Dia-Interacting Protein (DIP) Imposes Migratory Plasticity in mDia2-Dependent Tumor Cells in Three-Dimensional Matrices

Tumor cells rely upon membrane pliancy to escape primary lesions and invade secondary metastatic sites. This process relies upon localized assembly and disassembly cycles of F-actin that support and underlie the plasma membrane. Dynamic actin generates both spear-like and bleb structures respectively characterizing mesenchymal and amoeboid motility programs utilized by metastatic cells in three-dimensional matrices. The molecular mechanism and physiological trigger(s) driving membrane plasticity are poorly understood. mDia formins are F-actin assembly factors directing membrane pliancy in motile cells. mDia2 is functionally coupled with its binding partner DIP, regulating cortical actin and inducing membrane blebbing in amoeboid cells. Here we show that mDia2 and DIP co-tether to nascent blebs and this linkage is required for bleb formation. DIP controls mesenchymal/amoeboid cell interconvertability, while CXCL12 induces assembly of mDia2:DIP complexes to bleb cortices in 3D matrices. These results demonstrate how DIP-directed mDia2-dependent F-actin dynamics regulate morphological plasticity in motile cancer cells.


Introduction
Metastatic cancer cells are capable of migration and invasion through extracellular matrix (ECM) barriers present in tissues, intravasation into lymphatics or bloodstream, extravasation, and dissemination and growth at a new site and these processes depend upon dynamic modulation of the cytoskeleton. Regulation of the actin cytoskeleton is fundamentally important for driving cell migration and invasion. Much work has focused upon the role of the Rho family of GTPase proteins in regulating cancer cell migration, invasion and metastasis. One family of GTPase effecter proteins that are critical for regulating the F-actin cytoskeleton during migration is the mammalian diaphanous-related (mDia) formins, including mDia1-3. mDia formins are evolutionarily conserved proteins that nucleate, elongate and, in some cases, bundle F-actin filaments that underlie cytoskeletal structures, such as filopodia, lamellipodia and ruffles [1,2]. Like all mDia formins, mDia2 is maintained in an autoinhibited formation until a GTPase binds to the GTPase-binding domain [3]. GTPase binding to the formin releases the autoinhibited conformation, allowing protein effecters to bind. mDia2 and other formin family members are implicated in the formation of actin-rich structures important for cellular motility in both normal and cancer cells (reviewed in [4]). Knockdown of the zebrafish mDia2 homologue zDia2 revealed a role in the formation of actin-rich protrusions controlling cellular migration during gastrulation [5]; zDia2 directed the formation of non-apoptotic membrane blebs in marginal deep cells at the germ-ring stage of gastrulation. Membrane blebbing is an initial indicator of cellular motility accompanying the transformation of non-motile blastomeres into motile blastula cells, suggesting a role for Dia2 homologues in cellular migration in vivo. mDia2 regulates Focal Adhesion (FA) stability in migrating epithelial cells [6], and was localized to the lamellar region and maintained a stable pool of Factin critical to FA turnover. MDA-MB-231 breast cancer cell invasion was dependent upon mDia formins, and mDia2 was localized with Src to invadopodia; mDia2-depleted MDA-MB-231 cells had few invadopodia, pointing towards mDia2 as an important component in cancer cell invasion [7]. Deletion of the gene encoding murine mDia1, DRF1, revealed defects in assembly of the F-actin architecture important for T cell migration and trafficking [8,9], in neutrophil polarization and chemotactic responses [10,11], and in dendritic cell migration [12]. In vitro, formins control cortical actin contractility and their disruption promotes membrane blebbing and amoeboid morphology in cervical and prostate cancer cells [3,13]. In the latter study, mDia2 depletion enhanced EGF-dependent blebbing and increased motility and invasion. Analysis of human prostate tumors revealed that 20% of primary and 60% of metastatic tumors had lost DRF3/DIAPH3, the gene encoding mDia2, linking mDia2 expression and/or function with metastasis; genomic loss of DIAPH3 expression was subsequently linked to disease progression in invasive human breast and hepatocarcinoma, and siRNA-mediated suppression of DIAPH3 enhanced prostate cancer lung metastasis in a tail vein injection xenograft model [14]. Other formins, (i.e., mDia1, Formin-like 2, FHOD, FRL1/FMLN1) also drive membrane blebbing and amoeboid cancer cell motility in vitro [15,16,17,18,19], suggesting a conserved role for formins in regulating morphological plasticity during cancer cell motility. However, effecter proteins modifying formin-dependent signaling cues are poorly understood.
