required for CIS^Sc contraction in situ

To explore the role of CisA (SCO4242) in CIS^Sc contraction and function, we first generated an S. coelicolor ΔcisA null mutant. Negative stain electron microscopy (EM) imaging of CIS^Sc particles that were purified from ΔcisA mutant cells showed assemblies in the contracted conformation, similar to the CIS particles purified from the wild type (WT) (Fig. 1b/c). We then imaged vegetative hyphae of the WT, the ΔcisA mutant and the complemented mutant (ΔcisA/cisA⁺) by cryo-electron tomography (cryoET) (270 tomograms in total, n=3 experiments). The observed CIS^Sc particles in intact hyphae of these three imaged strains did not reveal discernable differences, indicating that CisA is not required for CIS^Sc assembly. We previously showed that CIS^Sc contraction is linked to cell death in S. coelicolor wildtype hyphae and we observed a clear shift in the number of extended to fully contracted CIS^Sc particles in damaged and dead hyphae (ghost cells) (18). Interestingly, and in contrast to the WT and complemented cisA mutant strain (Fig. 1d and f), cisA-deficient hyphae only contained CIS^Sc particles in the extended conformation irrespective of the cellular integrity of the hyphae (Fig. 1e, Supplementary Fig. 2a-d). While hyphal cell death can be induced by a range of different external or internal factors, our results suggest that CisA is required for the contraction of CIS^Sc in situ.

CryoEM structure of the extended CIS^Sc assembly

Contractile injection systems are generally classified into intracellular type VI secretions systems (T6SS) and the superfamily of extracellular CIS (eCIS), which is subdivided into six families with distinct genetic compositions (8). Many Streptomyces genomes contain a class IId eCIS locus, a subtype that is structurally less well understood compared to other eCIS subtypes, which includes ‘antifeeding prophages’ (AFPs) from Serratia, ‘metamorphosis-associated contractile structures’ (MACs) from Pseudoalteromonas luteoviolacea, AlgoCIS from A. machipongonensis or ‘Photorhabdus virulence cassettes’ (PVCs) from P. asymbiotica (1).

The CIS^Sc gene cluster consists of 19 predicted open reading frames (accessions SCO4242-4260) (Fig. 2a), including genes with sequence similarities to potential CIS structural components (cis1-cis16), as well as additional genes with unknown functions (including cisA) (18,19). To determine whether CisA is an integral component of CIS^Sc, we set out to elucidate the high-resolution structure of purified CIS^Sc particles from a non-contractile CIS^Sc mutant (18). The purified assemblies were subjected to single particle cryo-electron microscopy (cryoEM) for structure determination (Supplementary Fig. 3), yielding the first high-resolution model of the CIS^Sc in its extended conformation, which also presents the first example of a CIS IId subtype (8).

Figure 2:

Our model shows that extended CIS^Sc particles share the conserved eCIS architecture, including three modules: cap, sheath-tube and baseplate (Fig. 2b). We processed the dataset using an established workflow that was previously employed to produce the atomic model for other CIS (Supplementary Fig. 3) (13,17). The resulting maps of the baseplate and the cap reached resolutions of 3.5 and 3.4 Å, respectively. The quality of the final maps from the two modules enabled de novo structural modelling (Supplementary Fig. 4, Supplementary Table 1).

The final model of CIS^Sc comprised 783 polypeptide chains that assemble into a bullet-shaped particle measuring 268 nm in length (Fig. 2b/c). All analyzed particles were identical in length, consistent with our in situ data of extended CIS^Sc particles in vegetative hyphae of WT S. coelicolor (Fig. 1d-f). The length of eCIS particles is often controlled by tape measure proteins (26), we note, however, that the corresponding gene is absent from the CIS^Sc gene cluster. The distal end of the CIS^Sc is capped by a tail terminator complex (Fig. 2c/e). This cap comprises one protein, Cis16, forming a C6-symmetrical complex that terminates the inner tube and sheath, respectively. This minimal architecture resembles the cap complexes of PVC or Pyocin R2 CISs, but it is different from the more complex AlgoCIS from A. machipongonensis and AFP from S. entomophila (Supplementary Fig. 5) (13,16,25,27). The structures of the CIS^Sc sheath-tube module have been discussed previously (18). Here we found that the module consists of 61 sheath layers (Cis2-L₁–L₆₁) and 60 tube layers (Cis1-L₁-L₆₀) (Fig. 2b/d).

