PPM1F controls integrin activity via a conserved phospho-switch

Integrins are essential heterodimeric cell surface receptors that mediate extracellular matrix adhesion and instruct animal cells about the chemical and mechanical properties of their microenvironment (Gahmberg et al., 2009; Hynes, 2002; Morse et al., 2014). Accordingly, integrins are instrumental for cell adhesion during development, tissue regeneration, or leukocyte extravasation, but also contribute to pathological processes such as cancer cell invasion and metastasis (Bökel and Brown, 2002; Hamidi and Ivaska, 2018; Nieswandt et al., 2009; Sekine et al., 2012; Vestweber, 2002; Winograd-Katz et al., 2014).

A major regulatory principle of integrins involves an extensive conformational change, which has been termed integrin activation (Calderwood, 2004; Sims et al., 1991; Vinogradova et al., 2002). The active conformation of integrins can be stabilized either by the presence of an extracellular ligand (outside-in activation) or by a characteristic intracellular binding event of the scaffold protein talin to the cytoplasmic tail of the integrin β subunit (inside-out activation; Hughes et al., 1996; Shattil et al., 2010; Vinogradova et al., 2002; Wegener et al., 2007). During inside-out activation, the globular head of talin binds to a conserved NPxY amino acid sequence, thereby spatially separating the α and β subunits and forcing the extracellular domains into the extended, active conformation (Anthis et al., 2009; Calderwood et al., 2002; Wegener et al., 2007). This active conformation is a prerequisite for proper integrin-mediated cell attachment to the extracellular matrix (Harburger and Calderwood, 2009; Moser et al., 2009). Cell adhesion can be further promoted by integrin clustering (Bunch, 2010; Cluzel et al., 2005; van Kooyk and Figdor, 2000), which is supported by kindlin (Li et al., 2017; Ye et al., 2013), an additional binding partner of the integrin β subunit (Bledzka et al., 2012; Harburger et al., 2009; Li et al., 2017). Together, talin and kindlin initiate the formation of large, heteromeric protein complexes at integrin cytoplasmic tails, which are termed focal adhesion sites. These structures can comprise several hundred distinct proteins, the so-called integrin adhesome (Horton et al., 2015; Zaidel-Bar and Geiger, 2010; Zaidel-Bar et al., 2007).

Besides talin and kindlin as positive regulators of integrin function, several negative regulators of integrin activity such as filaminA, Dok1, Sharpin, or ICAP-1 have been described (Bouvard et al., 2003; Kiema et al., 2006; Liu et al., 2015; Oxley et al., 2008; Rantala et al., 2011). These nonenzymatic proteins are thought to act by competitive binding to the integrin β subunit, where they displace positive regulators of integrin activity. For example, filaminA and talin have overlapping binding sites in the leukocyte-specific integrin subunits β2 and β7, which they occupy in a mutually exclusive manner (Kiema et al., 2006; Takala et al., 2008). Interestingly, an evolutionary conserved threonine motif within the context of the filaminA and talin core binding sites is located in the cytoplasmic tails of most integrin β subunits (T788/T789 in the human integrin β1; Fig. 1 A and Fig. S1 A; García-Alvarez et al., 2003; Gingras et al., 2009; Kiema et al., 2006; Liu et al., 2015; Wegener et al., 2007). Upon cell stimulation, these threonine residues are phosphorylated (Buyon et al., 1990; Chatila et al., 1989; Craig et al., 2009; Hibbs et al., 1991; Hilden et al., 2003), and mutations mimicking Ser/Thr phosphorylation lead to enhanced integrin activity and integrin-based cell adhesion in vitro (Craig et al., 2009; Nilsson et al., 2006). In contrast, alanine substitution of this particular threonine motif severely compromises integrin function, leading to impaired integrin activation and abrogation of cell-matrix adhesion (Fagerholm et al., 2005; Hibbs et al., 1991; Nilsson et al., 2006; Wennerberg et al., 1998). These prior findings indicate that the conserved T788/T789 residues could form a phospho-switch to regulate integrin affinity and, thereby, control integrin-mediated cellular processes. However, the enzymatic machinery operating this phospho-switch within the cell is currently unknown.

Here we report that phosphorylation of the conserved threonine motif in the cytoplasmic tail of the integrin β1 subunit dissociates filaminA to allow access of talin to its canonical NPxY binding site. Using a focused genetic screen, we identify a member of the metal-dependent protein phosphatase (PPM) family, the serine/threonine phosphatase PPM1F, as the critical enzyme responsible for dephosphorylating the threonine motif. Our results uncover the mechanistic details of integrin activity regulation by this conserved phospho-switch and identify the underlying enzymatic machinery, thereby providing a novel access point to modulate integrin activity.