One mDia2 effecter protein is the diaphanous-interacting protein (DIP). DIP is a multi-domain protein implicated in stress fiber formation, cell migration, dendritic spine formation, vesicular trafficking and FA assembly [20,21,22,23,24]. DIP binds both mDia and WASp-family proteins, interacting through its aminoterminal SH3 domain with N-WASp or mDia via conserved proline-rich stretches within the FH1 domain [23,25]. The mouse DIP orthologue, WISH, induced Arp2/3-mediated nucleation by interacting in a cdc42-independent manner with N-WASp [25]. DIP reportedly directly binds and activates the Arp2/3 complex, although this was not confirmed in cells [20]. The DIP knockout mouse showed slight diminution of actin accumulation in dendritic spines on explanted neurons [26], enhanced synaptic functions and motor function testing, and impaired MEF migration [27,28].
A key understudied DIP feature is a conserved, leucine-rich repeat (LRR) region that directly binds mDia1 and mDia2 FH2 domains. We demonstrated that mDia2 binding to the DIP LRR is controlled by the Rho-governed autoregulatory mechanism mediated by Diaphanous-inhibitory domain (DID) and Diaphanous-autoregulatory domains (DAD) [3]. Full-length DIP and DIP LRR act as functional negative regulators of mDia2-mediated actin filament assembly and bundling, having no effect on mDia1directed F-actin assembly. DIP inhibited mDia2-directed filopodia assembly and triggered non-apoptotic membrane blebbing, independent of either mDia1 or Arp2/3/WASp. These data identified a role for mDia2 in regulation of the cortical F-actin cytoskeleton and suggested that mDia2 interaction with DIP destabilizes the F-actin cortex to drive membrane blebbing, a hallmark of amoeboid motility. It remained uncertain whether the spatial and temporal association of mDia2 and DIP during membrane blebbing supported this mechanism, and whether DIP was required for driving amoeboid transitions in mesenchymal cancer cells.
Physiological stimuli driving amoeboid motility in cancer cells have yet to be identified. However, the chemokine CXCL12 (SDF-1a) induces membrane blebbing and amoeboid transitions in multiple experimental systems including zebrafish [29]. Zebrafish primordial germ cells bleb upon CXCL12 stimulation potentially through polarized activation of the CXCL12 receptor CXCR4. Upon CXCL12 stimulation, cells showed increased F-actin and cellular calcium [30], the latter demonstrated to be involved in membrane blebbing [29].
Here we examine the role of the mDia2:DIP axis in directing amoeboid transitions and driving blebbing in cancer cells. mDia2 and DIP co-localize to membrane blebs in constitutively blebbing M2 cells, and DIP depletion inhibits blebbing. We show a central role for expression and/or function of the mDia2:DIP axis in driving morphological plasticity in mesenchymal MDA-MB-231 and amoeboid MDA-MB-435S cells in both 2D and in 3D matrices. Finally, we identify CXCL12 as a driver of amoeboid transitions in MDA-MB-231 cells within 3D matrices, inducing spatial and temporal association of mDia2 and DIP. These data reveal an important role for DIP-directed mDia2 F-actin dynamics in regulating morphological plasticity in motile cancer cells.

