Protein localization measurements increasingly rely on fluorescent protein (FP) fusions, while immunofluorescence – the former gold standard – is being phased out due to labor-intensive procedures, the requirement for antibodies, and the potential for fixation artifacts. In studies of bacterial cells, the convenience of FP fusions quickly led to a sea change, revealing that these cells have a high degree of spatial organization and are far from ‘bags of enzymes’1, 2
. Some of the proteins that were found to be localized are involved in processes like cell division where spatial aspects are central, but many studies3-5
have reported that numerous other proteins also form foci. However, few localization results have been independently validated. We designed a function-based validation assay for protein localization patterns in live cells and applied it to a canonical example of a bacterial protein reported to form biologically relevant foci across a range of bacterial species: the Clp proteases3-12
Conceptually, our approach exploits the fact that localization patterns determine the statistical differences between the two daughter cells right after cell division, and thereby influence the post-division heterogeneity in any affected downstream processes. By measuring the downstream heterogeneity in the presence and absence of a fluorescent tag to a protein of interest, a side-by-side comparison reveals whether the tag interferes with protein localization; for non-intrusive tags, the post-division single-cell heterogeneity in the downstream process should be the same with or without the FP ().
Figure 1 (a) Schematic depiction of the segregation assay. An upstream process (Clp protease localization) affects a downstream process (substrate degradation) that can be measured in daughter cells originating from cells with and without an FP tag on the upstream (more ...)
We first fused two commonly used FPs, Venus YFP and superfolder GFP (sfGFP), to the ClpX and ClpP proteins in E. coli
and confirmed the previously reported localization patterns, typically observing a single bright focus in roughly half of the cells during balanced exponential growth in rich media. Cells without a focus exhibit a low cytoplasmic FP signal, similar to the cytoplasmic signal in foci-harboring cells. Tracking cells through division showed how the focus segregated to one of the two daughters, while the other cell formed a new focus within a few generations (Supplementary Video 1
and Supplementary Note 1
). The observed localization of the Clp-FP protease foci should thus cause substantial post-division single-cell heterogeneity in the turnover rates of protease substrates.
We then used dual-color time-lapse microscopy to simultaneously measure substrate abundances and protease localization patterns in individual E. coli
cells over time. As the reporter substrate we fused mCherry to the E. coli
ssrA tag, which marks mCherry for proteolysis by ClpXP () and to a lesser extent by ClpAP13
. We expressed mCherry-ssrA from an inducible promoter in both foci-forming FP strains and in the wildtype strain and followed the fate of the mCherry-ssrA degradation reporter over time. Specifically, we measured the reporter degradation rate in daughter cells after cell division and analyzed the heterogeneity between individual daughter cell pairs.
When either ClpX or ClpP was tagged with Venus YFP or sfGFP, cells that contained the focus actively degraded mCherry-ssrA, whereas cells without a focus showed mild to extreme reduction in mCherry-ssrA degradation, thus producing two daughters with very different mCherry-ssrA degradation rates (, Supplementary Video 2
and Supplementary Figure 1
). However, both daughter cells in the wildtype strain continued proteolysis of mCherry-ssrA at very similar rates (, Supplementary Video 3
and Supplementary Fig. 2
). This shows that the FP tag causes clustering artifacts, and that the ClpX-FP and ClpP-FP fusions cannot be trusted for determining the localization of the native, untagged proteins.
To further validate our results we performed several independent tests. We used immunofluorescence microscopy against ClpX () with the strain expressing the foci-forming ClpX-Venus YFP fusion as a positive control and a ClpX knockout strain as a negative control, confirming that the anti-ClpX antibodies were specific and that fixation did not disassemble the ClpX-Venus YFP foci. In wildtype cells, the immunofluorescence images indicate that ClpX forms 20–50 complexes that are uniformly distributed in the cell. We also used the small monomeric SNAP tag fused to ClpP () and ClpX (data not shown), and again observed a uniform spatial distribution of these proteins.
Figure 2 (a) Immunofluorescence microscopy of ClpX in wildtype (left), ClpX-Venus YFP (middle) and ΔclpX (right) strains. Insets are phase images and a close-up is shown for the wildtype. (b) Fluorescence images show bacteria expressing the ClpP-SNAP tag (more ...)
