Estimates of differential toxin expression governing heterogeneous intracellular lifespans of Streptococcus pneumoniae Free

Host intracellular defence mechanisms ensure effective pathogen elimination and maintain cytosolic sterility. However, many intracellular bacterial pathogens reside within a customized vacuolar compartment that not only allows bacterial replication but also protects the microorganism from cytosolic antimicrobial strategies (Anand et al., 2020; Omotade and Roy, 2019). For many pathogens, residence inside these vacuolar compartments is compromised due to expression of pore-forming toxins by the pathogen. The loss of integrity of these vacuoles exposes the microbes to diverse fail-safe intracellular defence mechanisms. Maintenance of the integrity of these vacuoles is, therefore, critical for prolonged persistence of these pathogens, which could be essential for other aspects of their lifecycle (Brawn et al., 2007; Moreira et al., 1997). However, the dynamics of toxin expression and subsequent pore formation – which promote interaction with cytosolic immune systems leading to clearance of the pathogen, or which promote evasion from host defences, ensuring the safety of the pathogen and promoting intracellular life – are not known.

Streptococcus pneumoniae (SPN), the Gram-positive opportunistic pathogen, commonly resides in the upper respiratory tract of healthy individuals as commensal. But from time to time, it disperses from its niche and causes life-threatening diseases such as pneumonia, septicaemia, endocarditis and meningitis (Pereira et al., 2022). Although typically considered as an extracellular pathogen, it does have a brief intracellular stint that is particularly pertinent in trafficking across host barriers, such as the blood–brain barrier and lung epithelial barrier (Anil and Banerjee, 2020), which are key aspects of its disease pathogenesis. In recent years, a few studies have identified that SPN replicates and persists for prolonged periods inside splenic macrophages and that this serves as a transient reservoir for dissemination into the blood, causing septicaemia (Ercoli et al., 2018). Additionally, SPN has been shown to replicate within cardiomyocytes (Brissac et al., 2018) and few serotype I SPN strains have been shown to adapt to a benign intracellular niche in the lung epithelium (Badgujar et al., 2020). However, whether these intracellular niches constitute a reservoir for SPN to re-establish colonization, thereby facilitating transmission to new hosts, remains elusive. But these findings indicate that pneumococci might be evolving via an alternative evolutionary trajectory by embracing an intracellular lifestyle and inducing a milder form of disease for efficient transmission (Badgujar et al., 2020).

Central to the intracellular life of SPN is pneumolysin (Ply), the cholesterol-dependent pore-forming toxin secreted by all serotypes of SPN, which promotes cell lysis and extensive tissue damage in the host (Subramanian et al., 2019b). Although Ply is reported to bind to mannose receptor C type 1 (MRC1) and sialyl Lewis X antigen (Subramanian et al., 2019b), its cytolytic and inflammatory activity is primarily governed by membrane cholesterol binding (Nollmann et al., 2004). Like any other cholesterol-dependant cytolysin, 40 to 50 Ply monomers assemble on the cell membrane to form a pre-pore complex, which converts into ring-shaped functional pore of diameter 200–500 Å, sufficiently large to cause haemolysis (Lawrence et al., 2015; Marshall et al., 2015; van Pee et al., 2017). But the pore sizes are not always of similar pattern or of equal diameter. High-speed atomic force microscopy analysis has revealed that diverse shapes such as arcs or slits can also form functional pores and still exhibit cytolytic activity (Jiao et al., 2021). The functional significance of such structural diversification among pore sizes is yet to be determined. Moreover, these in vitro experiments performed with purified Ply do not provide the real scenario prevailing during the interaction of SPN with host, wherein intravacuolar accumulation of Ply presumably governs the fate of the pathogen.

We have previously established the key role of Ply in mediating the interaction of SPN with intracellular defence mechanisms (Surve and Banerjee, 2019). We demonstrated that Ply inflicted damage on endomembrane-triggered recruitment of various endomembrane damage sensors such as galectins or ubiquitin. These damage sensors can act as ‘eat-me’ signals that direct the host to shunt the SPN-containing damaged vacuoles towards autophagy or ubiquitin-mediated degradation pathways (Surve et al., 2018). However, the correlation between expression levels of Ply by SPN in the intravacuolar environment governing pore dynamics and the subsequent choice of fates fostering its intracellular life was not established. In this study, we quantitatively relate the dynamics of pore formation and SPN degradation to a model of ply gene expression, which provides an estimate of intravacuolar Ply accumulation dictating the fate of SPN. A threshold crossing of Ply quantity is a first-passage event, which in general refers to the occurrence of a relevant event for the first time in a stochastic process (Redner et al., 2014; Chou and D'Orsogna, 2014). Several cellular events in bacteria and yeast have been modelled in recent years as first-passage processes. For example, lysis of Escherichia coli cells due to the accumulation of the λ-phage holin protein beyond the threshold quantity has been mathematically treated as a first-passage process and its time statistics have been extensively studied (Blasi et al., 1999; Ghusinga et al., 2017; Rijal et al., 2020, 2022; Singh and Dennehy, 2014; White et al., 2011). Various other biological events such as the timing of kinetochore capture by microtubules of fission yeast, accumulation of the critical cell division protein FtsZ during cell division in E. coli and cell-size threshold regulating division timing in Caulobacter crescentus have also been treated as first-passage processes (Iyer-Biswas et al., 2014; Kalinina et al., 2013; Nayak et al., 2020; Sekar et al., 2018). However, no such quantitative studies have been performed on the degradation time of any intracellular pathogen using this approach. As Ply expression is established to be stochastic, its accumulation on the endomembrane beyond a threshold amount, forming a variety of pores, can be considered as a first-passage process. Here, using stochastic modelling, we provide quantitative estimate of the differential rate of Ply expression in two distinct subpopulations of SPN, undergoing degradation by two distinct pathways, i.e. vacuolar and cytosolic. Our analysis not only revealed the contribution of Ply expression regulating the bacterial life spans inside the host, but also revealed the strategy that SPN adopts to secure a safe intracellular life for prolonged periods.

An isogenic set of intracellular SPN gives rise to different intracellular subpopulations with respect to recruitment of various host-specific markers onto pathogen-containing vacuoles (PCVs). This differential recruitment of various markers was attributed to heterogeneous expression of Ply (Surve et al., 2018). However, the real-time kinetics of their recruitment and degradation dynamics of SPN in these PCVs remain unexplored. We therefore monitored a tRFP-expressing wild-type (WT) SPN strain trapped inside YFP–Gal8-positive compartments in A549 cells by time-lapse confocal microscopy for several hours post infection. Expectedly, two distinct phenomena based on the degradative fate of SPN were evident in our experiments. In one phenomenon, the tRFP signal of SPN gradually faded inside the Gal8 vacuole, suggestive of probable degradation of SPN inside Gal8-marked autophagosomes (Fig. 1A–C, Movie 1). Gal8 is an endosome damage-sensing marker that interacts with the autophagy marker LC3, ensuring sequestration of the damaged endosomes in autophagosomes and subsequent degradation of the cargo following fusion with lysosomes (Surve et al., 2018; Thurston et al., 2012). Quantitative analysis of a series of time-lapse imaging events (n=50) with WT SPN revealed that although the mean Gal8 recruitment time (t_g8) on PCVs was 80.4±15.95 min (±s.d.), the average lifespan of SPN inside these Gal8-marked compartments (t_v) was 184.84±19.49 min (Fig. 1D,E). Combined together, the mean lifetime of SPN in the Gal8-marked vacuolar environment (t_Vac) was found to be 265.24±21.39 min (Fig. 1F). The absence of the tRFP signal was found to be equivalent to bacterial death as revealed by time-lapse fluorescence imaging and quantification of colony-forming units (CFUs) of tRFP-expressing SPN following the addition of gentamicin, an antibiotic, in SPN culture medium (Fig. S1A–C). Moreover, a Δply mutant SPN strain, tagged similarly with tRFP, did not associate with Gal8 and remained fluorescent inside host cells for prolonged periods, alleviating concerns about photobleaching (Fig. S1D; Movie 2).

