The B. subtilis strains used in this study are listed in Table 1. E. coli DH5α (Sambrook et al., 1989) was used for plasmid constructions and transformation using standard techniques (Sambrook et al., 1989). Luria Bertani (LB) broth was used to grow E. coli and B. subtilis. When required, media were supplemented with antibiotics at the following concentrations: ampicillin 100 μg ml^-1 (for E. coli) and kanamycin (10 μg ml^-1), chloramphenicol (5 μg ml^-1), or spectinomycin 150 μg ml^-1 (for B. subtilis). Sporulation (SP) plates were prepared by the addition of 15 g l^-1 Bacto agar (Difco) to SP medium (8 g of nutrient broth per liter, 1 mM MgSO₄, 13 mM KCl, supplemented after sterilization with 2.5 μM FeSO₄, 500 μM CaCl₂, and 10 μM MnCl₂). B. subtilis was transformed with plasmid DNA according to the two-step protocol (Kunst and Rapoport, 1995). Transformants were selected on SP plates containing antibiotics as above.

For the fluorescence microscopy imaging, B. subtilis cultures in exponential phase (OD₆₀₀ 0.2–0.6) or stationary phase (OD₆₀₀ 1.5) were incubated for 5 min at 37°C and 200 rpm with DAPI and Nile Red at a final concentration of 1 μg ml^-1, to stain the bacterial chromosome and the cell membrane, respectively. To interrupt transcription in the cells, cultures were incubated for 10 min at 37°C and 200 rpm with rifampicin (dissolved in methanol) at a final concentration of 200 μg ml^-1. As a negative control, the same amount of methanol was added to the cells. The expression of GFP fusions under the control of a xylose-inducible promoter was induced by addition of 0.1% xylose.

All plasmids used in this study are listed in Table 2. To express B. subtilis proteins under the control of their native expression signals, fused to the monomeric GFP, we constructed plasmid pBP43. For this purpose, we amplified the gfp gene fragment using the primer pair ML220/ML221 (Rothe et al., 2013) and plasmid pSH3 as the template (Oliva et al., 2010). The PCR product was digested with HindIII and cloned into the vector pUS19 (Benson and Haldenwang, 1993). The vector pBP43 allows the construction of GFP fusions and integration of the plasmid into the chromosome via Campbell-type recombination between the cloned gene fragment and the chromosomal copy of the gene¹. For the fusion of proteins implicated in RNA degradation to the GFP, we amplified the 3′ 600 bp (lacking a stop codon) of each gene, with the respective oligonucleotide pair (see Supplementary Table S1), using chromosomal DNA of B. subtilis 168 as the template. The PCR products were cloned between the BamHI and SalI sites of pBP43. The resulting plasmids were pGP2806, pGP2807, pGP2808, pGP2810, pGP2811, pGP2812 for pnpA, rnjB, eno, pfkA, cshA, and rnjA, respectively. Integration of these plasmids into the chromosome of B. subtilis leads to the in-frame fusion of the gfp alleles to the entire genes lacking their stop codon.

For the ectopic inducible overexpression of N-terminal GFP fusions, we used plasmid pHJS105 that allows integration into the amyE locus and induction of the fusion by the addition of xylose (Jahn et al., 2015). For the construction of the plasmids we amplified the rnjA and rnjB genes from chromosomal DNA of B. subtilis 168 using the oligonucleotide pairs NC63/NC64 and NC65/NC2, respectively. We then cloned the PCR fragments between the BamHI and HindIII sites of pHJS105 to obtain plasmids pGP2802 and pGP2803, respectively. RNase Y encoded by the rny gene has a N-terminal transmembrane helix (Lehnik-Habrink et al., 2011). Therefore, we had to fuse the GFP to the C terminus of RNase Y. For this purpose, we amplified the rny gene from chromosomal DNA of B. subtilis 168 and the gfp gene from pHJS105 using the primer pairs NC51/NC52 and NC53/NC54, respectively. Both PCR fragments were digested with BamHI and fused together. Subsequently, we cloned the resulting fragment between the AvrII and NotI sites of pHJS105. The resulting plasmid, pGP1449, allows the overexpression of RNase Y with a C-terminal GFP upon induction with xylose.