The T788/T789 motif of the β1 integrin cytoplasmic tail, which is located within the context of talin, kindlin, and filaminA binding sites, is highly conserved across species and within different human β subunits (Fig. 1 A and Fig. S1 A). Previous studies using the leukocyte-specific integrins β2 and β7 already suggested that these threonine residues could operate as a phospho-switch to control binding of talin versus filaminA (Kiema et al., 2006; Takala et al., 2008). To evaluate the consequences of integrin β1 T788/T789 phosphorylation for filaminA or talin binding in vitro, we produced recombinant cytoplasmic domains mimicking T788 or T788/T789 phosphorylation (T/D or TT/DD) or harboring nonphosphorylatable alanine residues (TT/AA; Fig. S1, B and C). We also generated a Y783A mutant in the NPxY motif, which impairs talin and filaminA association, and mutated a tyrosine residue outside of the talin or filamin core binding sites to alanine (Y795A; Calderwood et al., 1999; Pfaff et al., 1998). Pull-down experiments with the various integrin cytoplasmic domains showed that His-Small Ubiquitin-Related Modifier (SUMO)-tagged yeast enolase as irrelevant control protein did not associate with any of the integrin β1 cytoplasmic domains (Fig. 1 B). The recombinant His-SUMO-tagged integrin binding domains of filaminA (Ig domain 19–21 of filaminA) and talin (F3 lobe of the talin head domain) bound the integrin β1 WT and TT/AA variants, but not the Y783A variant (Fig. 1 B), in line with previous reports (Calderwood et al., 1999; O’Toole et al., 1995; Pfaff et al., 1998; Tadokoro et al., 2003). Importantly, talin also showed unaltered association with the pseudo-phosphorylated integrin β1 variants T/D and TT/DD, while filaminA binding was reduced (T/D) or completely absent (TT/DD), indicating that filaminA–integrin interaction is controlled by the phosphorylation state of the T788/T789 motif (Fig. 1 B). Similar results were obtained by solid phase binding assays, where one of the binding partners was immobilized (Fig. 1 C). Again, the pseudo-phosphorylation of T788/T789 led to a complete loss of filaminA binding (Fig. 1 C). Furthermore, pull-down assays with biotinylated, synthetic peptides covering residues 762–798 of integrin β1 in either the unphosphorylated or the phosphorylated (pT788/pT789) form confirmed that phospho-threonine residues at these positions impede filaminA binding (Fig. 1 D). Modeling of the pT788/pT789 β1A integrin peptide based on known structures of the filaminA/integrin β7 complex (PDB2BRQ; Fig. 1 E) or the talin/integrin β1D complex (PDB3G9W; Fig. 1 F) suggested that phosphorylation of T788 and T789 not only sterically obstructs the binding interface with filaminA but also leads to charge repulsion. In contrast, integrin β1 pT788/pT789 phosphate groups do not interfere with the binding interface to talin (Fig. 1 E). These structural models are in line with data reported previously for β2 and β7 (Kiema et al., 2006; Takala et al., 2008) and strongly support our conclusion that phosphorylation of the integrin β1 T788/T789 motif disrupts filaminA binding, but does not impact talin association.

To validate our in vitro findings in the cellular context, we performed Opa-protein–triggered integrin clustering (OPTIC) assays (Baade et al., 2019; assay scheme, Fig. S1 D; construct expression, Fig. S1 E). Clustering of the WT integrin β1 cytoplasmic domain at the plasma membrane led to a strong recruitment of GFP-talin, while GFP alone was not enriched (Fig. 2 A). Furthermore, talin was equally well recruited to the pseudo-phosphorylated TT/DD and, to a slightly lesser extent, to the TT/AA variant (Fig. 2 A). Only disruption of the NPxY core binding motif (integrin Y783A) abolished talin recruitment (Fig. 2 A). In contrast, the filaminA integrin binding domain was only recruited in the case of the TT/AA mutant, where phosphorylation of this motif is impossible, while clustering of the WT integrin as well as the phospho-mimicking TT/DD variant did not support filaminA recruitment in the intact cell (Fig. 2 B). These results demonstrated that the phosphorylation status of the integrin β1 T788/T789 motif dictates the association with integrin activity regulators in intact cells. Moreover, these findings also suggested that filamin can only occupy the talin binding site if the T788/T789 motif in the integrin tail is dephosphorylated. To test this idea, we performed in vitro competition assays with recombinant talin, filaminA, and integrin β1 cytoplasmic domains (Fig. 2 C). While talin was not able to displace filaminA from integrin β1 (Fig. S1 F), increasing levels of filaminA led to a sharp drop in talin association with the WT integrin β1 tail (Fig. 2 C). However, in the case of the pseudo-phosphorylated integrin β1 tail, even a large excess of filaminA was not able to outcompete talin (Fig. 2 C). Identical results were obtained by solid phase binding assays (Fig. S1 G). These biochemical findings support the idea that dephosphorylation of integrin β1, and in particular of the T788/T789 motif, is a prerequisite to allow filaminA to displace talin and to inactivate integrins (Fig. 2 D).