The baseplate module comprises a heteromeric assembly of the proteins Cis1b/7/8/9/11/12 (Fig. 2b/c). The symmetry transition from the inner tube (Cis1a, C6) to the VgrG-like spike (Cis8, C3) is adapted by two layers of the tube initiators Cis1b and Cis7. The first layer of the sheath (Cis2) is bound to the conserved gp25-like protein Cis9, which connects to the baseplate ‘wedges’ composed of Cis11 and Cis12 (Fig. 2b/c). Importantly, the CIS^Sc ‘wedges’ consist of two copies of Cis11 and one copy of Cis12 (Fig. 2f), resembling the baseplate of Pyocin R2(27). This is in contrast to other previously reported eCIS, whose baseplate wedge comprises only one Cis11/12 component (AlgoCIS, AFP, PVC) and a Cis11 extension forming a cage around the spike (AlgoCIS) (Supplementary Fig. 6) (13,16,25).

Out of the 19 predicted open reading frames in the CIS^Sc gene cluster (Fig. 2a), nine encoded proteins were present in the final reconstruction, all with atomic models built (Fig. 2b/c). However, we did not detect a density for CisA in our model, suggesting that CisA is not part of purified CIS^Sc particles but may interact via a different mechanism. This is also in agreement with mass spectrometry results that indicate the absence of CisA peptides from purified CIS^Sc particles (Fig. 2a) (18). We speculate that the purification protocol used to isolate CIS^Sc particles for single-particle analysis may not preserve a CIS^Sc-CisA interaction, or the interaction is too transient and only occurs under certain growth conditions. To better understand how CisA may mediate CIS^Sc function, we set out to further characterize CisA in vivo.

CisA is a bitopic protein

Previous bioinformatic analyses (18,19) and Alphafold2 (28) modeling of CisA suggest that CisA consists of a largely unstructured N-terminal region, a transmembrane segment, and a conserved immunoglobulin-like (IgG) domain at the C-terminus (Fig. 3a). To characterize the intracellular localization of CisA, we first generated a strain in which cisA was fused to the N-terminus of mCherry and expressed in trans from a constitutive promoter in the S. coelicolor ΔcisA mutant. Using fluorescence light microscopy, we consistently observed a CisA-mCherry fluorescence signal along the hyphal periphery, indicative of a membrane-associated protein (Fig. 3b). In contrast, no fluorescence was observed in WT hyphae carrying an empty plasmid (Fig. 3c). Next, we tested if CisA is indeed localized to the membrane in vivo by performing cellular fractionation experiments. We separated soluble proteins from membrane proteins using whole cell lysates from the S. coelicolor WT or the ΔcisA mutant that was complemented with a cisA-3xFLAG fusion expressed in trans from a constitutive promoter. Our analysis confirmed that CisA-3xFLAG co-sedimented with the cell membranes, while the cytoplasmic transcription factor WhiA could only be detected in the soluble protein fraction (Fig. 3d).

Figure 3:

Next, we investigated the topology of CisA in the membrane using a dual phoA-lacZα reporter system in E. coli. The reporter activity can be directly visualized on indicator plates based on the complementary activities of the cytoplasmic reporter β-galactosidase LacZ (magenta coloration) and the periplasmic reporter alkaline phosphatase PhoA (blue coloration) (29). We engineered E. coli strains that expressed a C-terminal fusion of CisA variants to the PhoA-LacZα reporter, including full-length CisA and CisA variants that carried either a deletion in the predicted transmembrane domain and/or the putative periplasmic domain (Fig. 3e). In addition, we included three controls, including 1) cells carrying the empty reporter plasmid, resulting only in cytoplasmic β-galactosidase activity, 2) cells expressing a PhoA/LacZ fusion to the Streptomyces cytoplasmic cell division protein SepH, and 3) cells expressing a PhoA/LacZ fusion to the periplasmic domain of the Streptomyces membrane protein RsbN (30,31). The results of this assay confirmed that CisA is a bitopic protein with an N-terminus (amino acids 1-285) located in the cytoplasm and a C-terminus (amino acids 310-468) present in the periplasm of E. coli (Fig. 3f).

Collectively, our results demonstrate that CisA is a membrane protein comprised of a flexible, cytoplasmic N-terminal domain and a C-terminal globular IgG-like domain that is exposed to the outside of the cytoplasmic membrane.

CisA is essential for the cellular function of CIS^Sc

We previously established a fluorescence-based assay to demonstrate that the production of functional CIS^Sc particles results in a significantly reduced viability of Streptomyces hyphae exposed to exogenous stress, including treatment with the membrane-pore forming antibiotic nisin, UV radiation or the protonophore carbonyl cyanide 3-chlorophenylhydrazone (CCCP) (18). This viability assay is based on the measurement of the ratio of two fluorescent markers, namely cytoplasmic sfGFP produced by viable intact hyphae and the membrane dye FM5-95 to detect intact as well as partially lysed hyphae. Using this assay, we tested whether CisA was essential for CIS^Sc-mediated cell death. We therefore determined the relative viability of the WT cisA, the ΔcisA and the complemented cisA (ΔcisA/cisA⁺) strains expressing cytosolic sfGFP. We compared cells exposed to a sublethal concentration of nisin (1 µg/ml for 90 min) to cells not exposed to antibiotic stress. All three strains were found to express CIS^Sc assemblies under these growth conditions (Supplementary Fig. 7).