Deletion of the pfkA and pnpA genes was achieved by transformation with PCR products constructed using oligonucleotides (see Supplementary Table S1) to amplify DNA fragments flanking the target genes and intervening antibiotic resistance cassettes as described previously (Guérout-Fleury et al., 1995; Wach, 1996).

For fluorescence microscopy, cells were grown at 28°C in LB medium to an OD₆₀₀ of 0.2–0.5 (unless otherwise indicated). Fluorescence images were obtained with the AxioImager M2 fluorescence microscope, equipped with digital camera AxioCam MRm and AxioVision Rel 4.8 software for image processing and an EC Plan-NEOFLUAR 100X/1.3 objective (Carl Zeiss, Göttingen, Germany). The applied filter sets were the Filter set 49 (G 365, FT 395, BP 445/50; Carl Zeiss) for DAPI detection, the Filter set 38 (BP 470/40, FT 495, BP 525/50; Carl Zeiss) for GFP detection, and Filter set 43 (BP 545/25, FT 570, BP 605/70) for Nile Red visualization. The overlays of fluorescent and phase-contrast images were prepared for presentation with ImageJ 1.49p (National Institutes of Health, Bethesda, USA). Images were taken with an exposure time of 1 s for the GFP constructs (except for Eno-GFP, exposure time 500 ms), 500 ms for the membrane stain Nile Red, and 100 ms for the DNA stain DAPI.

To study the formation of the complex formed between the RNases J1 and J2, we used strain B. subtilis GP1048 which encodes RNase J1 fused to a Strep-tag to facilitate purification via a StrepTactin column (IBA GmbH, Göttingen, Germany) and RNase J2 fused to a triple FLAG tag to allow immunological detection. The bacteria were cultivated and Step-tagged protein purified as described previously (Meyer et al., 2011). The identity of the proteins was verified by Western blot analysis using antibodies directed against the Strep and the FLAG-tag, respectively.

Several RNases, RNA helicases, and glycolytic enzymes were proposed to form an RNA-degrading complex in B. subtilis. However, the subcellular localization of most of these proteins has not yet been determined by in vivo fluorescence labeling. The possibility of fusing fluorescent proteins to visualize the in vivo localization of proteins has greatly improved our understanding of the cell dynamics in the past years. However, the ability of the GFP to dimerize can cause artifacts in the localization of the fusion protein within the cells, which can be avoided by the use of monomeric fluorescent proteins (Miyawaki, 2011; Margolin, 2012). Moreover, the addition of a tag to a protein could affect its functionality and localization in vivo, by avoiding interactions, conformational changes, or proper folding (Gunka et al., 2013). Finally, overexpression of labeled proteins may result in aggregation and mislocalization (Miyawaki, 2011; Margolin, 2012). To study the localization of the proteins implicated in RNA degradation in B. subtilis, we therefore used a monomeric GFP variant (Oliva et al., 2010) and integrated the constructs into the chromosome, ensuring that the labeled protein is expressed from its native promoter. Additionally, we verified that these fusion proteins had retained their native activity in vivo.

