Spatiotemporal proteomics uncovers cathepsin-dependent macrophage cell death during Salmonella infection

Summary The interplay between host and pathogen relies heavily on rapid protein synthesis and accurate protein targeting to ensure pathogen destruction. To gain insight into this dynamic interface, we combined click-chemistry with pulsed stable isotope labeling of amino acids in cell culture (pSILAC-AHA) to quantify the host proteome response during macrophage infection with the intracellular bacterial pathogen, Salmonella enterica Typhimurium (STm). We monitored newly synthesised proteins across different host cell compartments and infection stages. Within this rich resource, we detected aberrant trafficking of lysosomal proteases to the extracellular space and the nucleus. We verified active cathepsins re-traffic to the nucleus and are linked to cell death. Pharmacological cathepsin inhibition and nuclear-targeting of a cellular cathepsin inhibitor (Stefin B) suppressed STm-induced cell death. We demonstrate that cathepsin activity is required for pyroptotic cell death via the non-canonical inflammasome, and that LPS transfection into the host cytoplasm is sufficient to trigger active cathepsin accumulation in the host nucleus and cathepsin-dependent cell death. Finally, cathepsin inhibition reduced Gasdermin D expression, thus revealing an unexpected role for cathepsin activity in non-canonical inflammasome regulation. Overall, our study illustrates how resolving host proteome dynamics during infection can drive the discovery of biological mechanisms at the host-microbe interface.


Supplementary Discussion
Dynamic proteome mapping unravels diverse host responses during STm infection.
We detected a strong response to bacteria at all time points in the lysatome, as well as cytokine stimulus (GO:0071345), interferon (GO:0035456), an extrinsic apoptotic pathway (GO:0097191) and a progressive decrease in cellular responses to lipopolysaccharide (LPS; GO:0071222). The latter behavior is consistent with macrophage adaptation to TLR activation by a highly immunogenic stimulus, such as LPS, in order to dampen detrimental impacts from a hyperactive pro-inflammatory response 1 . A strong and constant response to interferon-γ (GO: 0071346) and virus (GO:0009615) were also visible, with the response at 20 hours consisting of 33 proteins (Supplementary Table 2). This is in agreement with the Type I Interferon response being induced by bacterial pathogens [2][3][4][5] .
We also detected an early induction of the I-kappaB kinase/NF-kappaB signaling response, which was sustained throughout infection in the lysatome fraction (Fig.   1b). For instance, consistent with an early NF-κB response that relies on NF-κB1 translocation to the nucleus, we observed strong enrichment of NF-κB1 in the nucleome at 4 hours post infection (+2.12 Log 2 Fold change, adjusted p = 6.72x10 -5 ) (Supplementary Table 3). Furthermore, NF-κB1 expression peaked at 8 hpi in the lysatome, followed by a relative decrease at 20 hpi. This suggested an overall blunting of the NF-κB response over time by reducing nuclear NF-κB1 translocation and abundance within the lysatome. In addition, Nfkbiz (NF-κB inhibitor zeta), progressively increased in the nucleome with time (+1.7 Log 2 Fold change at 4 hours, adjusted p = 1.25x10 -4 ; +4.5 Log 2 fold change at 20 hpi, adjusted p = 3.8x10 -2 ). In addition to its role as an inhibitor, Nfkbiz can promote transcription of the siderophore LCN2 6 . Consistently, LCN2 expression in the lysatome followed a similar pattern in infected cells (Supplementary Table 3). These findings highlight the spatiotemporal dynamics of NF-κB signaling during STm infection, whereby early nuclear NF-κB translocation is followed by a second wave of regulation via Nfkbiz.
Moreover, several broader responses occurred at specific time-points, including an early reduction in the abundance of DNA binding proteins Uhrf1, Mki67, Tmpo and Zfp3 in the lysatome (GO:0003677; 4h) ( Fig. 1b). At later time-points, a decrease in proteins involved in the DNA replication fork (GO:0005657; 8h) preceded a decrease in the nucleic acid metabolic process (GO:0090304; 20h), suggesting multi-level remodeling of DNA replication, firstly by suppressing the DNA replication fork machinery and then the production of DNA building blocks to control the host cell cycle. This may indicate a host response aimed at suppressing pathogen proliferation by limiting vital nucleotide building blocks essential for pathogen survival [7][8][9] .

