|Home | About | Journals | Submit | Contact Us | Français|
The final stages in mammalian cytokinesis are poorly understood. Previously, we reported that the ADP-ribosyltransferase activity of Pseudomonas aeruginosa type III secreted toxin ExoT inhibits late stages of cytokinesis. Given that Crk adaptor proteins are the major substrates of ExoT ADP-ribosyltransferase activity, we tested the involvement of Crk in cytokinesis. We report that the focal adhesion-associated proteins, Crk and paxillin are essential for completion of cytokinesis. When their function is absent, the cytoplasmic bridge fails to resolve and the daughter cells fuse to form a binucleated cell. During cytokinesis, Crk is required for syntaxin-2 recruitment to the midbody, while paxillin is required for both Crk and syntaxin-2 localization to this compartment. Our data demonstrate that the subcellular localization and the activity of RhoA and citron K, which are essential for early stages of cytokinesis, are not dependent on paxillin, Crk or syntaxin-2. These studies reveal a novel role for Crk and paxillin in cytokinesis and suggest that focal adhesion complex, as a unit, may partake in this fundamental cellular process.
Cytokinesis is the final stage of cell division in which the daughter cells dissociate. Animal cells use an actin contractile ring that is attached to the plasma membrane and powered by myosin to create a cleavage furrow that partitions the dividing cell into two daughter cells (reviewed in ref. 1). Contractile ring assembly and its constriction are directed by RhoA, which recruits and activates various proteins including mDia, a formin homology domain containing protein, and citron kinase. After furrow formation, the two daughter cells remain connected through a cytoplasmic bridge containing a microtubule-based central spindle and a multitude of proteins that together form the midbody.1 Genetic and RNAi-based studies in simple model organisms, such as yeast, Dictyostelium, Drosophila and C. elegans, as well as the proteomic studies of midbodies isolated from mammalian cells, have led to the identification of many components of the midbody.2-5 The proteins required for the completion of cytokinesis exhibit diverse activities, including cytoskeleton remodeling, vesicular trafficking and membrane fusion.
Abscission, the final step in cytokinesis, is dependent on the endocytic machinery, the deposition of new membrane at the ingressing furrow, and ubiquitin-conjugating activity.6-8 Indeed, several midbody-associated proteins are involved in membrane trafficking, including components of the exocyst as well as the SNARE family proteins syntaxin-2 and endobrevin.7,9 Despite these recent advances in the field of cytokinesis, much remains to be learned about the organization and the hierarchical order of assembly of the molecular components of the midbody, particularly during late stages of cytokinesis.
We recently reported that ExoT, a type III secreted toxin produced by the Gram-negative opportunistic pathogen Pseudomonas aeruginosa, inhibits late stages of cytokinesis by blocking syntaxin-2 localization to the midbody.10 This phenotype was dependent on the ADP-ribosyltransferase activity (ADPRT) encoded in the C-terminal domain of ExoT. The major targets of the ADPRT activity are the CrkI and CrkII adapter proteins.11 These SH2 and SH3- containing proteins are widely expressed and are involved in various cellular processes including focal adhesion assembly, epithelial to mesenchymal transition, cell migration and phagocytosis.12,13
In this communication we explore the role of Crk in cytokinesis and report that Crk and paxillin, components of focal adhesion, play essential roles in cytokinesis. We show that syntaxin-2 midbody localization is dependent on Crk, and that paxillin mediates both Crk and syntaxin-2 midbody localization during cytokinesis. Our results reveal previously unappreciated functions for these focal adhesion-associated proteins and yield new insights into mammalian cytokinesis.