Although 55.5% of SPN were eliminated inside Gal8-marked autophagosomes, for a significantly high proportion (44%) of pneumococci, there was catastrophic damage of Gal8 structures, triggering escape of SPN into the cytosol (Fig. 2A–C, Movie 3). By measuring the kinetics of a large number of such events (n=40), we noticed that in these populations, the average time of Gal8 recruitment (t_g8) was 137.4±19.35 min, and within 69.9±7.072 min of Gal8 recruitment, the vacuoles collapsed (t_p, time required for catastrophic PCV puncture and release of SPN to the cytosol), allowing SPN to become cytosolic. In the cytosol (Gal8^–), these pneumococci were ubiquitinated (Fig. S1E,F) and, after persisting for 75.85±10.64 min in the cytosolic milieu (t_c, degradation time for SPN in the cytosol) (Fig. 2D–F), they were finally degraded either via autophagy or proteasomal pathway, depending on the type of ubiquitin chain topologies formed (Bhutda et al., 2022). Combining all of these, the average lifetime (t_Cyt, total time required for cytosol-bound SPN elimination) of these SPN populations was 283.15±18.21 min (Fig. 2G).

The two distinct intracellular fates of SPN (degradation in Gal8-marked compartments and vacuolar escape followed by cytosolic degradation) could be attributed to heterogeneity in Ply expression (Surve et al., 2018). To verify this, we created a GFP–Ply-expressing SPN strain, in which Ply expression could be estimated with GFP fluorescence and which retained haemolytic activity similar to that of WT SPN (Fig. S2A,B). Additionally, this strain exhibited heterogeneous GFP expression (Fig. S2C,D), similar to heterogeneous Ply expression in WT SPN. We used this engineered SPN strain to assess whether heterogeneity in Ply expression drives the differential fate of SPN (Movies 4 and 5). Indeed, we observed that although SPN with lower expression of GFP (or Ply; i.e. the GFP fluorescence intensity in these cells was ∼2- to 3-fold lower compared to that in SPN cells with high GFP expression) succumbed inside Gal8-marked vacuoles, SPN with higher GFP (or Ply) expression escaped into the cytosol from Gal8-marked vacuoles and was killed there (Fig. S2E,F). Collectively, these data suggest that the fate of SPN inside host cells is governed by the extent of Ply expression.

While closely analysing the clearance kinetics of different SPN populations, we noticed a discrepancy in Gal8 recruitment timing between the vacuolar and cytosolic subpopulations. The average time of Gal8 association (t_g8) for vacuolar events (80.4±15.95 min) was ∼1.7-fold lower than that for cytosolic events (137.4±19.35 min). This was dubious as within the SPN population with heterogeneous Ply expression (Fig. S3A), the eventual cytosolic population expressing higher Ply (Fig. S2F) should presumably trigger quicker pore formation and thus have faster Gal8 association. To further shed light on this paradoxical observation, we created two genetically modified strains of SPN expressing either high (SPN:Ply-High) or low (SPN:Ply-Low) amounts of Ply (Fig. S3B–D) (Amaral et al., 2015) and infected A549 cells with these two SPN strains, maintaining cell viability (Fig. S3E). As expected, the SPN:Ply-High strain damaged the endomembrane and escaped into the cytosol, whereas the SPN:Ply-Low strain was predominantly present in PCVs after prolonged incubation (Fig. S3F). Additionally, the SPN:Ply-Low strain exhibited ∼4-fold higher intracellular survival ability compared to that of the SPN:Ply-High strain (Fig. S3G). We then analysed the kinetics of Gal8 recruitment in these engineered SPN strains by time-lapse fluorescence imaging (Movies 6 and 7). Indeed, the SPN:Ply-High strain recruited Gal8 later (∼181.76 min) compared to the SPN:Ply-Low strain (∼144 min) (Fig. 3A), corroborating our earlier observations. These distributions of t_g8 are plotted in Fig. S4A,B. We made some speculations to explain this anomaly. As per our hypothesis, for the SPN:Ply-High strain, the initial burst in Ply expression formed small pores that might not have been large enough to recruit Gal8, but could create a pH imbalance. These small pores might have delayed PCV acidification, which was required for Ply activation and subsequent formation of larger pores detectable by Gal8. In the case of the SPN:Ply-Low, the PCVs could acidify progressively as there was not sufficient Ply initially to form pores and cause a pH imbalance. During acidification of these PCVs, sufficient levels of Ply accumulated, which could be activated owing to the acidic environment and form large pores detectable by Gal8. To prove these hypotheses, as a marker for initial small pores, we expressed mCherry-tagged lysenin, which detects sphingolipids, in A549 cells. Sphingolipids are typically present in the luminal leaflet of endosomes and undergo transbilayer movement to the cytosolic leaflet of the endomembrane due to the formation of very minute pores (Ellison et al., 2020). Time-lapse fluorescence imaging demonstrated that the time taken for SPN:Ply-High and SPN:Ply-Low strains to recruit lysenin (t_lys) was ∼26.5 and 99 min, respectively (Fig. 3A, Movies 8 and 9). This revealed that, indeed, the SPN:Ply-High strain generated functional pores quickly on the endomembrane, whereas the SPN:Ply-Low strain was unable to form pores at a similar time point. A recent report suggested that galectin 3 (Gal3) is also an endomembrane damage sensor and its role was found to be pivotal for shunting the damaged endosome towards either the repair pathway or sequestration by autophagosomes (Jia et al., 2020a). We therefore assessed Gal3 recruitment time (t_g3) to PCVs and found that Gal3 decoration happened at ∼31.8 and 102 min for the SPN:Ply-High and SPN:Ply-Low strains, respectively (Fig. 3A, Movies 10 and 11). The distributions of t_g3 are shown in Fig. S4C,D. We also analysed the t_g3 distribution for WT SPN (Fig. S4E, Movie 12) and observed that the mean time of Gal3 decoration for WT SPN is ∼65 min, which is between that for the SPN:Ply-High and SPN:Ply-Low strains.