Spatial Localization of mDia2 and DIP in Constitutively Blebbing Cells
M2 melanoma cells have diminished expression of Filamin A, an F-actin cross-linking protein [31,32], and constitutively extrude large blebs with extended half-lives. M2 cells were plated upon glass coverslips to stimulate profuse blebbing. The spatial localization of mDia2, DIP, and F-actin were assessed ( Figure 1). mDia2 localized to large punctae at the bleb base and to the rim of expanded blebs ( Figure 1B), where it co-localized with F-actin ( Figure 1A, D, F). DIP localization was highly vesicular and concentrated primarily at the bleb bases where it co-localized with mDia2 ( Figure 1C, E), implicating a potential role in bleb initiation and expansion. DIP also localized within the bleb and, to a lesser extent, to the rim of the bleb.

DIP is Required for Non-apoptotic Plasma Membrane Blebbing
DIP suppression enhanced mDia2-directed filopodia assembly in HeLa cells, while DIP overexpression promoted blebbing [3].
To determine the requirement for DIP in blebbing, we depleted DIP from M2 cells using either pooled siRNA or individual siRNAs (or, as a control, GAPDH) ( Figure 1G). Morphological evaluation revealed a dramatic shift towards an elongated, mesenchymal morphology devoid of blebs ( Figure 1I). A decrease (p,0.001) in DIP-depleted blebbing cells was revealed relative to control cells ( Figure 1H), indicating that DIP is a necessary component driving M2 blebbing and amoeboid morphology.  (Figure 2A and 2B), yet expression did not segregate with morphologies. MDA-MB-231 and MDA-MB-435S cells were utilized for the remainder of the studies as they are well characterized with morphological plasticity that can be experimentally manipulated [18,33,34]. We evaluated mDia2 and DIP localization in cells plated upon a thin 2D layer or within a thick 3D gel of type-I collagen ( Figure 2C, D). In 2D, MDA-MB-435S cells ( Figure 2D) were predominantly polarized with broad lamellae, mostly devoid of blebs. mDia2 expression was perinuclear with distinct tubular staining, whereas DIP staining was highly vesicular. MDA-MB-231 ( Figure 2C) morphologies were mesenchymal, with perinuclear and tubular mDia2 staining and punctate DIP staining, which overlapped slightly in the perinuclear region.

Spatial and Temporal Regulation of DIP and mDia2
In contrast, when embedded in 3D collagen matrices, MDA-MB-435S cells were predominantly amoeboid with hallmark blebs. mDia2 localized to blebs, and in particular, to bleb rims. DIP also localized to blebs, where it co-localized with mDia2. MDA-MB-231 cells were mesenchymal in 3D gels, with tubular mDia2 staining and associated perinuclear DIP.
We evaluated whether mDia2 and DIP complexed together upon EGF stimulation, which induced blebbing in HeLa [3] and in MDA-MB-231 cells ( Figure S1). We transfected MDA-MB-231 cells with both mDia2 and DIP and stimulated with 10 nM EGF for 0-20 min. Upon 10 min of stimulation, DIP associated with mDia2 ( Figure 2E), suggesting both a spatial and temporal

DIP Drives Morphological Plasticity in Migrating Cells in 3D Matrices
MDA-MB-435S cells were then treated with siRNA targeting GAPDH, DIP ( Figure 3A-C) or treated with Rho kinase inhibitor Y-27632. Cells were embedded in collagen gels for 24 hrs, fixed and the F-actin architecture assessed. While untreated MDA-MB-435S cells retained a characteristic amoeboid and bleb-enriched morphology in 3D, Rho-kinase suppression drove a mesenchymal morphological switch, seen previously in these and other amoeboid tumor cells [16,35,36,37]. DIP depletion revealed a mesenchymal transition, whereas GAPDH depletion yielded no transition. Elongation indices were calculated by dividing the long cellular axis by the short axis, revealing increases in MDA-MB-435S DIP-depleted cells relative to GAPDH ( Figure 3C, p,0.0005). These data indicate a requirement for DIP for maintaining MDA-MB-435S amoeboid morphology in 3D matrices. Phospho-MLC2 levels were then assessed in DIPdepleted cells in 3D matrices ( Figure 3D), as it is a hallmark characteristic of amoeboid morphologies. Indeed, control depleted amoeboid MDA-MB-435S revealed robust levels of pMLC2, while DIP-depleted, mesenchymal cells showed minimal pMLC2 levels, consistent with a loss of amoeboid morphology. Finally, the invasive capacity of DIP-depleted MDA-MB-435S cells was assessed using a transwell collagen invasion assay ( Figure 3E). DIP-depleted cells were allowed to invade through collagen I for 48 hrs in response to a 10% serum gradient. Upon DIP depletion using either pooled or an individual siRNA and a concomitant mesenchymal transition, a marked reduction in invasion was observed relative to controls.