ADP ribosylation of Crk by ExoT has been shown to interfere with its activity in cell migration, specifically by interfering with Crk interaction with paxillin.14 Similarly, expression of the DN CrkI mutants have been shown to prevent Crk interaction with paxillin and to interfere with the known cellular functions of both CrkI and CrkII, including cell migration and epithelial to mesenchymal transition.15,16 We thus examined the effect of Crk DN mutants on cytokinesis. To this end, we transfected C-terminal GFP fusions of wild type CrkI or variants with dominant negative (DN) mutations in the SH2 (CrkI/R38K), or SH3 (CrkI/W170K) domains (Table 1) into HeLa cells and analyzed the effect of these Crk variants on cytokinesis by timelapse videomicroscopy. Multiple movies of GFP-expressing cells undergoing cell division were examined and only cells which had sufficient time to complete cytokinesis were scored. As shown in Figure 1A and Supplemental movies 1 and 2, expression of DN CrkI/R38K-gfp or CrkI/W170K-gfp blocked cytokinesis in 43% (n = 39) or 35% (n = 43) of transfected dividing HeLa cells respectively (p < 0.001). Cytokinesis was inhibited in only 6% (n = 48) of cells transfected with the wild type CrkI, (Suppl. movie 3), a number that was not significantly elevated compared to the GFP vector control 2.7% (n = 73, p ≤ 1). Expression of DN Crk mutants inhibited cytokinesis after cleavage furrow completion, phenocopying the effect of the ADPRT domain of ExoT on cytokinesis.10 These effects were not due to differences in levels of expression of these Crk-GFP variants (data not shown). These results indicate that interfering with Crk function can adversely affect cytokinesis. Whether Crk plays a direct role in cytokinesis or whether its role effecting cytokinesis resulted from inhibition of the cell-matrix interaction remains to be determined.
Previously, we have shown that in HeLa cells, Crk was present in the midbody during cytokinesis.10 To determine whether localization to the midbody was isoform specific, HeLa cells were transiently transfected with CrkI- or CrkII-GFP fusion proteins and localization to the midbody was determined by immunofluorescent (IF) microscopy, using β-tubulin midbody staining as the marker. Both Crk isoforms localized to the midbody during cytokinesis (Suppl. Fig. 1). Interestingly, only CrkI stained focal adhesion like structures at the cell periphery. As would be predicted from proteins that possess both nuclear localization domain and nuclear export domain and consistent with previous reports,17 we also observed transfected Crk in both the nucleus and the cytoplasm (Suppl. Fig. 1).
To ensure that the presence of Crk in the midbody was not restricted to transformed cell lines, we examined the subcellular distribution pattern of endogenous Crk proteins in primary human epithelial keratinocytes (HEK). Because of antibody incompatibility we used citron K, a protein shown to localize to the midbody (reviewed in ref. 18 and Suppl. Fig. 2), as the marker for dividing cells. As shown in Figure 2A, endogenous Crk proteins also co-localized with citron K at the midbody after cleavage furrow completion, indicating that Crk midbody localization was not an artifact of GFP fusion, nor was it restricted to transformed cells.
To definitively show that Crk was necessary for cytokinesis, we transfected HeLa cells with an siRNA that targeted a conserved region present in both Crk isoforms (Materials and Methods). As an essential component of focal adhesions, complete CrkI/II protein depletion would be expected to result in cell detachment from the surface. We carried out a dose-response analysis and found a condition (100 nM Crk siRNA depletion for 48 hrs) under which Crk levels were reduced by 80% to 90% without causing excessive cell detachment and cytotoxicity (Fig. 2B and data not shown). Under these conditions, HeLa cells treated with CrkI/II siRNA exhibited greater than three-fold increase in multinucleated cells compared to control siRNA-treated cells (p < 0.005; Fig. 2C and D). These results reveal a previously unappreciated and indispensable role for Crk adaptor proteins in cytokinesis.
The common theme underlying the Crk-mediated complex assembly and signaling is the recruitment of Crk by tyrosine phosphorylated paxillin.12,13 To examine whether paxillin played a similar role in cytokinesis, we first examined the subcellular distribution of paxillin by IF. Similar to what we observed with Crk, paxillin also co-localized with citron K in the midbody after furrow completion in HeLa (data not shown) and primary HEK cells (Fig. 3A). We then tested whether paxillin in the midbody was tyrosine phosphorylated using an antibody that specifically recognizes paxillin phosphorylated at tyrosine 118. Similar to paxillin, phospho-paxillin was first detectable at the midbody after furrow completion and remained co-localized with citron K during abscission (Fig. 3B and data not shown). To determine whether paxillin was necessary for cytokinesis, paxillin was depleted by siRNA. Paxillin-specific siRNA treatment resulted in ~80% reduction in paxillin levels (Fig. 3C) and also increased the number of multinucleated cells by greater than 3 fold (Figs. 3D and E, compare with with2C,2C, p = 0.0025), indicating that paxillin, like Crk, is essential for cytokinesis.