Comparison of t_lys and t_g3 with t_g8 implied that initial minute damage to the endomembrane triggered lysenin recruitment, closely followed by Gal3 recruitment, and subsequent major damage led to Gal8 association. To prove this unambiguously, we examined the leakage of different sizes of FITC–dextrans (3 kDa and 60 kDa) from PCVs containing SPN:Ply-High and SPN:Ply-Low strains and marked with either Gal3 or Gal8 (Fig. 3B,C). We first validated our assay using the Δply mutant, which did not form pores and therefore did not lead to leakage of FITC–dextran (Fig. S5A). As expected, treatment of cells with L-leucyl-L-leucine methyl ester (LLOMe), which artificially forms pores, triggered leaching of FITC–dextran from Δply-containing vacuoles as well as their decoration by Gal8 (Fig. S5B). We also found that, irrespective of the strains used, Gal3-decorated PCVs leaked 3 kDa FITC–dextran but retained 60 kDa FITC–dextran. On the contrary, both 3 and 60 kDa dextrans leaked from Gal8-marked PCVs. These data highlight that Gal3 senses smaller endomembrane damage, whereas Gal8 association requires larger pore formation. We also observed that although SPN:Ply-High-containing vacuoles leaked 3 kDa FITC–dextran within ∼20 min, SPN:Ply-Low-containing vacuoles retained this dextran until ∼80 min. This implies that the SPN:Ply-High strain forms pores quicker, possibly due to initial high amount of Ply accumulation within PCVs, whereas the SPN:Ply-Low strain did not form functional pores that early. To further prove this, we incubated cholesterol-containing liposomes loaded with FITC–dextran (3 kDa) and rhodamine–dextran (60 kDa) and with various concentrations of Ply (Fig. 3D). We noticed that at low Ply concentrations (0.001 to 0.01 μM), 3 kDa FITC–dextran leakage enhanced gradually without any release of 60 kDa rhodamine–dextran. In contrast, at high Ply concentrations (0.2 to 1 μM), parallel release of both high- and low-molecular mass fluorescently labelled dextrans was noticed (Fig. 3E), clearly indicating that the pore size or diameter is governed by Ply concentration, where low Ply concentration can create only small pores, whereas high Ply concentration can create larger pores. As per our earlier speculation, these initial small pores formed by the SPN:Ply-High strain should cause pH imbalance, impeding further Ply activation and subsequent larger pore formation. To unequivocally prove this, we measured the temporal change in the pH of the PCVs containing both these strains. The pH of PCVs was assessed from a standard curve (Fig. S5C) and validated by measuring the pH of PCVs harbouring the Δply mutant, which gradually acidified (owing to the inability to form pores because of the absence of Ply) and de-acidified following treatment with LLOMe (Fig. S5D). For the SPN:Ply-High-containing vacuoles, starting from a near-neutral pH (∼7.18), there was no change until 50 min, whereas the pH of PCVs harbouring SPN:Ply-Low reduced to 5.7 during the same time (Fig. 3F). This suggests that initial small pores formed by the SPN:Ply-High strain inhibited the acidification of PCVs for a substantially long time. However, beyond 50 min, the pH of the PCVs started to drop, presumably due to the function of vATPase, which pumps protons into the lumen of vacuoles, triggering their acidification and activation of Ply. Indeed, the haemolytic activity of Ply, which reflects large pore formation, was found to be higher at acidic pH (Fig. S5E). Additionally, following incubation of fluorescently labelled dextran-loaded liposomes with very low concentration of Ply (0.001 μM) at different pH levels, leakage of both high- and low-molecular mass dextrans was observed between pH 5 and 5.5 (Fig. 3G). We also observed formation of higher-molecular mass oligomers of purified Ply, primarily at pH 5 and 5.5 (a low degree of oligomerization was also observed at pH 4.5), which was absent in near-neutral pH conditions (Fig. 3H). These findings imply that acidic pH can amplify pore-forming capability (in terms of kinetics as well as pore size) of Ply. As acidification of PCVs is restored by vATPase and acidic pH triggers Ply activation and larger pore formation, we treated the cells with bafilomycin A1, an inhibitor of vATPase, and observed 40% reduction of Gal8 association with SPN:Ply-High-containing vacuoles (Fig. 3I). These data provide evidence of vATPase-mediated lowering of vacuolar pH for Ply activation and larger pore formation, leading to Gal8 recruitment (Fig. 3J).

Extensive damage in the PCV membrane has been proven to be detrimental for SPN as, following cytosolic exposure, it is efficiently eliminated by host cytosolic surveillance and bacterial clearance mechanisms (Surve et al., 2018). We therefore hypothesised that remaining confined within PCVs could be a preferred option for prolonged intracellular persistence. Such a lifestyle can be ensured by very low or no Ply expression, which promoted longer intracellular persistence.

As the SPN:Ply-Low strain caused smaller pores and Gal3 detected such pores to trigger either vacuole repair or autophagic sequestration, we explored the fate of low-Ply-producing SPN variants following infection in A549 cells stably expressing mStrawberry–Gal3 and GFP–LC3. We observed that one subset of SPN:Ply-Low cells was marked with Gal3, but they did not associate with the autophagy marker LC3 (Fig. 4A,B, Movie 13). These bacteria were possibly driven towards the Gal3-dependent endomembrane repair pathway. Another subset was decorated with both Gal3 and LC3 (Fig. 4C,D, Movie 14) and possibly reflected the Gal3-mediated autophagy pathway, which might include clearance or clearance-independent pathways. The mean survival times for these two subpopulations were 199 and 267 min, respectively (Fig. 4G). Our time-lapse imaging also revealed the presence of another low-Ply-producing SPN subset, which was decorated with LC3 but was devoid of Gal3 (Fig. 4E,F, Movie 15). This highlighted that recruitment of LC3 in this SPN population did not depend on endomembrane damage sensing, but could possibly be attributed to a pH or osmotic imbalance created by extremely small pores of Ply produced by the SPN:Ply-Low strain, as has been suggested for the VacA toxin-producing Helicobacter pylori (Florey et al., 2015). Critically, this Gal3⁻ LC3⁺ subpopulation represented the largest fraction of the SPN:Ply-Low population (Fig. 4H) and exhibited prolonged persistence (9 h and beyond) (Fig. 4G). Collectively, these represent the different SPN subpopulations arising due to variance in low Ply expression (Fig. 4I).

We hypothesized that the Gal3⁻ LC3⁺ population originated due to extremely low Ply synthesis, resulting in negligible damage to the endomembrane. Such small pores are not sufficient for recognition by damage sensors such as Gal3 but are adequate to create pH imbalance, promoting non-canonical LC3 lipidation of PCVs and simultaneously halting PCV maturation to prolong SPN persistence. To prove this, we infected A549 cells with the Δply mutant SPN strain, which does not have any pore-forming ability and therefore did not associate with LC3 or any other damage-sensing markers, such as Gal3 or Gal8 (Fig. 5A). To mimic low-Ply-mediated osmotic imbalance in the PCVs containing the Δply mutant strain, A549 cells were treated with monensin sodium salt (50 μM), a Na⁺/H⁺ ionophore, after infection with the Δply mutant strain. We observed that with increasing non-cytotoxic concentration of monensin (Fig. 5C), higher percentages (37%) of Δply-containing vacuoles were marked with LC3, but they were devoid of the damage sensor Gal3 (Fig. 5B). Furthermore, time-lapse imaging showed that these SPN (LC3⁺ Gal3⁻ Δply) continued to persist until 10 h and beyond (Fig. 5D,E, Movie 16). These findings prove that the induction of osmotic imbalance either artificially by using an ionophore or due to low Ply production inside PCVs not only triggered LC3 lipidation without assistance from canonical autophagy components, but also prolonged the persistence of SPN in the host cells.