In contrast, MDA-MB-231 cells were treated with either DMSO, the MMP-inhibitor GM6001, or transfected with CFP or CFP-fused DIP LRR domains ( Figure 4A, B). MMP inhibition induced an amoeboid transition in MDA-MB-231 cells [33], as did CFP-DIP LRR-expressing cells ( Figure 4C, p,0.0001), relative to control CFP-expressing cells. CFP-DIP LRR-N555A, which abrogates binding to the mDia2 FH2 domain [3] and fails to disrupt FH2-dependent F-actin nucleation (K.M.E., unpublished observations), did not induce an amoeboid transition ( Figure 4C, p = 0.96) suggesting that interaction between DIP and mDia2 is required for the DIP-driven amoeboid transition in MDA-MB-231 cells in 3D matrices. Furthermore, DIP LRR expression enhances pMLC2 levels in cells in 3D matrices, while DIP LRR N555A cells failed to do so ( Figure 4D), confirming the amoeboid transition in DIP-LRR-expressing cells. Finally, invasion of amoeboid-transitioned CFP-DIP LRR-expressing cells was quantified by reverse invasion assay for 24 hrs ( Figure 4E). Indeed, relative to CFP expression alone, expression of CFP-DIP LRR (and suppression of mDia2 FH2 activity) dramatically enhanced the proportion of cells migrating to depths beyond 30 mm; this enhanced invasion was sensitive to ROCK1/2 inhibition, as treatment with Y27632 diminished the proportion of cells invading beyond 30 mm. These data collectively indicate that the DIP-directed mDia2-dependent amoeboid transition impacts morphological plasticity in motile tumor cells.

The Role of CXCL12 in Membrane Blebbing
The physiological stimuli driving amoeboid motility in cancer cells both in vitro and in vivo have yet to be identified; however, CXCL12 supports blebbing and amoeboid motility [29,38] and CXCL12 is highly expressed in tissues that are prevalent sites of distant breast cancer metastasis [39,40]. We examined if CXCL12 treatment induced blebbing. Expression of the receptor, CXCR4, was first confirmed in a panel of amoeboid and mesenchymal cells, with comparable expression across the panel ( Figure S2). MDA-MB-231 cells were treated with CXCL12 (15-100 ng/ml) ( Figure 5A and data not shown); 15-30 ng/ml stimulated profuse micro-blebbing ( Figure 5A, inset) in 25-30% of cells within 15 min and was sustained through 60 min.