We next examined whether Crk recruitment to the midbody was dependent on paxillin. HeLa cells were treated with paxillin siRNA, fixed and imaged by IF. To identify cells in which paxillin was efficiently depleted, we restricted our analysis to dividing cells that contained more than one nucleus in their daughter cells, reasoning that previous cytokinesis failure in these cells most likely occurred as a consequence of paxillin depletion. As expected, paxillin staining was significantly reduced in paxillin siRNA-treated cells both in the cytoplasm and at the midbody (Fig. 4D and data not shown). Paxillin depletion also prevented Crk midbody localization in over 60% of the cells (n = 17, p < 0.001) (Fig. 4A and C). Similarly, we examined the dependence of paxillin midbody localization on Crk. Crk siRNA-depleted cells exhibited no defect in paxillin enrichment at the midbody (Fig. 4B and D). Treatment with control siRNA had no effect on either paxillin or Crk midbody localization (Fig. 4C and D and data not shown). Collectively, these results indicate that Crk midbody recruitment requires paxillin and that paxillin localization to the midbody is an upstream event in cytokinesis.
The localization of the v-SNARE endobrevin/VAMP8 and the t-SNARE syntaxin-2 to the midbody occurs after furrow completion and midbody extension and is required to complete abscission.9 Previously, we demonstrated that syntaxin-2 fails to localize to the midbody in the presence of the ADPRT domain of ExoT.10 We examined whether Crk was required for localization of syntaxin-2 to the midbody. As shown in Figure 4E and F, Crk depletion by siRNA prevented syntaxin-2 localization to the midbody in over 88% of these cells (n = 25, p < 0.001), compared to only 20% when treated with control siRNA (n = 17). Since Crk localization to the midbody requires paxillin, we expected that paxillin would also be required for syntaxin-2 midbody recruitment. Indeed, paxillin depletion by siRNA blocked syntaxin-2 midbody localization in nearly 90% of the cells (n = 23, p < 0.001) (Fig. 4E and F). Crk or paxillin depletion had no effect on citron K or RhoA enrichment at the furrow (Fig. 4A and B and data not shown). We conclude that paxillin and Crk are required for syntaxin-2 midbody localization and that paxillin or Crk depletion interferes with cytokinesis at one or more step(s) after RhoA and citron K recruitment and function but prior to syntaxin-2 recruitment to the midbody.
We examined whether Crk or paxillin midbody recruitment and/or their persistence in this compartment may also be mutually dependent on syntaxin-2. HeLa cells were depleted of syntaxin-2 by siRNA (~85% depletion, Fig. 5A), fixed and the midbody localization of syntaxin-2, Crk, and paxillin was determined by IF. As expected, syntaxin-2 staining in the midbody was absent in over 95% of the syntaxin-2 siRNA treated cells (n = 20, p < 0.001) (Fig. 5B and E). Despite failed cytokinesis, as evidenced by multinucleation, depletion of syntaxin-2 did not affect Crk or paxillin recruitment to the midbody (Fig. 5C-G). We conclude that syntaxin-2 functions downstream from Crk and paxillin in cytokinesis.
Src, a known tyrosine kinase for paxillin,15,19-23 has been implicated in cytokinesis. Chemical inhibition of Src has been shown to inhibit separation of daughter cells after midbody extension and during abscission in HeLa cells.24 The presence of phosphopaxillin in the midbody prior to abscission, however, raised the possibility that Src may also be involved in earlier stages of cytokinesis in this compartment. To test this hypothesis, we first investigated Src subcellular distribution by IF. Consistent with a previous report,25 Src was enriched at the plasma membrane at metaphase (Fig. 6A). Upon furrow initiation however, the amount of Src in membranes adjacent to the furrowing site decreased. Src then co-localized with citron K during furrowing and remained in the midbody until abscission.
We then used timelapse videomicroscopy to determine at what step(s), Src activity is required for cytokinesis. In the presence of the Src-family inhibitor SU6656, at a concentration selective for Src family kinases (2 μM),26 nearly 78% of dividing cells failed to complete cell division compared to control-treated cells (n = 33, p < 0.005). Unexpectedly, 18% of these cells did not initiate furrowing (Fig. 6B and C, and data not shown). A larger fraction (60%) initiated furrowing but failed to complete this process during cytokinesis (Fig. 6B and C, and Suppl movie 4). SU6656 also resulted in nearly 4 fold increase in bi-/multinucleated cells (Fig. 6D). Gleevec, a known inhibitor of PDGFR, Abl, c-Kit and Arg tyrosine kinases27 had no effect on cytokinesis (Fig. 6B-D and Suppl movie 5). These results further expand the role of Src in cytokinesis, indicating that its activity is also required during furrowing.