Although we have been able to clearly delineate the role of Ply in governing the fate of intracellular SPN, the quantitative details of the ply gene expression pattern triggering such outcomes remained unknown. To build a simple mathematical model depicting the relationship between SPN lifespan and ply expression, we hypothesized that pore formation on PCVs and subsequent escape of SPN to the cytosol is a result of Ply accumulation crossing a certain threshold. Thus, apart from protein production, all other associated biophysical steps (such as protein deposition on the endosomal membrane, oligomerization, etc.) were ignored at the first approximation. Such a reduced picture of an effective first-passage process of protein threshold crossing has been previously reported to understand the timings of many significant cellular events such as lysis, cell division and kinetochore capture (Blasi et al., 1999; Ghusinga et al., 2017; Nayak et al., 2020; Rijal et al., 2020, 2022; Sekar et al., 2018; Singh and Dennehy, 2014; White et al., 2011).

When Ply accumulation in the proximity of the PCV wall causes damages to it, Gal8 molecules are recruited from the cytosol to those damaged points in random times (t_g8). The histograms for t_g8 for the vacuolar and cytosolic cases (from Figs 1D and 2D), are plotted in Fig. S4F,G. They follow exponential distributions with decay constants ∼0.0124 min⁻¹ and 0.0073 min⁻¹ (consistent with the mean times ∼80 min and ∼137 min as shown in Figs 1D and 2D). The distributions are expected to be exponential as the Gal8 attachment to a PCV is likely to be a single-step process.

For the SPN subpopulation that dies in the cytosol after escaping through the pores in PCVs, histograms for the total time of residence of the SPN inside PCVs before escaping to the cytosol (t_C′) and the time of larger pore formation after Gal8 attachment enabling escape to the cytosol (t_p) are shown in Fig. 6A,B (using data from Fig. 2D,E). We assumed that during the periods t_C′ or t_p, the Ply number increases from an initial value (N₀) to a final value (N) due to Ply translation. The number N−N₀ for Ply is expected to be larger for t_C′ compared to that for t_p. The N−N₀ number, the protein production rate (k) and the mean burst size (b) are unknown parameters that can be obtained by fitting the experimental histogram to the theoretical formula in Eqn 1 (see Materials and Methods). By best fitting (see Materials and Methods) the experimental data for t_p (Fig. 6B), we estimated the parameter values of Ply burst size (b) to be 2, the rate of Ply production (k) to be 0.107 min⁻¹ and the number of Ply required to reach the threshold (N−N₀) to be 13. We found that the theoretical formula (red line) fitted the experimental data quite well and reproduced the skewed non-Gaussian nature to the right of the maximum. With our fitted values of b, k and N−N₀, using Eqn 3 (see Materials and Methods), we get the theoretical standard deviation σ(t_p)=38.7 min (close to the experimental σ≈44.7 min). When we fitted the experimental data for t_C′ (Fig. 6A) in a similar way as the theoretical formula, the value N−N₀ turned out to be 35. This provided us with a rough estimate of the intravacuolar number of Ply (35−13=22) accumulated during the time of Gal8 attachment (t_g8≈ 137 min). Thus, such a model not only provided the rate of Ply synthesis (k≈0.107 min⁻¹), but also enabled quantitative estimates of Ply accumulated during Gal8 attachment (∼22) and larger pore formation on PCV membrane post Gal8 attachment to facilitate cytosolic escape (∼13). The histogram (using data from Fig. 2F) of timescales of SPN degradation in the cytosol after escape from PCVs (t_c) is shown in Fig. S4H, but as pneumococcal degradation post cytosolic escape is independent of Ply, we did not attempt to model the statistics of these timescales.

In the case of vacuolar degradation, the bacteria were eventually degraded within the Gal8-marked autophagosomes. We hypothesized that due to relatively low Ply production, this subset of SPN could not create significantly large pores on the PCV membrane to escape to the cytosol. Thus, we expected that our theoretical model would provide a lower threshold number of Ply in this case compared to that of the cytosolic SPN population discussed above. We first plotted the histograms of the total vacuolar degradation times (t_Vac=t_g8+t_v) in Fig. 6C (using Fig. 1F), as well as the degradation times after Gal8 attachment (t_v) in Fig. 6D (using Fig. 1E). The histogram in Fig. 6C exhibits bimodality, suggesting the existence of two separate peaks at different time points. This could presumably be due to distinct subpopulations of WT SPN that produce Ply at distinct rates within different vacuoles, and because of which they have shorter or longer lifetimes. This is supported by our experimental study of a low-Ply-expressing mutant (SPN:Ply-Low) having the distinct markers Gal3⁺ LC3⁺, Gal3⁺ LC3⁻ and Gal3⁻ LC3⁻ associated with different SPN subpopulations. Among these, we found that the Gal3⁻ LC3⁺-marked subset was the majority and longest living, with lifetimes (t_m) ranging from ∼250 min to >600 min, with a significant rise beyond 400 min (Fig. 6E). Hence, in Fig. 6D, we intentionally plotted histogram of timescales below 400 min, which presumably represent a moderate-Ply-producing subpopulation. Those above 400 min (corresponding to extremely low-Ply-producing subsets) were ignored. Upon fitting the experimental histogram (Fig. 6D) with the theoretical Eqn 1 (Materials and Methods), we obtained the mean Ply burst size (b) to be 2, the rate of Ply production (k) to be 0.0412 min⁻¹ and increment of Ply number (N−N₀) to be 10, during the lifespan of SPN following Gal8 attachment and its eventual death. The theoretical fit is shown as a solid red line in Fig. 6D. With the fitted values of b, k and N−N₀, we obtained a standard deviation (using Eqn 3 in the Materials and Methods) of 89.2 min (compared to the experimental σ≈85.5 min).

Similar to the cytosolic population, a fit to the experimental data for t_Vac (Fig. 6C) yielded N−N₀=19. Again, this helped to find a rough estimate for the number of Ply (19−10=9) that accumulated in the vacuole before (t_g8≈80 min) Gal8 attachment. Critically, our analysis suggests that (1) the rate of Ply synthesis in the case of the SPN subset undergoing vacuolar degradation is 2.6-fold lower (0.107 min⁻¹ versus 0.0412 min⁻¹) than that for the SPN subpopulation undergoing cytosolic death and (2) the number of Ply accumulated in the case of the SPN subset undergoing vacuolar degradation (∼19) is almost half the number of Ply needed for cytosolic escape (∼35).