Without nisin stress, all three strains exhibited a similar sfGFP/FM5-95 ratio, indicating no significant difference in viability (Fig. 4a/c). As previously observed (18), the viability of the WT was dramatically reduced in response to nisin stress. In contrast, the viability of ΔcisA hyphae was unaffected by nisin stress and comparable to the viability of untreated cells. This phenotype could be completely reversed by complementing the ΔcisA deletion in trans (Fig. 4b/d). Notably, the ΔcisA strain phenocopies CIS^Sc knockout and non-contractile mutants we analyzed previously (18). Thus, we conclude that CisA is required for CIS^Sc-mediated cell death upon stress.

Figure 4:

To further test the dependence of CIS^Sc function on CisA, we conducted additional in vivo assays. Previous studies showed that the absence of functional CIS^Sc impacted the timely differentiation of Streptomyces hyphae into chains of spores, resulting in accelerated cellular development and reduced secondary metabolite production (18,19). Here we repeated these analyses using the WT, the ΔcisA mutant and the complemented mutant (ΔcisA/cisA⁺). All strains consistently completed their life cycle and synthesized spores after 96 h of growth on solid medium (Fig. 4e). Importantly, unlike the WT and the ΔcisA/cisA⁺ strain, ΔcisA mutant colonies sporulated markedly earlier (72 h). These findings were further corroborated by quantifying the number of spores produced by each strain under the same experimental conditions (Fig. 4f). Finally, we also tested the production of the two characteristic pigmented secondary metabolites of S. coelicolor, actinorhodin (blue) (32) and undecylprodigiosin (red) (33). Compared to the WT and the ΔcisA/cisA⁺ strain, the ΔcisA mutant showed significantly reduced levels of secondary metabolite production (Supplementary Fig. 8). Our data therefore support the conclusion that the cellular function of CIS^Sc in S. coelicolor depends on CisA, and its absence impacts the timely cellular and chemical differentiation of S. coelicolor hyphae.

Discussion

Here, we identify CisA as an essential membrane-associated factor that mediates the cellular function of CIS^Sc particles in Streptomyces coelicolor. In the absence of CisA, CIS^Sc are not able to contract in the cellular context (Fig. 1). Similar to a CIS^Sc deletion or non-contractile mutant (18), this decreases the induction of cell death upon stress, which in turn affects cellular development and secondary metabolite production (Fig. 5). The mechanistic aspects of this distinct mode of action of a CIS are discussed below.

Figure 5:

A conserved feature of all previously studied CIS is the requirement of the contractile apparatus to bind to a membrane before firing. This can be achieved in different ways, leading to the classification of CIS into different modes of action. First, T6SS are anchored to the cytoplasmic membrane of the host cell by a trans-envelope complex spanning from the cytoplasm to the inner membrane, periplasm, and outer membrane (34–36). Second, contractile phages and eCIS bind to the surface of their target cell via tail fibers, which are typically connected to the baseplate and transmit the signal for firing to the baseplate (13,16,37). Third, tCIS are anchored to thylakoid membrane stacks in cyanobacteria by an extension of their baseplate (38). Finally, and in contrast to the previously described CIS, we show that intracellular CIS^Sc particles from Streptomyces do not include structural components that can directly bind the host membrane. Our data suggest that CisA may be required to mediate an interaction of CIS^Sc with the cytoplasmic membrane prior to firing (Fig. 5). This interaction may occur directly via the binding of the CIS^Sc particle to the membrane protein CisA, or through the interaction with a yet unknown factor. To explore this idea, we performed an in silico prediction of protein-protein interactions between monomeric CisA and CIS^Sc components using Alphafold2-Multimer (39) (Supplementary Fig. 9a). Interestingly, this analysis identified the baseplate component Cis11 as a significant hit and possible interaction partner of CisA (Supplementary Fig. 9b-d). Importantly, such a protein-protein interaction would be consistent with our cryoEM structure (Fig. 2b-f), showing a peripheral surface-exposed position of Cis11 in the baseplate complex of extended CIS^Sc.

Additional structural modelling using AlphaFold2-Multimer with truncated versions of CisA in complex with Cis11 further support the idea that the largely unstructured cytosolic portion of CisA is required to interact with the Cis11 (Supplementary Fig. 10). We have been unable to confirm a direct interaction between CisA and Cis11 because the expression of cisA is toxic in E. coli, which prevented the experimental analysis of the CisA-Cis11 interaction using co-purification approaches and bacterial two-hybrid studies. However, we did detect CisA peptides in crude purifications of CIS^Sc from nisin-stressed cells (Supplementary Table 2). This would be consistent with a transient and/or short-lived interaction of CIS^Sc particles with CisA, triggered by an exogenous or cellular signal. In agreement with this idea, we rarely observed extended CIS^Sc assemblies associated with the cytoplasmic membrane via their baseplate in WT cells (Fig. 1d).