We assessed if CXCL12 induced a complex between DIP and mDia2, driving the DIP-mediated blebbing mechanism. MDA-MB-231 cells were transfected with FLAG-mDia2 and stimulated with CXCL12 for 0-8 hrs. Co-immunoprecipitation revealed that CXCL12 treatment increases the association between DIP and mDia2 through 1 hr, followed by sharp disassociation of the complex ( Figure 5B). We then evaluated how CXCL12 treatment affected the spatial localization of DIP and mDia2. In untreated MDA-MB-231 cells, DIP localized to vesicular structures, while mDia2 was predominantly perinuclear and localized to tubular structures, with some perinuclear co-localization of proteins ( Figure 5C). Upon CXCL12 treatment, copious F-actin-enriched small blebs formed with mDia2 and DIP co-localized to the blebs ( Figure 5D), demonstrating that CXCL12 treatment induces the association of mDia2 with DIP to blebs. mDia2 or DIP were depleted from MDA-MB-231 cells to examine if DIP or mDia2 suppression affected the kinetics of CXCL12-induced blebbing. DIP depletion, via pooled or individual siRNAs, suppressed CXCL12-mediated blebbing, relative to controls ( Figure 6A, B). To evaluate the contribution of the DIP LRR:mDia2 FH2 interaction upon morphological plasticity, YFP, YFP-DIP (full length) or YFP-DIP harboring an N555A point mutation was expressed and CXCL12-induced blebbing was assessed ( Figure 6C-E). Despite the disruption in the LRR-FH2 interaction, DIP N555A robustly interacts with mDia2 in HEK293T cells, likely through the DIP SH3-mDia2 FH1 interaction [23]. Indeed, DIP and mDia2 co-expression in MDA-MB-231 cells revealed a robust CXCL12-inducible mDia2:-DIP complex through 60 min ( Figure 6D). Impairing the DIP LRR:FH2 interaction through the N555A mutation revealed delayed and diminished mDia2:DIP complex induction upon CXCL12 stimulation. Concurrently, a dramatic increase in CXCL12-induced blebbing was observed (p,0.0001 vs. YFP) upon DIP overexpression, while a modest, yet non-statistically significant (p.0.05 vs. YFP) increase was observed upon disruption of the DIP LRR-FH2 interaction via the N555A mutant ( Figure 6E), indicating the importance of the mDia2 FH2:DIP LRR axis in driving morphological plasticity in response to CXCL12.
Furthermore, mDia2 depletion revealed both sustained and enhanced CXCL12-mediated blebbing ( Figure 6F; p,0.01 relative to uninduced). Suppression of FH2 activity alone via treatment with the FH2 inhibitor SMIFH2, which targets both mDia1 and mDia2 FH2 activity [41], was sufficient to drive amoeboid transition and blebbing in the absence of CXCL12 ( Figure 6G), consistent with findings in other tumor cells [41]. Interestingly, we demonstrated in Drf1-knockout MEFs that loss of mDia1 alone is not sufficient to drive blebbing in the presence or absence of EGF [3], indicating suppression of mDia2 and not mDia1 FH2 activity is critical for the blebbing mechanism. These data collectively reveal a role for DIP in the CXCL12-driven blebbing mechanism, through interaction with its binding partner mDia, therein acting as a functional negative regulator of FH2 activity; furthermore, suppression of mDia2 expression is sufficient to drive blebbing.
Finally, we examined whether stimuli such as CXCL12 and FBS drive mDia2 and DIP to co-localize near/into the induced     Figure 5C), CXCL12 induced a dramatic amoeboid transition in invading cells, with cells enriched for large single and compound blebs ( Figure 7A, C), and denoted by *), with punctate mDia2 and DIP co-localization to the blebs. MDA-MB-435S cells exposed to a 10% FBS gradient retained an amoeboid morphology with very large, exaggerated blebs ( Figure 7B), and subsequent invasion through 3D collagen I matrices was sensitive to ROCK inhibition ( Figure 7D). mDia2 and DIP were localized to large punctae at the bases of extruded blebs in MDA-MB-435S cells. Collectively, these data indicate that CXCL12 induces the spatial/temporal association of an mDia2 and DIP complex, therein driving DIPdependent blebbing and amoeboid transitions in motile tumor cells in both 2D and 3D matrices.

Discussion
Primary tumor cells continue to evolve novel modes of metastatic escape into distant tissues. These escape strategies require migratory plasticity, and are characterized by interconvertability between mesenchymal and amoeboid cancer cell motility programs. Morphological plasticity relies upon localized disassembly/reassembly of cytoskeletal structures that either drive and/or support the traction and protrusion cycles. Central to this process are Rho family GTP-binding proteins and their downstream effecters, formins. Here we demonstrate a functional role for mDia2 and its binding partner DIP in the interconversion of mesenchymal and amoeboid motility modes in cancer cells in 2D and 3D matrices. Co-recruitment of DIP and mDia2 to the plasma membrane is co-incident with bleb extrusion and blebbing is dependent upon DIP expression and mDia2 function. Our data reveal a physiological trigger for DIP-directed mDia2-dependent blebbing in mesenchymal breast cancer cells-CXCL12-which stimulates blebbing both in 2D and 3D matrices, while driving the assembly of the DIP:mDia2 complex to blebs.