Intracellular defences in the form of the endolysosomal pathway and various forms of autophagy or the ubiquitin–proteasome system safely dispose of almost all invasive pathogens. However, few pathogens have devised strategies to circumvent these defence lines and create a safe sanctuary for themselves within the host cells, from where they can periodically disseminate into the host. However, the dynamics of host–pathogen interactions governing development of such pathogenic reservoirs remain poorly understood. SPN, which is generally regarded as an extracellular pathogen, has been shown in recent years to occasionally establish intracellular niches within the body to evade immune surveillance and disseminate within the host (Ercoli et al., 2018; Subramanian et al., 2019a). Particularly pertinent are observations of its intracellular replication inside splenic macrophages and cardiomyocytes, as well as prolonged persistence in the lower respiratory tract (Ercoli et al., 2018), (Badgujar et al., 2020). Central to its intracellular life is the expression of the pore-forming toxin Ply, which has been shown to form variable-sized pores on biological membranes. Although the differential pore-forming ability of Ply is directly dependent on the monomer concentration of the toxin in in vitro conditions (El-Rachkidy et al., 2008; Gilbert and Sonnen, 2016; Sonnen et al., 2014; van Pee et al., 2016), reports depicting a correlation between its expression levels and pore dynamics, fostering the intracellular life of SPN, are not available.

Few other pore-forming proteins, such as holin, are reported to perforate cell membrane only after accumulating to a threshold level (White et al., 2011). Such proteins are usually uniformly distributed in a relatively mobile state in the cell membrane. A sudden transition from this uniformly distributed state to accumulation at specific sites occurs during membrane damage, which eventually results in formation of holes in the membrane. Therefore, to calculate the membrane perforation rate of a protein toxin, it is necessary to keep track of the timescale of such a transition (Dennehy and Wang, 2011). On similar lines, we hypothesized that various events observed by us, ranging from vacuolar degradation to cytosolic escape and subsequent cytosolic degradation, occurred due to Ply accumulation inside PCVs reaching a threshold value.

Compared to many earlier studies on pathogen invasion of host cells and their subsequent survival against the host defence system, which typically have been qualitative, here, we quantified not only the overall survival times (separately for distinct subpopulations of SPN), but also provided a separate breakup of these times into sub-events such as pore formation time (Gal8 attachment), vacuolar death or cytosolic escape time, and total degradation time. This level of detailed breakup helps in further understanding the parts of the process that are or are not dependent on toxin accumulation. Both high- and moderate-Ply-producing subpopulations were marked with Gal8 and the corresponding timescales for Gal8 recruitment were ∼137 and 80 min, respectively. But post Gal8 attachment, although the vacuolar subset stayed alive for an additional ∼104 min, the high-Ply-producing variants escaped to the cytosol in a relatively shorter time (∼70 min). On the contrary, the very low-Ply-producing variants showed extremely prolonged persistence (>400 min). Our main aim in this study was to correlate the shorter versus longer lifespans of distinct SPN subpopulations with the corresponding amount of Ply accumulation inside the vacuoles.

Mathematical models for the kinetics of protein production within the gene expression literature have been available for the past two decades. Yet, only very recently (Rijal et. al., 2020, 2022), the full distributions of timescales associated with ‘threshold crossing’ of protein quantity have been mathematically determined – these were applied to data on cell lysis timings caused by bacteriophage infection. Here, we applied this contemporary theoretical result to threshold crossing of the toxin (Ply) in the intravacuolar environment, leading to either vacuolar death or catastrophic membrane damage. Through the modelling, although indirectly, we can quantitatively estimate toxin threshold quantities and rates of toxin production of different bacterial subpopulations and compare them, without any direct experimental probing of Ply. Our findings suggested that Ply accumulation was proportional to the protein production rate (k_c and k_v for cytosolic and vacuolar life stages, respectively). Critically, we found that k_v was 2.6 times slower than k_c. At the same time, the Ply threshold numbers associated with the cytosol-escaped population was ∼2-fold higher than that in the case of the SPN subset being degraded in the vacuole. Jointly, these justify our claim that the vacuolar and cytosolic subpopulations are indeed moderate- and high-Ply-producing populations. For the cytosolic population, the high rate of Ply synthesis quickly led to catastrophic damage of Gal8-marked endosomes, triggering escape of SPN to the cytosol even before the host could sequester it in autophagosomes. However, in the case of the vacuolar subset, owing to slower production rate of Ply, the time taken for threshold Ply synthesis could be sufficiently long, providing the host a window of opportunity to sequester the Gal8-marked damaged endosomes in autophagosomes. We would like to highlight that these quantitative estimates of intravacuolar toxin accumulation triggering engagement with diverse host intracellular defence mechanisms promoting distinct phenotypic outcomes is a novel finding.

Our results also suggested that Gal3 can detect smaller pores and is recruited to the damaged endosomes first, compared to Gal8. Unlike other galectins, not only is Gal3 widely distributed in different tissues (Gal8 is also widely distributed), but it is also the only galectin that has a long unique N-terminal domain attached to a single carbohydrate recognition domain (Johannes et al., 2018). Gal3 can exist as monomer or can associate via the non-lectin domain into multivalent complexes up to a pentameric form (Liu and Rabinovich, 2010). It is speculated that this allows Gal3 to bridge different ligands effectively and form adhesive networks. Although the diverse role of Gal3 in mediating endomembrane repair or targeting the damaged endosomes towards autophagy is already known (Jia et al., 2020a), how Gal3 switches between these roles remains elusive. On the one hand, Gal3 is known to interact with the ESCRT complex, including Alix, to augment endomembrane repair; on the other hand, it interacts with TRIM16, an autophagy receptor, for efficient sequestration of damaged endosomes or lysosomes (Chauhan et al., 2016). We also similarly observed that subsets of the low-Ply-producing SPN population, beyond Gal3 decoration, either associated with the autophagy marker LC3 or remained free of LC3 (repair pathway). However, irrespective of the routes adopted, both pathways finally resulted in bacterial degradation, possibly following fusion with lysosomes.