We further speculate that CisA may not only serve as a mediator for the interaction of CIS^Sc with the cytoplasmic membrane but also as a sensor and checkpoint for inducing cell death only under specific stress conditions or in response to cellular signals. The CisA C-terminal IgG-like domain that is localized outside the cytoplasm and in proximity of the Gram-positive cell wall may sense stress signals that could potentially trigger conformational changes in CisA, which in turn could mediate the recruitment of CIS^Sc assemblies to the cytoplasmic membrane, followed by firing. The fatal outcome of firing for the producing cell may have driven the evolution of a stepwise mechanism involving a checkpoint that prevents self-inflicted cell death by accidental firing. Such a mechanism would keep the CIS^Sc in a free-floating, non-functional and safe state in the cytoplasm.

Methods

Bacterial strains, plasmids, and oligonucleotides

Bacterial strains, plasmids, and oligonucleotides used in this study are listed in Supplementary Tables 3-4. Plasmids were generated using either standard restriction-ligation or assembled using the Gibson Assembly Master Mix (NEB) or In Vivo Assembly. All plasmids were verified by DNA sequencing. Escherichia coli strains were cultured in LB, SOB, or DNA medium. E. coli strains TOP10, DH5α and NEB5α were used to propagate plasmids and cosmids, E. coli strain BW25113/pIJ790 for recombineering cosmids and E. coli ET12567/pUZ8002 for interspecies conjugation. When required, media was supplemented with antibiotics at the following concentrations: 100 µg/ml carbenicillin, 50 µg/ml apramycin, 50 µg/ml kanamycin, 50 µg/ml hygromycin.

Streptomyces coelicolor strains were cultivated in LB, TSB, TSB-YEME, or R2YE liquid medium at 30 °C in baffled flasks or flasks with springs, at 250 rpm or grown on LB, SFM, R2YE medium solidified with 1.5 % (w/v) Difco agar (40). Antibiotics were added at the following concentrations: 25 µg/ml apramycin, 50 µg/ml kanamycin, 25 µg/ml hygromycin, 12.5 µg/ml nalidixic acid.

Construction of the S. coelicolor ΔcisA mutant

The λ RED homologous recombination system was used to isolate gene replacement mutations using PCR-directed mutagenesis (ReDirect) of the S. coelicolor cosmid StD8A containing the cis gene cluster(41),(42). The cisA coding sequence (SCO4242) was replaced with the aac3(IV)-oriT resistance cassette from pIJ773. The mutant cosmid (pSS684) was introduced into E. coli ET12567/pUZ8002, followed by conjugation into S. coelicolor M145. Exconjugants that had successfully undergone double-homologous recombination were identified by screening for apramycin resistance and kanamycin sensitivity. The deletion of cisA was subsequently verified by PCR.

Sheath preparation of CIS^Sc

S. coelicolor strains were grown in 30 ml TSB, TSB-YEME or R2YE liquid medium for 48 h. Cells were pelleted by centrifugation (7,000 x g, 10 min, 4 °C), resuspended in 5 ml lysis buffer (150 mM NaCl, 50 mM Tris-HCl, 0.5× CellLytic B (Sigma-Aldrich), 1 % Triton X 100, 200 µg/ml lysozyme, 50 μg/ml DNAse I, pH 7.4), and incubated for 1 h at 37 °C. Cell debris was removed by centrifugation (15,000 × g, 15 min, 4 °C) and cleared lysates were subjected to ultra-centrifugation (150,000 × g, 1 h, 4 °C). Pellets were resuspended in 150 µl resuspension buffer (150 mM NaCl, 50 mM Tris-HCl, supplemented with protease inhibitor cocktail [Roche], pH 7.4). CIS^Sc sheath preparations were analyzed by negative stain EM imaging (43) and mass spectrometry at the Functional Genomics Center Zürich.

To purify non-contractile CIS^Sc particles from S. coelicolor (SS393), cleared cell lysates were subjected to ultracentrifugation using a sucrose cushion (20 mM Tris pH 8.0, 150 mM NaCl, 50 mM EDTA, 1 % Triton X 100, 50 % [w/v] sucrose) (18). The sucrose cushion alongside 1 mm of liquid above and residual bacterial contaminants were removed by centrifugation at 15,000 x g. Samples were then subjected to a second round of ultracentrifugation without a sucrose cushion (150,000 × g, 1 h, 4 °C) and resulting cell pellets were resuspended in buffer (50 mM Tris, 150 mM NaCl). These crude samples were then purified further by gradient ultracentrifugation (10 %-50 % continuous gradient made with BIOCOMP gradient master IP model 107 gradient forming instrument) at 100,000 x g for 1 h using a SW55 Ti rotor. The gradient was analyzed in 11 fractions of 500 µl and checked by negative stain EM. The fractions, which contained CIS^Sc particles, were pooled and used for further experiments.