Based on the observations that a phosphorylated T788/T789 motif impedes filamin binding and that the pseudophosphorylated integrin β1 T788D promotes cell adhesion (Craig et al., 2009; Nilsson et al., 2006), we hypothesized that an integrin-directed protein phosphatase(s) counteracts integrin activation. Accordingly, deletion of such a putative protein phosphatase should lead to a gain of function with regard to integrin-based cell-matrix adhesion. Therefore, we performed a focused genetic knock-down screen with shRNAs individually targeting all protein phosphatases reported in the integrin adhesome (Zaidel-Bar et al., 2007; Fig. S2 A). As the cellular system, we deliberately chose 293T cells, a human cell line exhibiting weak adhesion to extracellular matrix proteins under tissue culture conditions. Stable knock-down cells were plated on the integrin ligands fibronectin or collagen or on poly-L-lysine, to which cells attach independently of integrins (Fig. 3 A). Compared with control cells, depletion of the protein tyrosine phosphatases (PTPs) PTP-1B, PTP-PEST, and RPTPα as well as depletion of the serine/threonine phosphatase PPM1F (also known as POPX2 [Koh et al., 2002], Ca²⁺/calmodulin-dependent protein kinase phosphatase [CaMKP; Ishida et al., 2008], and hFEM2 [Tan et al., 2001]) resulted in enhanced cell adhesion to collagen and/or fibronectin, but not poly-L-lysine (Fig. 3, A and B). PTP-1B, PTP-PEST, and RPTPα dephosphorylate the focal adhesion proteins paxillin, p130^CAS, and c-Src, respectively, which could indirectly affect integrin-mediated adhesion (Arias-Salgado et al., 2005; Garton et al., 1996; Shen et al., 2000). As PPM1F, a member of the PPM (Moorhead et al., 2009), dephosphorylates serine/threonine residues and has not been implicated in cell adhesion, we decided to focus on this enzyme. Western blotting (WB) confirmed the depletion of the phosphatase in knock-down cells and demonstrated that levels of several key focal adhesion proteins such as integrin β1, talin, filamin, kindlin-2, focal adhesion kinase, paxillin, vinculin, zyxin, integrin-linked kinase (ILK), ezrin, or α-actinin were not altered (Fig. S2 B). These results indicated that reduction of PPM1F is directly connected to increased integrin-mediated cell adhesion.

PPM1F knock-down in normal human dermal fibroblasts (NHDF) recapitulated the phenotype observed in 293T cells leading to enhanced cell adhesion on integrin ligands (Fig. 3, C and D). Depletion of PPM1F did not affect expression of integrin subunits or other cytosolic focal adhesion proteins (Fig. S2, C and D), suggesting that the increased adhesion might be due to alterations in integrin activity. Indeed, PPM1F knock-down NHDFs exhibited elevated levels of active integrin β1 (Fig. 3 E), and the active receptor was enriched at peripheral focal adhesions, where prominent recruitment of talin was observed (Fig. 3 F). Similar levels of active integrin β1 and strong recruitment of talin were also seen in NHDF cells with knock-down of filaminA, while knock-down of integrin β1 eliminated the integrin staining and confirmed the specificity of the used integrin antibody (Fig. S2, E–G). Together, reduction of PPM1F results in elevated levels of active integrin accompanied by increased accumulation of talin phenocopying the depletion of the negative integrin regulator filaminA.