STm infection induces distinct host proteome responses compared to LPS
A number of infection-related GO terms, including response to virus (GO:0009615) and defense response (GO:0006952) were significantly enriched at both pre-SPI-2 dependent and active SPI-2 dependent proliferation ( Fig. 2a- Table 5). The antiviral response includes proteins (Stat1, Oas2, Tlr7, Tllr13, Lgals8, Lgals9, Samhd1 and Eif2ak2), which may provide insights into additional molecular players involved in this seemingly universal response towards intracellular bacterial pathogens 3 . The remaining defense response, even after normalizing out the LPS response, points to the production of these proteins being significantly higher during  Table 4). STm may suppress this part of the innate immune defense to avoid cell lysis 10,11 . We also detected an even larger number of core cellular processes that were specifically activated or repressed during STm-infection. At 8 hours, there was a significant upregulation of the apoptotic process (GO:0006915;  Table 5). This response may hint to an interesting host-microbe antagonism revolving around purines, as their synthesis appears to be a strong limiting factor for STm infection 14,15 16 . The suppression of host cell pathways important for DNA synthesis and replication may signify a host-driven response to counter STm G2/M arrest.
Together, these findings demonstrate modulation of core cellular processes spanning from DNA replication to cell death specifically during STm infection however, the functional significance of many of these proteins will require further study.
Similar to the lysatome, STm infection triggered a number of distinct responses from LPS in the secretome. It is now well appreciated that the SCV contains reduced levels of vacuolar hydrolases, including cathepsins, likely due to fusion of the SCV with lysosomes whose hydrolytic activity has been deactivated by the STm effector

Previous reports have indicated Osteopontin is detrimental to the host during
Pneumococci or viral infection 18,19 , and its intracellular form induces the antiviral response by stabilising TRAF3 20 . Our findings indicate Osteopontin is specifically regulated during STm infection, although its functional relevance in this context remains to be elucidated.

infection.
In addition to cathepsin secretion, we noticed a pronounced enrichment for lysosomal proteases in the nucleus during STm infection. Although cathepsins and lysosomal hydrolases have previously been shown to be targeted to the nucleus during cellular stress, in cancer cells, or different phases of the cell cycle [21][22][23][24][25][26][27][28] , to our knowledge this is the first report of active cathepsin delivery to the nucleus during infection. Despite the considerable overlap in the lysosomal proteases enriched in both the secretome and the nucleome during infection, CtsB and CtsS displayed pronounced enrichment only in the nucleus. This distinct cathepsin enrichment profile between these two cellular compartments suggests that STm-induced changes to subcellular cathepsin targeting are not identical for nuclear and extracellular targeted proteins. This is consistent with the finding that the two events have different triggers (while extracellular re-trafficking is Salmonella-triggered by SifA 17 , nuclear retrafficking seems to be host-triggered, as a response to cytosolic LPS) and outcomes (non-active cathepsins extracellularly versus active nuclear cathepsins).
Analogous to the previously reported SPI-2 dependent cathepsin secretion 17 , cathepsin delivery to the nucleus was enhanced by the STm SPI-2 secretion system. However, as cathepsin dependent cell death and nuclear cathepsin activity could be triggered by LPS transfection alone, we interpret the apparent SPI-2 dependent effect to be (at least partially) a consequence of higher LPS doses presented in the cytoplasm from wild-type proliferating STm compared to the non-proliferating SPI-2 mutant or dead STm. At this point, it remains unknown whether the increased ER pool of non-targeted cathepsins, generated by the SPI-2-mediated short-cutting of cathepsin trafficking from the ER to the vacuole 17 , contributes also to the increased nuclear re-trafficking we observe for wild-type cells.
Nuclear cathepsins were of higher molecular weight compared to lysosomal forms found in the Tx-100 soluble fraction, akin to previously reported SPI-2 dependent CtsD secretion 17 . However, in this instance, high molecular weight nuclear cathepsins are active. Nuclear cathepsins often exist as higher molecular weight proforms [21][22][23]25 and have also been reported to be active in thyroid carcinoma cells and during the S-phase of the cell cycle 21,29 . This suggests a non-canonical form of protein trafficking delivers cathepsins to the nucleus during STm infection, possibly as a result of alternate splicing events leading to signal-peptide-devoid N-termini as previously observed for CtsL and CtsB 29,30 . Presumably, this is then followed by a subsequent pH-independent maturation and/or cathepsin activation (currently, cathepsin S is the only known example that remains active at neutral pH 31 ). We can exclude that the nuclear cathepsins are simply proteins derived from ruptured vacuoles, as these are newly synthesised proteins and have not undergone maturation in the low pH of a lysosome as judged by their high molecular weight.
General breakdown of cellular integrity of dying cells can also be excluded as an explanation for nuclear cathepsin localization for several other reasons: nuclear relocalization is specific to cathepsins and a few other cytosolic proteins (compared to >700 proteins we can detect in both the lysatome and nucleome); occurs early at 8 hpi, even before strong signs of cell death; trafficking to the nucleus occurs independently of cathepsin activity, which itself is related to cell death (Extended Data Fig. 3a, c). Taken together, nuclear cathepsins represent a branch of cathepsin trafficking distinct from those destined for the endosomal/lysosomal compartment or those being re-trafficked to the extracellular milieu. However, the precise mechanism for nuclear cathepsin targeting and activation during infection will require further investigation.