Negative stain electron microscopy

4 µl of purified CIS^Sc sheath particles were adsorbed to glow-discharged, carbon-coated copper grids (Electron Microscopy Sciences) for 60 s, washed twice with milli-Q water and stained with 2 % phosphotungstic acid for 45 s. The grids were imaged at room temperature using a Thermo Fisher Scientific Morgagni transmission electron microscope (TEM) operated at 80 kV.

Mass spectrometry analysis

To confirm the presence of predicted CIS^Sc components, isolated sheath particles were subjected to liquid chromatography–mass spectrometry analysis (LC–MS/MS). These experiments were conducted at the Functional Genomics Center Zürich. First, the samples were digested with 5 µl of trypsin (100 ng/µl in 10 mM HCl) and microwaved for 30 min at 60 °C. The samples were then dried, dissolved in 20 µl ddH₂0 with 0.1 % formic acid, diluted in 1:10 and transferred to autosampler vials for liquid chromatography with tandem mass spectrometry analysis. A total of 1 µl was injected on a nanoAcquity UPLC coupled to a Q-Exactive mass spectrometer (ThermoFisher). Database searches were performed by using the Mascot swissprot and tremble_streptomycetes search programs. For search results, stringent settings have been applied in Scaffold (1 % protein false discovery rate, a minimum of two peptides per protein, 0.1 % peptide false discovery rate). The results were visualized by Scaffold software ([Proteome Software Inc.], Version 4.11.1).

Protein gel electrophoresis and western blotting

For general protein analysis, protein samples were boiled at 95 °C for 5 min and separated by a 4–20% Mini-PROTEAN® TGX Stain-Free™ SDS PAGE (BioRad) or 12 % Tris-Glycine SDS PAGA (Invitrogen) and visualized using InstantBlue Coomassie Protein Stain or ReadyBlue Protein Gel Stain (Sigma Aldrich).

For western blotting, cells were grown in biological triplicate for 48 h in TSB, 2 ml aliquots of each culture were pelleted and washed once with 1x PBS. Cell pellets were resuspended in 0.4 ml of sonication buffer (20 mM Tris pH 8.0, 5 mM EDTA, 1x EDTA-free protease inhibitors [Sigma Aldrich]) and subjected to sonication at 4.5-micron amplitude for 7 cycles of 15 s on / 15 s off. Samples were centrifuged at 17,000 x g for 15 min at 4 °C. Protein concentration in cleared cell lysate supernatants was determined by Bradford Assay (Biorad). Equivalent total protein concentrations (1 mg/ml) were assayed using the semi-dry western blot transfer. The gel was transferred to a PVDF membrane (Biorad) and probed with the appropriate antibody diluted in TBS-T or using the iBIND buffer system (Thermo Fisher): Rabbit anti-FLAG ([Sigma Aldrich] F7425, 1:2500), Rabbit anti-WhiA ((44), 1:2500), and Goat Anti-Rabbit-HRP (Abcam Ab6721, 1:5000). To visualize HRP-conjugated antibodies, membranes were incubated with SuperSignal West Femto ECL solution (Thermo Fisher).

Cellular fractionation

To fractionate membrane and soluble proteins from S. coelicolor (WT, SS576), cells were grown for 48 h in 50 ml TSB/YEME and harvested by centrifugation. Cell pellets were resuspended in 1/10 volume of lysis buffer (0.2 M Tris-HCl, pH 8, 10 mg/mL lysozyme, and 1× EDTA-free protease inhibitors; [Roche]), incubated for 30 min at 37 °C, and then briefly cooled on ice before lysed by sonication (11 x 15 sec on/off at 50 % power at 8 microns on ice). Cell debris was removed by centrifugation at 16,000 × g for 20 min and supernatant was subsequently subjected to two additional rounds of centrifugation. The cleared cell lysate was subjected to ultracentrifugation for 1 h at 150,000 × g at 4 °C to separate the soluble proteins from membrane proteins. Sedimented membrane proteins were resuspended and washed twice in 1 volume of wash buffer (60 mM Tris-HCl, 200 mM NaCl, pH 8, 0.2 mM EDTA, and 0.2 M sucrose) at 150,000 × g at 4 °C for 1 h. The wash step was repeated one final time with the wash buffer containing 8 M urea to remove traces of membrane-associated proteins (45). The final pellet was dissolved in 1/10 of the initial volume with wash buffer (no urea). Equi-volume amounts of fractions were mixed with 2x SDS sample buffer and analyzed by immunoblotting.