Src activity may be required for assembly of the furrowing components, such as citron K, for its own recruitment, or for the recruitment or activity of targets such as paxillin. IF microscopy indicated that SU6656 did not affect citron K distribution at the site of furrow formation or its localization to the midbody (Suppl Fig. 3A). However, the drug inhibited Src localization to the membrane prior to furrow initiation and reduced its enrichment in the midbody (Suppl Fig. 3A). Since paxillin midbody localization could be detected only after cleavage furrow completion and since SU6656-treated cells failed cytokinesis prior to that time point, we could not directly examine the role of Src in paxillin phosphorylation in the midbody. However, immunoblot analysis using an antibody specific to tyrosine phosphorylated paxillin demonstrated that inhibition of Src activity by SU6656 prevented the vast majority of tyrosine phosphorylation of paxillin (Suppl Fig. 3B), indicating that in HeLa cells, Src is the major kinase for paxillin. Combined, these results suggest that Src is a major contributor of paxillin phosphorylation in HeLa cells and may likely be responsible for paxillin phosphorylation in the midbody.
We have recently shown that the ADP-ribosyltransferase activity of P. aeruginosa ExoT toxin inhibits late stages of cytokinesis.10 As Crk adapter proteins are the primary targets of ExoT,11 we tested the hypothesis that Crk proteins are involved in cytokinesis. The studies reported in this communication confirm our hypothesis and reveal the requirement for other focal adhesion proteins in mammalian cytokinesis.
Despite lack of previous data implicating Crk in cytokinesis, we provide compelling evidence that Crk function is essential to this process. We propose that Crk is required at steps after the completion of the cleavage furrow and is a prerequisite for the terminal steps of cytokinesis. This interpretation is supported by the findings that (i) Crk DN mutants inhibit cytokinesis post-cleavage furrow formation; (ii) Crk is detectable at the midbody after cleavage furrow; and that (iii) siRNA depletion of Crk prevents syntaxin-2 midbody localization, a requirement for the final step of cytokinesis, without affecting RhoA or citron K localization to the midbody, events which occur prior to ingression and are required for earlier steps.
Tyrosine phosphorylation of paxillin by Src tyrosine kinase and the subsequent phospho-paxillin recruitment of Crk are among the common features that underlie the Crk-mediated signaling cascades in cellular processes, such as focal adhesion assembly and cell migration.20,23,28 Our data suggest that similar strategy may be used in cytokinesis. We now show that phospho-paxillin is also present at the midbody and that paxillin also serves an essential role in cytokinesis. Our data demonstrate that paxillin mediates intermediate steps in cytokinesis, post RhoA-mediated cleavage furrowing and prior to Crk midbody localization which occurs after furrow completion. Paxillin siRNA depletion prevents Crk enrichment at the midbody but has no effect on midbody localization of RhoA and citron K. As would be expected from a protein that is required for Crk localization to the midbody, paxillin is also essential for syntaxin-2 midbody recruitment during cytokinesis. The requirement of paxillin for Crk midbody localization is consistent with its role in recruiting Crk in other cellular contexts such as focal adhesion assembly and cell migration.20,23,28 The involvement of paxillin in cytokinesis is further supported by recent reports demonstrating that Pxl1p, a paxillin homolog in the fission yeast Schizosaccharomyces pombe, plays a major role in cytokinesis.29,30 In our studies reported here, depletion of Paxillin or Crk by siRNA or the expression of Crk DN mutants in HeLa cells did not grossly affect the cell matrix interaction yet inhibited cytokinesis. Taken together, these data suggest that the role of paxillin and Crk in cytokinesis is likely independent of their role in focal adhesion assembly and cell migration or that focal adhesion as a unit may partake in this cellular process.
Src is a known tyrosine kinase for paxillin.15,19-23 While it has been demonstrated that Src activity is required to initiate furrowing25 and to dissociate daughter cells during the final step in cytokinesis, abscission,24 our data further expand the role of Src family kinases in cytokinesis. We demonstrate that Src localizes to the midbody early during cytokinesis and that its kinase activity is also required for cleavage furrow completion during cytokinesis. Although our data indicate that Src is a major contributor to paxillin phosphorylation in HeLa cells, more studies are needed to definitively address whether paxillin phosphorylation at the midbody is Src-dependent and critical for recruitment of Crk to this compartment.