The migratory plasticity in both 2D and 3D was dependent upon mDia2 and DIP expression/activation. As the Rhodependent release of autoinhibited mDia2 is required for DIP:mDia2 interaction [3], these data suggest an important interplay between Rho activation and the DIP-driven mDia2dependent amoeboid transition in both 2D and 3D matrices. This contrasts with 2D studies suggesting that DIP, through association with p190RhoGAP and VAV2, suppresses RhoA activity to negatively impact stress fiber formation and cellular migration [22]. Forced DIP overexpression in COS cells induced a transient decrease in GTP-RhoA upon EGF stimulation that was recoverable within 20 min. Rac-GTP levels were simultaneous elevated, consistent with lamellipodia formation, although DIP expression alone was not sufficient to activate Rac in another study [42]. In our study, in 2D and 3D matrices, EGF ( Figure S1, data not shown) or CXCL12 induced an amoeboid transition in MDA-MB-231 cells through 60 min and DIP depletion in MDA-MB-435S cells induced a mesenchymal transition. These data suggest that Rho activity predominates the cellular signaling milieu to induce the formation of the mDia2:DIP signaling axis, thus maintaining an amoeboid morphology. However, it must be considered that spatial and temporal localization of DIP with its binding partners (e.g., mDia2, N-WASp, palladin, p53 IRES or Src [23,25,42,43]) may control the assembly of DIP-directed complexes that functionally regulate actin dynamics. Importantly, our studies examine the contribution of DIP upon membrane dynamics in the context of 3D matrices, where the activity of Rho GTPases controlling migratory plasticity and the nature of actin-rich protrusion formed can be very distinct from those observed in 2D.
DIP and mDia2 predominantly co-associate at membranes associated with blebs in M2, MDA-MB-231 and MDA-MB-435S cells. These data point to a potential mechanism for blebbing and amoeboid transitions (Figure 8). In this model, localized GTP-Rho binds to and releases the autoinhibited mDia2 (Step 1), allowing it to complex with DIP at the cell cortex (Step 2). The consequence is disruption of F-actin assembly at the membrane and localized weakening of the actin cortex to induce bleb extrusion. Our data demonstrate distinct localization of mDia2 alone to the rim of extruded blebs (Figure 1B), revealing a second potential function for mDia2 in the re-assembly of the F-actin cortex preceding bleb retraction [44,45,46,47] (Step 3, 4a). RhoA localized with F-actin to the rims of extruded and retracting blebs [47], suggesting a role for RhoA effecter proteins like mDia2 in maintaining blebs. However, mDia2 interacts with multiple Rho GTPases, including RhoA-C (reviewed in [48]) and Cdc42 [3,49]. While it is uncertain if specific mDia2-interacting Rho GTPases drive the blebbing process, FHOD-mediated blebbing was dependent upon interaction with Rac [16], while RhoA was required for SH4domain-induced blebbing [19]. Yet, GTP-Rac was shown to negatively regulate pMLC2, diminishing actinomyosin contractility and inhibiting the amoeboid motility program; Rho/ROCK pro-amoeboid signaling conversely directed Rac inhibition, via activation of the Rac GAP ARH-GAP22 [50], indicating a tightly controlled balance in Rho/Rac signaling in driving morphological transitions. Finally, we observed punctate DIP and mDia2 colocalization to the bleb rims and propose that this interaction induces a secondary break in the re-establishing F-actin cortex to mediate formation of compound blebs, or lobopodia [44,51] (Step  4b).