Membrane protein topology analysis in E. coli

The coding sequence of CisA was inserted in the dual pho-lac reporter plasmid pKTop, which consists of an E. coli alkaline phosphatase fragment PhoA_22-472 fused in frame after the α-fragment of β-galactosidase LacZ_4-60 (Supplementary Table 3) (29). For membrane protein topology, E. coli TG1 was transformed with the pKTop and derivatives. Membrane protein topology was assayed by plating the resulting reporter strains on dual-indicator plates containing LB agar supplemented with 80 μg/ml 5-bromo-4-chloro-3-indolyl phosphate disodium salt (X-Phos) ([Sigma Aldrich], RES1364C-A101X) and 100 μg/mL 6-chloro-3-indolyl-β-d-galactoside (Red-Gal) ([Sigma Aldrich], B6149) as indicators, 1 mM IPTG, and 50 μg/ml kanamycin. A periplasmic or extracellular location of the reporter fusion results in higher alkaline phosphatase activity (blue color), whereas a cytosolic location of the reporter leads to higher β-galactosidase activity (magenta color). Plates were incubated for two days at 37 °C and scanned.

Vitrified sample preparations

For single-particle cryoEM (SPA), the S. coelicolor non-contractile CIS particles were purified as described above and vitrified using a Vitrobot Mark IV (Thermo Fisher Scientific). 4 µl of samples were applied twice on glow-discharged 200 mesh Quantifoil copper grids (R 2/2) which were manually coated with a layer of 1 nm carbon. Grids were blotted for 5.5 s and plunge-frozen in liquid ethane-propane mix (37 %/63 %). Frozen grids were stored in liquid nitrogen until loading into the microscope.

For cryo-electron tomography (cryoET), Streptomyces cells were mixed with 10 nm Protein A conjugated colloidal gold particles (1:10 v/v, [Cytodiagnostics]) and 4 µl of the mixture was applied to a glow-discharged holey-carbon copper EM grid (R2/1 or R2/2, [Quantifoil]). The grid was automatically blotted from the backside for 4-6 s in a Mark IV Vitrobot by using a Teflon sheet on the front pad, and plunge-frozen in a liquid ethane-propane mixture (37 %/63 %) cooled by a liquid nitrogen bath.

SPA data collection and image processing

44,925 movies were collected on Titan Krios microscope (Thermo Fisher Scientific), operated at 300 keV equipped with K3 direct electron detector, operating in counting mode, and using a slit width of 20 eV on a GIF-Quantum energy filter (Gatan). The automated data collection was conducted with EPU software (Thermo Fisher Scientific) with a final pixel size of 1.065Å/pix over 40 frames with a total dose of 50 e^-/Ų. The targeted defocus was set between -1.6 and -2.8 µm with 0.2 µm increment.

Movie-alignments with dose-weighting were performed in MotionCor2 software (46) with a gain reference estimated a posteriori (47) followed by standard in-house pipeline of manual inspection and selection of 38,845 micrographs (48) for further processing in cryoSPARC package v4.0.3 (49). Due to the low concentration of particles and the presence of other contaminants in the sample, automatic particle picking appeared challenging, and we developed CIS-specific approach for efficient picking of the cap and baseplates. In this approach, CIS^Sc particles were manually selected as filaments (start-to-end) on 920 micrographs and used for training a crYOLO (50) picking model (Supplementary Fig. 2). The large size of the training set appeared essential for successful picking; we could not get rid of large number of false-positive picks, but there were not so many false-negative picks. Then, CIS^Sc particles were picked in crYOLO as filaments, followed by the extraction of the ends of the “filaments” using star_modif.py script (48). Particle coordinates were imported in cryoSPARC, where 211,041 particles were extracted at a large box size of 756 pixels binned to 64 pixels for initial 2D-classification, which allowed us to separate initial particle sets corresponding to the CIS^Sc cap (36,569 particles) and CIS^Sc baseplate (43,087 particles) (Supplementary Fig. 2).

Two subsets of this dataset were processed in cryoSPARC separately. For the cap subset, a standard cryoSPARC single-particle processing approach was applied, including gradual iterative 2D-classification rounds, followed by selection of good classes and particle re-extraction at finer sampling. The main challenge was the proper centering of the particles selected on the end of elongated filament-like objects with repeating patterns (sheath-tube module of CIS), hampering the alignments. Ab-initio model generation with applied C6 symmetry was hampered by poor centering of 2D-classes due to elongated nature of the particles (Supplementary Fig. 2). Homo-refinement with this reference was performed followed by particle re-extraction and one more round of ab-initio model generation with applied C6 symmetry (Supplementary Fig. 2). The new 3D-reconstruction appeared reliable and was used for further iterative 3D-refinements (Homo-refinement and NU-refinements) accompanied by 2D-classification and CTF-refinement rounds to exclude particles with wrong Euler angle assignments. The final 3.4 Å resolution map with imposed C6 symmetry resulted from 19,218 particles.