PPM1F is ubiquitously expressed, but with high levels in neuronal cells (Ishida et al., 2018). To confirm the phenotype of PPM1F knock-down cells, we disrupted the Ppm1f gene in human glioblastoma A172 cells by CRISPR/Cas9. Compared with A172 WT cells and A172 control cells, which were transduced with a vector lacking the PPM1F sgRNA, the derived A172 PPM1F KO cells completely lacked expression of PPM1F (Fig. 4 A). Expression of focal adhesion proteins was unaltered and surface expression of different integrin subunits was not increased in A172 PPM1F KO cells (Fig. S3, A and B). Similar to PPM1F knock-down 293T and NHDF cells, the A172 PPM1F KO cells showed a 1.5–2-fold increase in cell adhesion compared with A172 WT or control cells at different time points after plating on substrates coated with low, medium, or high concentrations of extracellular matrix ligands (Fig. 4 B; and Fig. S3, C and D). Furthermore, A172 PPM1F KO cells displayed elevated levels of active integrin β1 (Fig. 4 C) and exhibited a prominent increase of active integrin β1 in the form of a peripheral “active integrin belt” (Fig. 4 D), which colocalized with enlarged clusters of talin (Fig. 4 D and Fig. S3 E). This phenotype was seen in around 80–90% of A172 PPM1F KO cells during the first 1–2 h of spreading (Fig. 4 E and Fig. S3 F). In general, PPM1F KO cells did not spread as fast as A172 WT cells and, therefore, covered a smaller area (Fig. 4, F and G; and Fig. S3 G), suggesting that cell spreading might be compromised due to intensified integrin–matrix interaction. This observation also indicates that other PPM1F substrates such as p21-activated kinase (PAK) or mammalian Diaphanous-related formin 1 (mDia1), which promote actin-based cell protrusions and which are negatively regulated by PPM1F, might not be responsible for the spreading defect of PPM1F KO cells (Koh et al., 2002; Parrini et al., 2009; Xie et al., 2008). To further confirm that the increased cell adhesion of PPM1F-deficient cells is connected to filamin-dependent activity regulation of integrins, we performed epistasis experiments. Therefore, A172 control cells and PPM1F KO cells received either a control shRNA or shRNA targeting human filaminA (Fig. 4 H). Similar to the KO of PPM1F and consistent with the known inhibitory role of filaminA (Liu et al., 2015; Takala et al., 2008; Waldt et al., 2018), shRNA-mediated knock-down of filaminA in A172 control cells increased cell adhesion, reduced cell spreading, and elevated integrin activity (Fig. 4, I and J; and Fig. S4, A–D). However, depletion of filaminA in PPM1F KO cells did not further elevate the increased integrin-dependent adhesion or the enhanced integrin activity in these cells, nor did it further reduce cell spreading (Fig. 4, I and J; and Fig. S4, A–D). The results of these epistasis experiments highlight the strong similarities in the phenotype of PPM1F KO cells and filaminA knock-down cells and suggest that PPM1F and filaminA work together in the same pathway controlling integrin activity (Fig. S4 E). Together, the absence of PPM1F results in a gain of function with regard to integrin-based cell adhesion due to enhanced integrin activity, elevated talin recruitment, and reduced filaminA association with integrin β1.

Association of filaminA and talin with the integrin β subunit as well as integrin activity are regulated by phosphorylation of the T788/T789 motif. Therefore, we wondered about the phosphorylation status of these residues in PPM1F-deficient cells. Interestingly, while suspended A172 WT cells showed low levels of integrin β1 T788/T789 phosphorylation, KO of PPM1F resulted in constitutively elevated levels of pT788/pT789 (Fig. 5 A). Upon seeding onto fibronectin, the level of pT788/pT789 increased transiently in WT A172 cells during the initial attachment phase up to 45 min (Fig. 5 B). In contrast, PPM1F KO cells permanently exhibited substantially elevated T788/T789 phosphorylation (Fig. 5 C). Quantification of multiple blots demonstrated a four- to fivefold higher phosphorylation level of integrin β1 T788/T789 in A172 PPM1F KO cells compared with WT cells (Fig. 5 D). To rigorously demonstrate that this phenotype is due to the lack of PPM1F activity, we complemented the PPM1F KO A172 cells with either WT monomeric (m)Kate-PPM1F or the phosphatase-dead mutant mKate-PPM1F D360A (Fig. S3 H). As seen before, expression of core focal adhesion proteins or surface expression of integrin subunits was not altered by this genetic manipulation (Fig. S3, A and B). However, expression of PPM1F WT, but not expression of PPM1F D360A, reverted the increased phosphorylation of integrin β1 T788/T789 back to levels seen in WT A172 cells (Fig. 5 D). The increased integrin T788/T789 phosphorylation seen in the PPM1F KO cells correlated well with the elevated integrin activity and enhanced cell adhesion to integrin ligands, which was also reverted back to basic levels upon reexpression of WT PPM1F, but not PPM1F D360A (Fig. 5, E–G). Together, these findings are consistent with the idea that PPM1F regulates the phosphorylation state of the T788/T789 motif, thereby controlling association of talin versus filaminA with the cytoplasmic tail of integrin β1 and determining cell-matrix adhesion strength (Fig. 5 H). The uniform phenotype of enhanced integrin activity observed upon depletion or disruption of PPM1F in multiple cell types also indicated that PPM1F might act directly on the integrin β1 subunit.