Based on the rationale that if the fusions rendered the proteins inactive the strains would show a phenotype similar to the respective deletion mutants, we decided to compare the reporter strains to the respective wild type and deletion mutant strains. The deletion of the pnpA gene encoding PNPase causes the cells to grow in long cell chains and results in cold sensitivity (Wang and Bechhofer, 1996). B. subtilis GP1698 harboring the PnpA-GFP fusion was able to grow with a normal cell length (Figure 1A). RNase J1, encoded by the rnjA gene, has been regarded as essential until recently. Although the construction of a deletion mutant is possible, this strain shows severely affected cell morphology, with cells growing in curly long chains (Figaro et al., 2013). Our strain GP1722 (RNase J1-GFP) showed normal cell morphology and no chains (Figure 1B). Although RNase J2, the paralog of RNase J1, can be deleted without any phenotype, it has been shown to interact in vitro with RNase J1 with a 1:1 stoichiometry (Mathy et al., 2010). Moreover, we have observed a strong interaction of both proteins by in vivo co-purification (see Figure 2). As there is no assay for the functionality of RNase J2, we would take a similar localization of RNases J1 and J2 as an indication that the GFP tag did not affect J2 localization (see below). RNase Y is encoded by rny, the first gene of the bicistronic rny-ymdB operon (Diethmaier et al., 2011). To avoid interference with the expression of the phosphodiesterase YmdB, we constructed fusion expressing RNase Y-GFP under the control of a xylose-inducible promoter. As membrane localization of RNase Y has been demonstrated in several independent studies (Hunt et al., 2006; Lehnik-Habrink et al., 2011; Bührmann et al., 2012), no further functional analyses were performed. The cshA deletion mutant shows poor growth at temperatures lower than 28°C (Lehnik-Habrink et al., 2013) but strain GP1721 was able to grow at 28°C like the wild type (Figure 3A). The essentiality of the glycolytic enzymes Eno and PfkA has recently been revisited (Commichau et al., 2013). Eno was shown to be essential for the growth of B. subtilis in LB medium, while the pfkA gene is not essential but is required for growth on minimal medium with glucose as the carbon source. For strain GP1720 expressing the PfkA fused to GFP, we observed growth on minimal medium with glucose as the carbon source, indicating activity of the fusion protein (Figure 3B). The strain encoding the fusion of Eno to GFP (GP1700) was viable and grew as the wild type strain indicating functionality of Eno when fused to GFP (Figure 3C).

RNase Y has been proposed to provide the scaffold for RNA-degrading enzymes in B. subtilis. Several studies have reported membrane localization of this protein (Hunt et al., 2006; Zweers et al., 2009; Lehnik-Habrink et al., 2011; Bührmann et al., 2012), and our results are in excellent agreement with those observations (Figure 4). Interestingly, we observed a higher intensity of RNase Y localization at the division septa. In a recent study, RNase Y was found to interact with the dynamin protein DynA, which also localizes to the division septa (Bührmann et al., 2012).

Next, we investigated the localization of the exoribonucleases PnpA, RNases J1 and J2. For PnpA, we observed a cytoplasmic localization with an even distribution throughout the cell (see Figure 5A). RNases J1 and J2 were found to localize in the cytoplasm except in the central region of the cell. A DAPI stain identified this latter region as the nucleoid, the RNases J1 and J2 are therefore localized in the cytoplasm around the nucleoid (Figure 6). In order to confirm the exclusion of RNases J1 and J2 from the nucleoid region, we treated cultures expressing the fusion proteins with rifampicin to stop transcription. This transcription stop results in relaxation of the nucleoid; as a result the nucleoid delocalizes from the central region of the cell to the whole cytoplasm (Hunt et al., 2006). This delocalization of the nucleoid is shown in Figure 7. It is accompanied by a complete delocalization of RNases J1 and J2, which now appeared evenly distributed throughout the cytoplasm (see Figure 7). Similar localization and delocalization pattern upon rifampicin have been shown previously for ribosomes in B. subtilis (Mascarenhas et al., 2001; Lehnik-Habrink et al., 2013). Taken together, these observations suggest that both RNases J1 and J2 are localized to those areas in the cytoplasm where the RNA target molecules are present. Interestingly, very similar results have been obtained for the localization of the RNase Y-associated DEAD box helicase, CshA (Hunger et al., 2006), and we were able to confirm the localization of CshA in the cytoplasmic area that does also harbor the ribosomes (Mascarenhas et al., 2001; see Figure 6). Interestingly, for E. coli RNase E delocalization has been observed in response to rifampicin treatment (Strahl et al., 2015); such a delocalization was not detected for RNase Y in B. subtilis (data not shown). This difference may be caused by the different attachment of the two RNases to the membrane: RNase E is membrane-attached via an amphipathic helix (Khemici et al., 2008), whereas RNase Y is inserted to the membrane via a trans-membrane helix (Lehnik-Habrink et al., 2011).

For the glycolytic enzymes Eno and PfkA, we detected an even distribution in the cytoplasm (see Figure 5). Interestingly, Eno formed bright spots at the polar regions of some cells at 37°C but not at 28°C. Polar localization of Eno has been reported before and was shown to depend on phosphorylation of the enzyme on a tyrosine residue by the protein tyrosine kinase PtkA (Jers et al., 2010).