Genetic screens in Drosophila and C. elegans, together with proteomic analysis of purified mammalian midbodies, have led to the identification of many proteins that contribute to cytokinesis, including vinculin, talin and enigma, components of focal adhesions, but failed to identify Crk or paxillin.2-5 It should be noted that these methods also failed to identify other essential components required late in cytokinesis, including syntaxin-2, endobrevin, Cep55, centriolin and the exocyst components, highlighting the need for multipronged approaches to dissect the molecular mechanisms of cytokinesis.
Since syntaxin-2 lacks identifiable Crk interacting domains, it is unlikely that Crk recruits syntaxin-2 directly to midbody. Indeed, we have been unable to co-immunoprecipitate syntaxin-2 or its partner endobrevin with Crk or paxillin (unpublished results). Other Crk and/or paxillin interacting proteins may recruit syntaxin-2 to the midbody. Intriguingly, Crk interacts with pleckstrin homology domain (PH)- containing proteins such as DHR and Gab.22 These proteins interact with membrane inositol phospholipids.31,32 Thus, the DHR and/or Gab proteins, in association with Crk, could potentially mediate the recruitment of membrane-associated exocyst and syntaxin-2 proteins to the midbody during cytokinesis. More experiments are needed to test this hypothesis.
Based on their critical roles in cytokinesis, Crk or paxillin deficient mice would be expected to manifest an embryonic lethal phenotype. Indeed, attempts to construct mice deficient in both CrkI and CrkII have failed, while CrkII null mice are viable,33 indicating that CrkI may be sufficient for all Crk-associated functions including cytokinesis. In contrast, paxillin deficient embryos develop normally until dying at gastrulation (E6.5-E7.5),34 interestingly a time at which paxillin begins to be expressed during normal mouse development.34 Prior to this time point, Hic-5 and leupaxin, two closely related paxillin family members,35 are expressed34 and potentially could compensate for the absence of paxillin in cytokinesis and other critical processes. Interestingly, paxillin versus Hic-5 and leupaxin exhibit mutually exclusive tissuespecific expression. While paxillin is expressed in all epithelial cell lines examined thus far, Hic-5 and leupaxin expression have not been detected in epithelial tissues or cell lines, including HeLa cells,35-37 providing an explanation for the requirement for paxillin in cytokinesis in HeLa cells.
The hierarchical order of assembly of midbody components and their interactions remain largely unknown. The presence of hundreds of proteins in this compartment, however, highlights the need for adaptor proteins to act as molecular glue during cytokinesis. Scaffolding proteins such as paxillin and Crk possess multiple protein interaction domains12,38,39 and are highly suited to function as centers for the assembly of the multi-protein signaling and structural complexes at the midbody.
A growing body of evidence suggest that traction forces and/or cell migration may contribute to cytokinesis. Traction forces have been suggested to contribute to cytokinesis in dividing amoebae.40 During furrowing and midbody extension in dividing human fibroblasts, force is generated,41 supporting the notion that traction-mediated cell migration may contribute to daughter-cell scission.41 Since Crk, paxillin and Src play pivotal role in cell migration, it is intriguing to postulate that in addition to their role in recruiting syntaxin-2 and perhaps other essential components of late cytokinetic apparatus, Crk and paxillin may also indirectly contribute to cytokinesis by providing traction-mediated force during abscission. We are currently conducting research aimed at examining this hypothesis.
In summary, our results demonstrate that paxillin and Crk play key roles in late stages of cytokinesis. Our findings provide a solid foundation to further elucidate the role of focal adhesion components in this fundamental biological process.
Antibodies were obtained from the following sources: CrkI/II, paxillin and pY118-paxillin (BD Transduction Laboratories, San Jose, CA), Src (Abgent, San Diego, CA), β-tubulin (Sigma, St. Louis, MO), RhoA and citron K (Santa Cruz Biotechnology, Santa Cruz, CA), syntaxin-2 for IF (Synaptic Systems, Goettingen, Germany) and for Western blot (Calbiochem, San Diego, CA), GFP (Novus Biologicals,Littleton, CO), Alexa-conjugated IgG (Molecular Probes, Eugene, OR), GAPDH (Chemicon International, Temecula, CA), and HRP-conjugated IgG (Jackson Immunoresearch, West Grove, PA).