On the contrary, the unique SPN subpopulation that exhibited prolonged intracellular persistence remained devoid of Gal3 as well as other damage sensing markers throughout its lifetime but was unusually decorated with LC3. Moreover, knocking down the expression of FIP200 (RB1CC1), a key canonical autophagosome pre-initiator protein, did not perturb LC3 decoration of this SPN subpopulation (Fig. S6A,B). Recently, infection by some pathogens such as Listeria monocytogenes and Aspergillus fumigatus has been reported to induce decoration of pathogen-containing phagosomes by LC3 in a process called LC3-associated phagocytosis (LAP) (Gluschko et al., 2018; Martinez et al., 2015; Sanjuan et al., 2007). These processes have been reported to result in enhanced degradation of the cargo for antigen presentation by MHC-II molecules (Romao et al., 2013) as well as contribute to establishment of an intracellular niche (Cheng et al., 2019; Lam et al., 2013; Mitchell et al., 2018; Prajsnar et al., 2021). In LAP or related non-canonical autophagy pathways, LC3 lipidation is observed to happen on phosphatidylserine (PS) along with phosphatidylethanolamine (PE) (Durgan et al., 2021). These parallel pathways are known to be activated in a vATPase-dependant manner wherein a direct vATPase–ATG16L1 interaction contributes to non-classical LC3 lipidation (Hooper et al., 2022; Xu et al., 2019). However, non-canonical lipidation of LC3 also occurs via ubiquitin ligase activity of TECPR1, which interacts with ATG5, thus allowing the transfer of LC3 to PE and/or PS (Boyle et al., 2023). It is assumed that such conjugation might provide a molecular signature for non-canonical autophagy, enabling its distinction from closely related, parallel pathways. However, the LC3⁺ Gal3⁻ PCVs we observed are not generated by LAP as recruitment of LC3 to these endosomes occurred even after inhibition of PI3K, a prerequisite for induction of LAP (Fig. S6B). A few pathogens possessing pore-forming toxins are also known to trigger a variant of non-canonical autophagy, which is distinct from LAP (Gluschko et al., 2022; Ma et al., 2012). This process, termed pore-forming toxin-induced non-canonical autophagy (PINCA), can also be initiated by the needle-like type three secretion system (T3SS) of Shigella flexneri or Salmonella Typhimurium (Xu et al., 2019). In these cases, ion or osmotic imbalance induced by bacterial toxins or the T3SS is responsible for LC3 lipidation of PCVs. This is typified by the H. pylori toxin VacA, which induces osmotic imbalance to cause unconventional LC3 lipidation on intact phagosomes (Florey et al., 2015). Drawing from this, we speculated that the long-persisting PCVs originated due to extremely low Ply expression, which resulted in very minute damage to the endomembrane. Such disruptions are not capable of recruiting any damage sensors but are adept in inducing pH or ionic imbalance. Indeed, we could trigger decoration of Δply mutant-containing vacuoles with LC3 following treatment with an ionophore, proving the key role of ion or osmotic imbalance induction in marking the perturbed endomembrane with LC3.

However, the LC3⁺ Gal3⁻ PCVs that we observed differ fundamentally from other PCVs that are marked by non-canonical autophagy with respect to persistence time and ultimate fate. For example, in PINCA, LC3-positive phagosomes fused more often with lysosomes, indicating that PINCA promotes phagolysosomal fusion (Herb et al., 2022). It is also speculated that PINCA might represent an attempt of macrophages to repair damaged phagosomal membranes as a last resort against pathogens that are yet to escape from the phagosome (Kreibich et al., 2015). Contrary to this, in our case, tiny pores produced by low Ply expression help to sustain SPN for prolonged periods. Moreover, these unique PCVs (LC3⁺ Gal3⁻) did not fuse with lysosomes as revealed by the absence of the lysosomal enzyme cathepsin B (CTSB) from these structures (Fig. S6C,D). Our observations could be correlated to a situation with L. monocytogenes in which lesser listeriolysin O production gave rise to a unique subpopulation that resided inside the phagosomes and persisted for a prolonged period (Birmingham et al., 2008). These spacious Listeria-containing phagosomes serve as a replicative niche for Listeria by maintaining a neutral intraluminal pH. This highlights that residing inside a vacuole by lowering metabolic activity can be a pathogenic counterstrategy to evade host defences. Our findings also point towards a similar situation in which an SPN subpopulation with low toxin expression ensures longer survival inside host cells. Prolonged persistence associated with neutralization of bacteria-bearing lysosomes is reported to trigger lysosome exocytosis, resulting in expulsion of exosome-encased bacteria (Miao et al., 2015). Although such a process is touted to be a cell-autonomous defence program to clear recalcitrant pathogens, this could potentially be exploited by pathogens for egressing out of the hostile intracellular environment. Whether pneumococci follow suit for egressing out of lung epithelia, facilitating onward transmission, remains to be explored.

SPN strain R6 (serotype 2, gift from Prof. Tim J. Mitchell, University of Birmingham, UK, retired) and its derivatives were routinely grown in Todd-Hewitt broth (HiMedia) supplemented with 1.5% yeast extract at 37°C in 5% CO₂. When necessary, the following antibiotics were used: spectinomycin (100 μg/ml; HiMedia) and chloramphenicol (4.5 μg/ml; Sigma). The generation of engineered pneumococcal strains, such as the Δply mutant and low (SPN:Ply-Low) and high (SPN:Ply-High) Ply-expressing strains, have been described earlier (Surve et al., 2018). SPN strains were made fluorescent either by integration of hlpA-tRFP fusion cassette (kind gift from Prof. Jan-Willem Veening, University of Groningen, Netherlands) into the genome of SPN or by staining with DRAQ5 (BD Biosciences) and Hoechst (HiMedia). All gene deletions and cassette insertions were verified by PCR amplification of the gene locus, followed by DNA sequencing.

A gfp-ply translationally fused SPN strain was created. To create this, flanking regions of the ply gene (ply-up, 556 bp upstream, and ply-down, 529 bp downstream), following amplification with the Ply-up-KpnI-F2/Ply-up-XhoI-R2 and Ply-down-BamHI-F2/Ply-down-XbaI-R primer pairs (Table S1), were sequentially cloned using KpnI/XhoI and BamHI/XbaI enzymes in pBSK(+) (kind gift from Prof. Indranil Biswas, University of Kansas Medical Centre, USA). Then gfp gene was fused upstream of the ply gene using overlap extension PCR (GFP-Xho1-F/Linker-GFP-R/Linker-Ply-F/Ply-BamH1-R primers) and the gfp-ply construct was cloned in the previously created recombinant plasmid (between the flanking sequences of ply gene) using BamHI/XhoI. The amplified chloramphenicol resistance cassette was then cloned in the BamHI site. The resultant plasmid was used for transformation of the Δply mutant (Δply mutant strain possessing DNA sequence flanking ply gene and a spectinomycin resistance cassette in between) using CSP-1 (Genpro Biotech), and the recombinant colonies were selected on brain heart infusion (BHI) agar plates containing chloramphenicol. Double crossover recombination was confirmed by PCR amplification of the ply region with the Ply-up-flnk-F/Ply-down-flnk-R primers. GFP expression in recombinant colonies was also visualized by fluorescence microscopy and GFP–Ply translational fusion was confirmed by immunoblotting with anti-Ply (1:1000, Santa Cruz Biotechnology, sc-80500) and anti-GFP (1:2000, Invitrogen, A11122) antibodies.

The haemolysis assay was performed as previously described (Surve et al., 2018). Briefly, mid exponentially grown SPN cultures were pelleted and lysed by sonication. Cell lysates were centrifuged at 21,000 g at 4°C for 30 min to collect cell-free extracts. Dilutions of the cell-free extracts were incubated with a 2% RBC suspension for 60 min at 37°C. Absorbance of the released haemoglobin was determined at 450 nm and 540 nm using a microplate reader. PBS- and Triton X-100-treated RBC suspensions were used as negative and positive controls, respectively.

The heterogeneity in Ply expression was assessed by flow cytometry. Mid-exponentially grown WT SPN, SPN:Ply-High and SPN:Ply-Low strains were fixed with 2% paraformaldehyde and permeabilized using 0.1% Triton X-100. Fixed bacteria were incubated overnight with the anti-Ply antibody (1:1000, Santa Cruz Biotechnology, sc-80500) at 4°C followed by anti-mouse-IgG secondary antibody conjugated to Alexa Fluor 555 (1:2000, Invitrogen, A31570) for 1 h. Cells were then thoroughly washed with PBS and analysed by a flow cytometer (BD Biosciences). For analysis, FlowJo (Version 10) was used.