The baseplate subset was processed using the same approach, resulting in a 3D-reconstruction with imposed C6 symmetry with a resolution of 3.5Å from 22,980 particles (Supplementary Fig. 2). The tip of the baseplate has a C3 symmetry, therefore to solve that part of the map correctly, we exported that particle stack to Relion4 (51) and took advantage of 2D-classification in Relion to select 18,124 particles, corresponding to the best 2D classes averages. 3D-refinement with relaxed symmetry resulted in a 3.8 Å 3D-resolution reconstruction with imposed C3 symmetry.

Cryo-electron tomography

Intact Streptomyces cells were imaged by cryoET as described previously (52). Images were recorded using a Titan Krios 300 kV microscope (Thermo Fisher Scientific) equipped with a Quantum LS imaging filter operated at a 20 eV slit width and with K3 Summit direct electron detectors (Gatan). Tilt series were collected using a bidirectional tilt-scheme from -60 to +60° in 2° increments. Total dose was 130-150 e^-/Ų and defocus was kept at -8 µm. Tilt series were acquired using SerialEM (53), drift-corrected using alignframes, and reconstructed using IMOD program suite (54). To enhance contrast, tomograms were deconvolved with a Wiener-like filter ‘tom_deconv’ (55).

Structural modeling of the CIS^Sc complex

The map qualities of different modules of CIS^Sc (cap and baseplate) allowed for de novo structural modelling (Supplementary Fig. 2-3). The atomic models of SCO4243-4248, SCO4252-4253 and SCO4260 were built de novo using COOT (56). The resulting models were iteratively refined using RosettaCM (57) and real-space refinement implemented in PHENIX (58). In cases where protein domains could only be partially modeled, side chains were not assigned. Final model validation was done using MolProbity (58) and correlation between models and the corresponding maps were estimated using mtriage (58). To illustrate the complete CIS^Sc structure, we generated a composite model by integrating symmetry-related protein subunits into a full model of the CIS^Sc, based on the consistent CIS^Sc length observed both in situ and in vitro (Fig. 2b).

Fluorescence microscopy

Fluorescence-based cell viability assays were performed as described previously (18). Briefly, to produce cytoplasmic sfGFP in Streptomyces, the coding sequence for sfGFP was introduced downstream of the constitutive promoter ermE* on an integrating plasmid. The plasmid was introduced by conjugation to S. coelicolor strains (SS430, SS575, SS576. An equal level of cytoplasmic sfGFP in the different strains was confirmed by Western blotting (Supplementary Fig. 11). Streptomyces strains were grown in 30 ml of TSB liquid culture at 30 °C with shaking at 250 rpm for 48 h. Where appropriate, nisin was added to a final concentration of 1 µg/ml 90 min prior to imaging. One ml aliquots of each culture were centrifuged for 5 min at 15,000 x g, washed twice with 1xPBS, resuspended in 1 ml of 1xPBS containing 5 µg/ml of the red-fluorescent membrane dye FM5-95 (Invitrogen) and then incubated in the dark at room temperature for 10 min followed by two wash steps with 1x PBS. Washed cell pellets were resuspended in a total volume of 50 µl PBS and 10 µl of each sample was spotted onto 1 % agar pads and subsequently imaged using the Thunder imager 3D cell culture microscope (Leica). First, tile scan images were acquired using the Las X Navigator plug-in software (Leica Application Suite X, Version 3.7.4.23463), and 100 regions of interest were picked manually. Then Z-stack images were acquired using a HC PL APO 100x objective with the following excitation wavelengths: GFP (475 nm) and TRX (555 nm). Images were processed using LasX software to apply thunder processing and maximum projection and FIJI to create segmentation and quantify the live (sfGFP)/total cells (FM5-95) area ratio (59). Statistical analyses were performed on data from biological triplicates (n=100 images for each experiment) using a one-way ANOVA and Tukey’s post-test in GraphPad Prism 9 (Version 9.3.1).

For imaging the subcellular localization of the CIS adaptor protein in Streptomyces, cells (strain JS69 or S. coelicolor M145) were grown for 48 h in TSB/YEME liquid medium. Two μl of each strain were spotted on 1 % agarose pads and subsequently imaged. Images were acquired using a Zeiss Axio Observer Z.1 inverted epifluorescence microscope fitted with a sCMOS camera (Hamamatsu Orca FLASH 4), a Zeiss Colibri 7 LED light source, and a Hamamatsu Orca Flash 4.0v3 sCMOS camera. mCherry fluorescence was detected using an excitation/emission bandwidth of 577–603 nm/614–659 nm. Images were collected using Zen Blue (Zeiss) and analyzed using Fiji (59).