Crk I/II siRNA (human) with sense sequence; 5′-CUGCUUACCCUGAUUUAUUtt-3′, and antisense sequence; 5′-AAUAAAUCAGGGUAAGCAGtg-3′, was designed by Ambion, (ID#: 202984), (Austin, TX). Paxillin siRNA (sc-29439), syntaxin-2 siRNA (sc-41326), and control siRNA (sc-37007) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). SU6656 (Sigma) and used at 2 μM final concentration. Gleevec was used at 10 μM concentration.
In all experiments, wells or coverslips were precoated with poly-L-lysine (Sigma), followed by human fibronectin as described.10 HeLa cells were grown and maintained as described.10 HEK primary cells and its growth medium were purchased from Cell Applications, INC., (San Diego, CA) and used in described experiments after the second passage. For experiments involving IF videomicroscopy, cells were grown in the same media without phenol red for 3 days prior to seeding. Cells were then seeded in the same media without phenol red and that media was augmented with ZVAD-fmk (R&D Systems, Minneapolis, MN) at 60 μM to block cell death and propidium iodide (PI) at 7 μg/ml (Sigma, Saint Louis, MO) to identify dying cells.
Restriction enzymes were from New England Biolabs (Ipswich, MA). Primer sequences shown in lowercase below are not homologous to the chromosomal sequences and contain the engineered restriction sites. The pIRES2-EGFP mammalian vector (BD Biosciences/Clontech; Mountain View, CA) was modified as follows. Using long PCR42 and primers; EGFP-f3, 5′-aaaactagtATGGTGAGCAAGGGCGAGG-3′ and EGFP-r2, 5′-aaaagctagcGGATCTGACGGTTCAC-3′, which contain Spe1 and Nhe1 respectively, the intervening sequences between the promoter and the EGFP coding sequence including the IRES were removed.
To construct CrkI, CrkI/R38K, CrkI/W170K or CrkII C-terminal fusions to GFP primers; crk-f2, 5′-ttttgctagcATGGCCGGGCAGTT-3′, containing Nhe1, and either crkI-r3, 5′-tttgggactagtGCGACCTCCAGTCAG-3′, with engineered Spe1 site, for wild type CrkI, CrkI/R38K and CrkI/W170K, or crkII-r3, 5′-accccactagtGCTGAAGTCCTCATC-3′, with engineered Spe1 site, for CrkII were used to amplify these Crk variant, using a PCR program described previously43 and the pEBB plasmid series containing Crk variants, provided by Drs. Morag Park and Bruce Mayer,15,16 as templates in these PCR reactions. These PCR products were then digested with Nhe1 and Spe1 and directionally cloned into the modified pIRES2 vector.
Transient transfection and IF microscopy were done as described.10 All primary antibodies were added at 1:50 dilution except anti β-tubulin (1:200) and goat and rabbit polyclonal anti-GFP (1:300). The secondary antibodies (all at 1:400 dilution) and Hoechst 33342 nuclear stain (at 5 μg/ml) were added in the same blocking solution.
All video timelapse microscopy studies were performed as described previously,44 except video images were captured every 5 min for phase only, or every 15 min for phase and IF.
HeLa cells were plated into 6-well dishes and transfected with 100 nM CrkI/II, paxillin, syntaxin-2, or control siRNA (see above) or transfected with the indicated expression vectors. 48 hrs after siRNA transfection or 24 hrs post-transfection, the rounded, non-adherent cells contained in the media and washes were collected by centrifugation at 1,000 rpm for 5 min. The cell pellet and remaining adherent cells were lysed and analyzed by Western blot as described.44
The statistical analyses were based on either the two-tailed students t test and represented as mean ± SEM, or the Chi-square test that was used to determine the pair-wise statistical significance. p < 0.05 was taken as significant.
We thank Dr. Andy Finch and the Mt. Zion Cancer center for technical assistance and Drs. Zena Werb, Pat O’Farrell, Eric Brown, Daniel Portnoy, Zhien Wang, Jessica Ray, and members of Engel lab for their advice and suggestions. We appreciate the generous gift of reagents from Drs. Morag Park and Bruce Mayer. This work was supported by the NIH grants F32 AI054056 (to SS), R01 AI053194 (to JE and KM) and R01 AI42806 (to JE).