Human lung alveolar adenocarcinoma (type II pneumocyte) cells or A549 (American Type Culture Collection, CCL-185) were routinely cultured in Dulbecco's modified Eagle medium (DMEM; HiMedia) supplemented with 10% fetal bovine serum (FBS, Gibco) at 37°C and 5% CO₂. Stably transfected A549 cells overexpressing the YFP–Gal8, GFP–LC3, mStrawberry–Gal3, mCherry–lyseninW20A fusion proteins were generated following transfection with M6P-Blast-YFP-LGALS8 (kindly provided by Prof. Felix Randow, MRC Laboratory of Molecular Biology, UK), pMRX-IRE-Blast-GFP-LC3 (sub-cloned from pBABE-Puro-GFP-LC3 vector, Addgene #22405), pMRX-IRE-Puro-mStrawberry–Gal3 (kindly provided by Prof. ‪Tamotsu Yoshimori, Osaka University, Japan) and pMRX-IRE-Blast-mCherry-LyseninW20A (sub-cloned from M6P-blast-mcherry-lyseninW20A, kindly provided by Prof. Felix Randow). All transfections were performed using Lipofectamine 3000 (Invitrogen) as per the manufacturer's instructions. Following transfection, cells were selected with 2 μg/ml of puromycin (Sigma-Aldrich) and 2 μg/ml of blasticidin hydrochloride (HiMedia). For treatment studies, cells were incubated in 50 μM of monensin sodium salt (Sigma-Aldrich) post infection.

A549 cells were treated with 1 mM of the PI3K inhibitor 3-Methyladenine (3-MA, Sigma) for 24 h, 150 nM of the v-ATPase inhibitor Bafilomycin A1 for 2 h (BafA1, Sigma), 0.1 mM LLOMe (Sigma) or 50 mM of Monensin (Sigma) for further experiments. Monensin and LLOMe were present in the cell culture medium throughout the course of experiments.

A549 cells were transfected with 100 pmol siFIP200 (Qiagen, SI02664578) using Lipofectamine 3000 as per the manufacturer’s protocol. At 6 h after siRNA addition, cells were replaced with fresh culture medium. For negative control, scrambled siRNA (Invitrogen, 4457289) was used.

The following antibodies were used in this study: anti-Gal3 (1:500, R&D systems, MAB11541), anti-Gal8 (1:1000, R&D systems, AF1305), anti-cathepsin B (1:500 dilution, Abcam, ab58802), anti-goat IgG Alexa Fluor 633 (1:1500, Invitrogen, A21082) and anti-mouse IgG Alexa Fluor 555 (1:2000, Invitrogen, A31570). The fine chemicals used were: DiO (3,3′-dioctadecyloxacarbocyanine, Invitrogen), monensin sodium salt (Sigma), DRAQ5 (BD Biosciences), FITC (Sigma-Aldrich), bafilomycin A1 (Sigma) and Flipper-TR (Cytoskeleton, Inc.).

To find out the working concentration of monensin, A549 cells were incubated with varying concentrations of monensin from 50 μM to 300 μM. At 9 h post incubation, cell viability was checked using MTT assay kit (HiMedia). Cells incubated with ethanol (vehicle) or 0.1% Triton X-100 were considered as negative and positive controls, respectively. Similarly, the MTT assay was performed to assess cell viability post infection with SPN:Ply-High and SPN:Ply-Low strains for 12 h.

Mid-log-phase-grown SPN strains were re-suspended in PBS to adjust the optical density at 600 nm (OD_600nm) to 0.4, comprising 10⁸ CFU/ml. These were used to infect A549 cells at a multiplicity of infection (MOI) of 25 and incubated for 1 h, followed by treatment with penicillin (10 µg/ml) and gentamicin (400 µg/ml) to kill extracellular SPN. At the indicated time points cells were thoroughly washed with DMEM, lysed with PBS containing 0.025% Triton X-100 and the lysates were plated on BHI agar plates to enumerate the surviving bacteria. The percentage survival at the indicated time points was calculated relative to that at 0 h and the survival ability at the indicated time points was represented as the fold change in percentage survival relative to that of control.

For immunofluorescence, A549 or its transfected derivatives grown on glass coverslips were infected with SPN at a MOI of 25 for 1 h, followed by treatment with penicillin (2 μg/ml) and gentamicin (400 μg/ml) for 2 h to kill the extracellular bacteria. At the desired time points, the infected cells were washed several times with DMEM, fixed with 4% paraformaldehyde for 15 min, permeabilized with 0.1% Triton X-100 for 10 min and blocked with 3% bovine serum albumin (BSA) for 2 h. The cells were incubated overnight in the appropriate primary antibodies prepared in 1% BSA in PBS at 4°C. The next day, the cells were washed with PBS and incubated in suitable secondary antibodies for 1 h. Finally, cells were incubated in Hoechst (10 μg/ml) for 15 min and mounted over a clean glass slide using VectaShield without DAPI (Vector Laboratories) for visualization using a laser scanning confocal microscope (LSM 780, Carl Zeiss) under 40× or 63× oil objectives. Images were acquired after optical sectioning and processed using Zen Lite software (version 5.0).

Live fluorescence imaging was performed using a spinning-disk confocal microscope (Zeiss Axio-Observer Z1, Carl Zeiss) or the laser scanning confocal microscope (LSM780, Carl Zeiss). Monolayers of transfected cells grown in 35 mm glass-bottomed Petri dishes were infected with DRAQ5 (3 μg/ml)-treated SPN as mentioned above. Time-lapse imaging was performed at multiple positions along with optical sectioning under the 63× oil immersion objective and the images acquired were processed and analysed using Zen Lite Software 3.1.

For generation of a pH calibration graph, cell monolayers grown in 35 mm glass-bottomed dishes were incubated in FITC–dextran (500 kDa, 200 μg/ml, Sigma) solubilized in buffers of various pH for 1 h at 37°C and 5% CO₂. Post incubation, the cells were extensively washed, followed by sequential incubation in different isotonic K⁺ solutions (140 mM KCl, 1 mM MgCl₂, 0.2 mM EGTA, 20 mM NaCl, and 20 mM of HEPES, MES or Tris based on the pH) buffered to pH values from 4.5 to 7.5 containing no or 5 μM monensin. After 5 min, ratiometric imaging of cells was performed with the laser scanning confocal microscope using the 63× oil immersion objective. Light was transmitted alternately through 490 nm and 440 nm excitation filters, and the emitted light was passed through a 520 nm emission filter and captured using a GaSP detector. The resulting fluorescence intensity ratio (490/440) was extrapolated to its corresponding pH values. For measurement of the pH of SPN-containing endosomes, cells were infected with SPN in FITC–dextran (500 kDa, 200 μg/ml) and rhodamine–dextran (60 kDa, 200 μg/ml)-laden media, and ratiometric imaging of FITC present inside bacteria-containing endosomes was performed as described. Rhodamine fluorescence, which is unaltered due to change in pH, was used to track PCVs throughout imaging.