Contrary to our previous biochemical studies [3] and the above mechanism, recent studies in S. pombe revealed that the yeast DIP orthologue, dip1p, failed to inhibit assembly of actin cable at the cell tips [52] and did not act as a formin regulator. However, the for3p FH2 domain shares less than 40% identity with the FH2 domain of its murine orthologue, mDia2, and subtle differences between mDia FH2 domains influence DIP's ability to negatively regulate FH2-mediated F-actin assembly; despite their similarity, DIP binds both the mDia1 and mDia2 FH2 domains, yet only YFP-DIP N555A. Lysates were co-immunoprecipitated for DIP-associated mDia2. Tubulin is shown as a loading control for i.p. lysates. (D, E) MDA-MB-231 cells co-transfected for 48 hrs with 3XFLAG-mDia2 and YFP, YFP-DIP or YFP-DIP N555A were stimulated with 25 ng/ml CXCL12. Lysates were coimmunoprecipitated for DIP-associated mDia2 upon CXCL12 stimulation (D) and blebbing (E) was quantified as above. (F) MDA-MB-231 LN Luc cells stably expressing inducible miRNA directed against mDia2 were uninduced or induced with Doxycycline for 72 hrs and protein depletion validated (inset) by western blotting. (Un)induced cells were plated upon glass coverslips after 72 hrs induction and stimulated for the indicated times with 30 ng/ml CXCL12 and blebbing cells enumerated, where n.105 cells. (G) MDA-MB-231 cells were plated upon glass coverslips and, upon addition of either DMSO or 10 mm SMIFH2, blebbing was quantified in n.100 cells. Inset shows representative treated cells that were fixed and stained with phalloidin. doi:10.1371/journal.pone.0045085.g006 negatively regulates the latter [3]. As with mDia1, the for3p FH2 domain may preclude dip3p from acting as a negative regulator, whereas other S. pombe FH2 domain-containing formins may be potent targets for dip1p-mediated inhibition. Currently, the exact nature of subtleties in the primary sequence or secondary structures of FH2 domains conferring both the ability to bind and be regulated by DIP are unclear. Interestingly, while both the mDia1 and mDia2 FH2 domains polymerize F-actin, only mDia2 bundles F-actin filaments [53]. Further studies would be interesting to determine the parameters FH2 domains must conform to in order to be regulated by DIP.
It is uncertain if DIP-directed blebs in invading amoeboid cells are polarized towards a chemotactic gradient to drive protrusion and directional migration. Invading MDA-MB-231 cells formed copious blebs that polarized to the contracting cell rear proximal to the uropod, and migration speed and uropod formation were mDia dependent [7,54], suggesting that blebbing may drive polarized contraction as opposed to forward membrane extension. However, Marshall and colleagues revealed the presence of a rigid, ezrin-enriched uropod at the rear of amoeboid, invasive A375 melanoma cells [55]; the presence of the rearward ROCKdependent uropod inhibited proximal bleb formation, essentially establishing rear polarity through bleb inhibition. Our data show that MDA-MB-231 cells invading towards CXCL12 through collagen matrices have polarized compound blebs extending towards the gradient (Figure 7A), indicating a potential role for CXCL12 in directing bleb and/or cellular polarity. In fact, during zebrafish germline migration, CXCL12 gradients drive localized increases in Ca 2+ to establish and stabilize protrusions that become the leading edge of the cell [29]. Experiments directed at understanding contributions of chemokine gradients to uropod formation and/or bleb polarization at the front/rear edge of migrating cancer cells are indeed warranted.
Finally, our data reveal a novel role for CXCL12 in inducing amoeboid transitions in MDA-MB-231 cells, and in recruiting the mDia2:DIP complex to blebs. CXCL12 and its receptor CXCR4 are implicated in the metastasis of several cancers including breast cancer metastasis to lung and lymph node [30,56,57]. CXCL12 is highly expressed in tissues that are prevalent sites of distant breast cancer metastasis [39,40]. CXCR4 knockdown decreased MDA-MB-231 cell invasion [58,59]. Therefore, amoeboid transitions in migrating breast cancer cells induced via exposure to CXCL12 may play a role in one or multiple steps in metastasis. Indeed, Rho degradation via the ubiquitin ligase Smurf1 locally impaired Rho/ ROCK signaling and promoted mesenchymal motility, while Smurf1 inhibition induced amoeboid motility in tumor cells in subcutaneous xenografts [60,61]. This transition was accompanied by dramatic increases in the number of tumor cells in and proximal to blood vessels, suggesting a linkage between amoeboid motility and intravasation. Ultimately, dynamic transition between mesenchymal and amoeboid motility may function during distinct stages of the metastatic process (e.g., localized invasion) when the microenvironment requires morphological adaptation of tumor cells. Together, these studies and data presented herein demonstrate an important role for DIP-directed mDia2-dependent Factin dynamics in regulating morphological plasticity in motile cancer cells, and reveal a novel role for CXCL12 in inducing amoeboid transitions within 3D matrices. These studies underscore the importance of understanding amoeboid motility in promoting mechanistic steps driving tumor metastasis.