Cover glass impression of Streptomyces spore chains

Spore titers of relevant strains were determined by dilution plating (18). 10⁷ colony forming units (CFU) of S. coelicolor strains (WT, SS539, SS557) were spread onto R2YE agar plates and grown at 30 °C. Sterile glass coverslips were gently applied to the top surface of each bacterial lawn after 48 h, 72 h and 96 h post inoculation. Coverslips were then mounted onto glass microscope slides and imaged using a 40x objective on a Leica Thunder Imager 3D Cell Culture. Images were processed using FIJI (59).

Actinorhodin production assay

S. coelicolor strains (WT, SS539, SS557) were inoculated into 30 ml R2YE liquid media at a final concentration of 1.5 x 10⁶ CFU/ml. Cultures were grown in baffled flasks at 30 °C overnight. Cultures were standardized to an OD₄₅₀ of 0.5 and inoculated in 30 ml of fresh R2YE liquid medium. For visual comparison of pigment production, images of the growing culture were taken between t = 0 and t = 72 h (as indicated in Supplementary Fig. 7). For quantification of total actinorhodin production, 480 µl of samples were collected at the same time points where images were taken. 120 µl of 5M KOH was added, samples were vortexed and centrifuged at 5,000 x g for 5 min. The weight of each tube was recorded. A Synergy 2 plate reader (Biotek) was used then to measure the absorbance of the supernatant at 640 nm. The absorbance was normalized by the weight of the wet pellet.

Structure prediction and in silico protein-protein interaction screen

CisA structure was predicted using AlphaFold2 (28) and further processed using UCSF Chimera (60) or ChimeraX-1.7.1 (61).

A pairwise in silico screen for possible interactions between CisA and components encoded by the S. coelicolor cis gene cluster (SCO4242-SCO4260) was conducted using the AlphaFold2 Multimer-based LazyAF pipeline (28,39,62). The confidence metric (ipTM) for the top model from each pairwise interaction was tabulated and visualised using Prism (Version 10.2.2) with ipTM scores >0.7, indicating a possible protein-protein interaction (63).

Bioinformatic analysis of CisA homologs

To identify CisA homologs, we performed a reciprocal BLAST search in 120 genomes of Streptomyces and other Actinomycete species that were previously reported to include a type IId eCIS (8,19) using the closely related cisA homologue from Streptomyces albus J1074 (YP_007743954.1) as a query (72% identical to CisA from S. coelicolor) (Supplementary Data 1). Of these 120 genomes, 75 had a reciprocal hit for CisA (Supplementary Data 2). Protein sequences of CisA homologs were aligned using Clustal Omega (64) and visualized with JalView (Version 2.11.3.3).

Data availability

Representative reconstructed tomograms (EMD-50935, EMD-50939, EMD-50940, EMD-50944, EMD-50948, EMD-50949) and SPA cryoEM maps (EMD-51564, EMD-51565 and EMD-51566) have been deposited in the Electron Microscopy Data Bank. Atomic models (PDB: 9GTP, PDB: 9GTR, PDB: 9GTS) have been deposited in the Protein Data Bank.

Acknowledgements

We thank ScopeM for instrument access at ETH Zürich. We thank the Functional Genomics Center Zürich for mass spectrometry support. Pilhofer Lab members are acknowledged for discussions. M.P. was supported by the Swiss National Science Foundation (31003A_179255/310030_212592), the European Research Council (679209), and the NOMIS foundation. Work in the lab of S.S. was supported by a Royal Society University Research Fellowship (URF\R1\180075 and URF\R\231009) and BBSRC grant BB/T015349/1 to S.S. and by the BBSRC Institute Strategic Program grant BB/X01097X/1 to the John Innes Centre.

Author contributions statement

B.C., S.S. and M.P. conceived the project. B.C. optimized the sample preparation for SPA; B.C. and P.A. collected and processed the cryoEM data, reconstructed the cryoEM map; B.C., P.E.H. and J.X. built and refined the structural models; B.C. and P.E.H. conducted and processed cryoET; B.C. conducted automated fluorescent light microscopy, determined sporulation efficiency and actinorhodin production; J.W.S. and S.S. generated plasmids and Streptomyces strains; J.W.S. designed constructs and performed membrane topology assay. J.W.S. prepared samples and performed cellular fractionation experiment. J.W.S. imaged strains expressing fluorescently tagged CisA. S.S. and G.C. performed bioinformatic analysis of CisA. B.C., J.W.S., S.S. and M.P. wrote the manuscript.

Competing interests statement

The authors declare no competing interests.