Loss of integrity of the endomembrane was assessed by leakage of FITC–dextrans of various molecular masses: 3 kDa (Sigma, FD4) and 70 kDa (Sigma, 46945). A549 cells expressing mCherry–Gal8 and mStrawberry–Gal3 were infected with SPN and FITC–dextran solution (10 mg/ml) was then immediately added to the medium. At 30 min post infection, the extracellular bacteria and residual FITC–dextran were washed off by DMEM, replenished with fresh assay medium containing penicillin (10 µg/ml) and gentamicin (400 µg/ml), and were imaged under the confocal microscope using the 63× oil objective with incubation at 37°C and 5% CO₂.

Ply was purified as described earlier (Badgujar et al., 2020). Briefly, the gene encoding Ply was amplified from the genomic DNA of SPN strain D39 and cloned into the NdeI and XhoI sites of the pET28a vector (Novagen). The plasmid was verified by DNA sequencing and transformed into E. coli BL21 (DE3) cells for protein expression. Freshly transformed colonies were grown in Luria Bertani (LB) broth containing 50 μg/ml kanamycin at 37°C on a shaker incubator for 12 h. 1% of the primary culture was added to 1 l of LB broth and incubated at 37°C on a shaker incubator until the OD_600nm reached 0.6–0.8. Protein expression was induced by the addition of 400 μM isopropyl-1-thiogalactopyranoside (IPTG) and the culture was further grown at 22°C for 5–6 h with agitation at 150 rpm. The cells were harvested by centrifugation at 3000 g for 10 min at 4°C. The cell pellet was resuspended in buffer A (25 mM Tris, pH 8.0, and 300 mM NaCl) and lysed by sonication. Following separation of cell debris by centrifugation (21,000 g, 50 min, 4°C), the supernatant was applied to a 5 ml His-Trap column, equilibrated with buffer A. The column was washed with ten column volumes of buffer A and Ply was eluted with buffer A containing 250 mM imidazole. The eluted fractions were pulled, concentrated with Amicon ultrafiltration units (10 kDa) and buffer exchanged in 200 mM HEPES, 1 M KCl and 100 mM MgCl₂, pH 7.4, using PD-10 columns (GE Healthcare). The purity of Ply was analysed using SDS-PAGE.

To check the oligomerization status of Ply, we incubated purified Ply (20 µg/ml) at different pH levels (ranging from pH 4 to 7), at 37°C for 1 h. Following incubation, the Ply solution was further examined in a 1.5% horizontal agarose gel with native gel buffer (25 mM Tris-HCl, pH 8.5, and 19.2 mM glycine), and resolved at 80 V for 1 h. The gel was further stained with Coomassie Brilliant Blue solution for 30 min.

Liposomes loaded with fluorescently labelled dextrans of different molecular mass were prepared by the thin-film hydration technique (Omarova et al., 2021). 12 mg 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC; Sigma, 850855C) and 6 mg cholesterol (Sigma, C8667) were dissolved in 3 ml chloroform and 1.5 ml methanol mixture (2:1; v/v) in a round-bottomed flask. The solvent was then evaporated on a rotary evaporator at 40°C temperature at 100 mbar for 4 h to form a thin lipid film. The film was further maintained in a vacuum at 10 mbar for 30 min to remove residual solvent. The formed thin lipid film was then hydrated with 1 mg/ml of 60 kDa rhodamine–dextran and 3 kDa FITC–dextran solution in PBS at 50°C for 45 min. This yielded a suspension of large lipid vesicles. The vesicle suspension was treated with eight consecutive cold-heat sonication cycles to enrich uniform liposomes with an average diameter of 100 nm. These dual-size-dextran-loaded liposomes were incubated with different concentrations of purified Ply (0.001 to 1 μM) at 37°C for 1 h. Following incubation, the supernatant was collected by centrifugation at 20,000 g and the release of FITC–dextran or rhodamine–dextran was assessed by fluorometric scanning (488 nm for FITC, 555 nm for rhodamine) to reveal the status of pore size.

From live-cell images, we quantified the times representing the three kinds of events observed in the degradation pathways of different subpopulations of WT SPN. For the subpopulation that degrades in the cytosol, the relevant event timescales are as follows: (1) Gal8 attachment times (t_g8), which are seen for both cytosolic and vacuolar degradation pathways; (2) time for the formation of large pores on the Gal8-marked PCVs for cytosolic escape (t_p) (applicable for the cytosolic degradation pathway); (3) the lifetime of SPN within the autophagosome after Gal8 recruitment (t_v) (in the case of the vacuolar pathway).

The Gal8 attachment depends on sensing a minimal rupture in the membrane of PCVs caused by Ply. As the attachment process is expected to be a single-step binding, the statistics of the timescales t_g8 are given by P(t_g8), which follows an exponential distribution. In contrast, the cytosolic escape time t_p is expected to depend on Ply accumulation on the inner membrane of autophagosome and its number crossing a certain threshold. Hence, t_p is distributed according to Eqn 1. Likewise, for vacuolar degradation, the vacuolar degradation time t_v, after Gal8 recruitment, is also expected to be distributed according to Eqn 1.

For the quantitative analysis of data, normalized histograms of the above time points were plotted using MATLAB software. We fitted these histograms with normalized exponential distributions ∼λ^(−λtg8). The inverse of the means of t_g8 (80 min and 137 min for the vacuolar and cytosolic cases, respectively) were used as the decay constants λ for the fitting.

For cytosolic degradation, we plotted the histograms for total cytosolic escape times as t_C′=t_g8+t_p and t_p. Note that for every cytosolic escape event, t_C′>t_g8, thus t_p=t_C′−t_g8>0. We fit the histograms of t_p and t_C′ using Eqn 1. We first fixed a burst size b. We then equated t_mean in Eqn 2 with the experimental mean t_p and varied N−N₀ as a free parameter (integer), the rate k gets fixed from Eqn 2. For our choice of b=2, and certain N−N₀ and k values as mentioned in the text, we found that the error between the theoretical fit in Eqn 1 and the experimental histogram in terms of the sum of squared residuals at mid timepoints of each bin in the histograms (Archdeacon, 1994) was the least. We evaluated the σ from Eqn 3 for the best-fitted values of b, k and N−N₀. For the vacuolar degradation pathway, we plotted histograms of total degradation times, t_Vac=t_g8+t_v and t_v. Again, note that for every vacuolar degradation event, t_Vac>t_g8, thus t_v= t_Vac−t_g8>0. To best fit the histogram of t_v, a fitting method similar to that for t_p was used, as described above. To verify the goodness of the fit, the standard deviation σ from the theoretical fit and experimental histogram were compared.

GraphPad Prism version 5 was used for statistical analysis. Statistical tests undertaken for individual experiments are mentioned in the respective figure legends. P<0.05 was considered to be statistically significant. Data were tested for normality and to define the variance of each group tested. All multi-parameter analyses included corrections for multiple comparisons and data are presented as mean±standard deviation (s.d.) unless otherwise stated.

We acknowledge the biosafety level 2 facility and confocal microscopy facility at the Indian Institute of Technology Bombay.

All relevant data can be found within the article and its supplementary information.

This article has an associated First Person interview with the first author of the paper.

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.260891.reviewer-comments.pdf