These findings motivated us to evaluate other FPs (Supplementary Table 1
) fused to ClpP or ClpX. We found that sfGFP, Venus YFP, mCherry, and mCherry2 all cause substantial foci formation in the majority of cells, despite being monomers or very weak dimers when expressed alone. mKate2 and TagRFP-T caused intermediate clustering, while for mVenus YFP and mYPet most of the fluorescence signal was spatially uniform, although foci were observed in a few cells. The mTagBFP and mEos2 fusions resulted in a weak signal with infrequent dim foci. We detected no foci for PS-CFP2, rsFastLime (data not shown) and GFP(–30) but the signal was very dim. Finally, mGFPmut3, Dronpa, and Dendra2 displayed an essentially uniform signal. FP fusions to ClpP generally caused more foci formation than fusions to ClpX, in particular for mYPet (). Because foci-forming tendencies could also be affected by protein expression levels, which in turn could be affected by the FP tags, we expressed two separate copies of the gene for ClpP-mGFPmut3 in the same strain. We observed no increase in clustering despite the higher level (Supplementary Fig. 3 and 4
We further analyzed the ClpP-SNAP tag, ClpP-Dronpa, ClpP-Dendra2 and ClpP-mGFPmut3 fusions using our single-cell segregation assay and observed very little post-division cell-to-cell variability, confirming that these tags, though not perfectly mimicking the wildtype, are less prone to clustering artifacts (). All ClpP-FP fusions also showed a somewhat reduced degradation activity when compared to the wildtype, presumably because the bulky FP tags interfere with protease activity. Of all the reporters tested, the SNAP tag was both the most active and least intrusive in terms of localization.
Figure 3 The plots show single-cell segregation assays for bacteria expressing the indicated proteins. Post-division single-cell degradation rates were measured by time-lapse fluorescence microscopy at 37 °C (upper row) and 30 °C (lower row) for (more ...)
Gentle fixation of cells harboring the ClpX-mGFPmut3 and ClpP-mGFPmut3 fusions also revealed uniformly distributed complexes (). We further used HILO microscopy (Online Methods) to perform real-time single-molecule imaging in live cells. ClpA-mGFPmut3, ClpP-mGFPmut3 and ClpX-mGFPmut3 complexes were all observed to move freely and rapidly in the cytoplasm (, Supplementary Videos 4
). Individual ClpP-Dronpa molecules could also be detected in live cells with HILO imaging and were also uniformly distributed (Supplementary Fig. 10
The bright ClpP-FP foci are proteolytically active and highly fluorescent, showing that the fusions are functional and not misfolded. Introducing the monomeric A206K mutation14
into an FP also substantially reduced foci formation, again demonstrating that the foci are not caused by spontaneously misfolded FPs. Even the strongest foci-forming FPs, like Venus YFP, are spatially uniform when expressed alone, even at high levels (Supplementary Fig. 5
). Both the Clp complexes and the FPs are thus spatially uniform on their own, and only form foci when fused to each other. We hypothesize that this is due to avidity effects. In FP fusions, the homo-oligomeric proteins could act as scaffolds, bringing several FPs into close proximity. This would prevent the oligomers from diffusing apart after a single FP-FP dissociation event and allow them to rebind before the remaining links are broken (), thereby driving the coalescence of tagged oligomers into visible foci.
These results raise the question of how many other reported foci are caused or greatly exaggerated by FP fusions. In fact, the FPs we observed to be prone to clustering are used in the three main bacterial FP fusion libraries – mCherry in C. crescentus5
, Venus YFP in E. coli4
, and GFPuv4 in the E. coli
– that all report numerous foci. The ClpX and ClpP foci have even been used as positive controls in genome-wide localization studies5
. To investigate whether FPs cause false localization patterns more generally, we used one of our most monomeric FPs, mGFPmut3, to re-tag five E. coli
proteins – Hfq, PepP, IbpA, FruK, and MviM – that previously were reported to form bright foci in multiple FP libraries3, 4
. Fusions PepP-mGFPmut3, FruK-mGFPmut3, and MviM-mGFPmut3 showed no foci, while Hfq-mGFPmut3 and IbpA-mGFPmut3 were uniformly distributed in most cells and only showed dim foci in a small fraction of cells (Supplementary Fig. 6
, Supplementary Video 8
and Supplementary Table 2
). Our results strongly suggest that FP-mediated clustering is a widespread phenomenon although further tests, as presented for the Clp proteins, will be necessary to prove this unequivocally.
The segregation-based assay described here cannot be used for all cellular components since we cannot always measure the heterogeneity in an affected downstream process with existing reporters. However, the assay could be used for any factor that directly or indirectly affects transcription, translation, RNA degradation or proteolysis: protein localization patterns can be analyzed with FP reporters and mRNA localization patterns with the MS2 tagging system or FISH (although it should be noted that these patterns themselves may be prone to artifacts). It may also be possible to probe segregation of factors involved in other types of processes, using light microscopy to determine cell morphology, for instance, or using FRET biosensors to measure pH, metabolites, ATP levels or Ca2+ ions. Because the assay is based on a relative comparison with and without a tag, it is insensitive to systematic measurement errors and can resolve small statistical differences. For example, if two different FPs produce different localization patterns (as expected from ), testing which FP interferes less with the heterogeneity of a downstream process could suggest which reporter is more trustworthy.
We hope the results described here will lead to a reinvestigation of protein localization in bacteria, that the FP survey will guide the choice of fluorescent reporters both for conventional and super-resolution localization measurements, and that the segregation-based assay will prove useful in other biological systems.