Time lapse imaging of M2 cells was acquired with an Olympus IX-81 (63x or 40x oil objective) with an environmental chamber that maintained cells at 37uC in a 5% CO 2containing atmosphere. Images were acquired at 10-30 sec intervals.
For 3D IF, cells were mixed with 2mg/ml Type I collagen (BD Biosciences), allowed to form a gel for 1 hr at 37uC and then incubated for 24 hrs in full growth medium. Cells were then fixed with 4% PFA, permeabilized with 0.2% TX100 and incubated with primary antibody overnight at 4uC. Cells were incubated with fluorescently labeled secondary antibody and phalloidin for 3 hrs at 37uC. Gels were covered with fluoromount G (Southern Biotech) and stored in the dark. Images were acquired using a TCS SP5 multiphoton laser scanning confocal microscope (Leica Microsystems) with laser lines of 458, 488, 514 and 633 in a sequential manner. Z stack images were acquired using 0.5 mm optical sections and creating a 3D reconstruction for image analysis using MetaMorph imaging software.

Elongation Index
For calculation of elongation indices, cells were stained with Calcein AM FITC (Invitrogen), mixed into a 3D collagen matrix and gelled as above. Where indicated, control cells were treated with Y-27632 Rho-kinase inhibitor (Sigma) or MMP-inhibitor GM6001 (Sigma). 24 hrs post-embedding, cells were imaged by confocal microscopy with a 20x objective and 0.5 mm optical sections. Z-stacks were used to make projection images and Metamorph Image Analysis software was used to determine elongation index (cell long axis divided by cell short axis). Measurements were made for at least 20 cells per condition.

Invasion Assays
Invasion assays were performed as described [64], casting 2.0 mg/ml collagen I gels upon Fluoroblock transwell inserts with 8 mm pores (BD Biosciences) using 50,000 cells per insert. 10%FBS was placed in the lower chamber to stimulate invasion. Cells were enumerated by staining inserts for 15 min with 0.5 mg/ ml Calcein AM-FITC (Invitrogen) in PBS and enumerating using a fluorescent plate reader using 488ex:535em. Reversed invasion assays were performed as previously described with minor modifications [65]. Cells invaded through 2 mg/ml collagen I gels for 24 hrs, whereupon they were stained using antibodies directed against mDia2, DIP or with phalloidin. Cells were imaged by confocal microscopy using a 20x objective and 0.5 mm optical sections. Z-stacks were used to make 3D projections. When quantifying invasion, optical sections of 10 mm were analyzed for pixel intensities per slice. A 0 mm distance invaded was assigned based upon the appearance of both CFP pixels within the optical slice greater than the assigned background, and the appearance of pores as visualized by transmitted white light.

Statistical Analysis
For blebbing quantification, the average percentage of blebbing cells for each sample was taken 6 standard deviation. Elongation indices were calculated for at least 20 cells per sample and the average 6 standard error was determined. Invasion was enumerated in triplicate and the average % of control (standard invasion) or % of total pixels (reversed invasion) was calculated. Data are expressed as average 6standard deviation.
A one-tail t-test was performed to evaluate statistical significance with a 95% confidence interval. A p value of 0.05 or less was considered statistically significant. Ms. Sarah DeRaedt-Banks for technical assistance with plasmid and virus preparation; Ms. Stephanie Weber for technical assistance with screening knockdowns; and the University of Toledo Advanced Microscopy and Imaging Center for imaging assistance. We thank Dr. Art Alberts for supporting the initial stages of this work and for helpful discussions.