To test the ability of PPM1F to directly dephosphorylate pT788/pT789 of integrin β1, human PPM1F WT and PPM1F D360A were produced in Escherichia coli (Fig. 6 A). Using the generic phosphatase substrate 4-methylumbelliferylphosphate (4-MUP), maximum velocity (V_max) and Michaelis constant (K_m) values of PPM1F were comparable to other Ser/Thr phosphatases (Gee et al., 1999), while PPM1F D360A was inactive (Fig. 6 B). PPM1F also dephosphorylated synthetic peptides spanning the integrin β1 T788/T789 motif (Fig. 6, C and D). Though the doubly phosphorylated peptide (β1-pT788/pT789) served as a suitable substrate for PPM1F, dephosphorylation was more effective with the mono-phosphorylated integrin β1 peptides β1-pT788 or β1-pT789 (Fig. 6 D). PPM1F acted specifically on integrin pT788/pT789, since a phospho-peptide derived from myosin light chain (MLC; pMLC) was not dephosphorylated (Fig. 6, C and D). In a complementary approach, PPM1F was overexpressed and purified from human cells. Again, we observed dephosphorylation of synthetic integrin phospho-peptides by WT GST-PPM1F isolated from human cells, but not by GST-PPM1F D360A (Fig. S5, A–E). In a further approach, we used purified Ca²⁺/calmodulin-dependent kinase II (CaMKII) β, which has been reported as a potential kinase of integrin β1 (Suzuki and Takahashi, 2003), to phosphorylate the recombinant cytoplasmic domain of integrin β1 in the presence of ATP, Ca²⁺, and calmodulin. In contrast to WT integrin β1, the T788A/T789A mutant was not phosphorylated, demonstrating that CamKIIβ selectively acts on the integrin β1 T788/T789 motif in vitro (Fig. S5 F). Recombinant WT PPM1F, but not PPM1F D360A, dephosphorylated the resulting integrin β1 pT788/pT789 (Fig. 6 E). To check if other protein phosphatases also act on the T788/T789 motif, we recombinantly expressed several enzymes, including integrin-linked kinase-associated protein (ILKAP, an additional member of the PPM family present at integrin adhesion sites; Leung-Hagesteijn et al., 2001), PP5, and PTP1B (Fig. 6 F), and verified the activity of the purified proteins with 4-MUP (Fig. S5 G). Next, PPM1F, ILKAP, PP5, and PTP1B were incubated with the doubly phosphorylated β1-pT788/pT789 peptide. While PTP1B, PP5, and ILKAP failed to dephosphorylate integrin β1 pT788/pT789 to a significant extent, PPM1F was highly active (Fig. 6 F). Moreover, ILKAP did not dephosphorylate the GST-β1 integrin cytoplasmic domain phosphorylated in vitro by CaMKIIβ, while PPM1F was active against this phospho-protein substrate (Fig. 6, G and H). Therefore, PPM1F is the integrin phosphatase that specifically controls the phosphorylation state of the conserved T788/T789 motif in the cytoplasmic domain of integrin β subunits.

The T788/T789 motif is also at the core of the kindlin binding site in integrin β1 (Fig. 1 A), and kindlin, together with talin, is the major positive regulator of integrin function (Harburger et al., 2009; Li et al., 2017; Theodosiou et al., 2016). Therefore, we wondered if the phosphorylation state of T788/T789 also influences kindlin association with integrin β1. In vitro, recombinant kindlin2 bound the WT integrin β1 cytoplasmic domain and the Y783A mutant with the corrupted NPxY talin binding motif, but not the Y795A variant, which disrupts the core NPKY kindlin binding motif (Li et al., 2017; Fig. 7 A). Kindlin2 also did not associate with the A788/A789 variant, which is in line with previous studies showing that the hydroxyl-groups of the conserved threonine residues are involved in H-bonds between kindlin and integrin β1 (Li et al., 2017). Unexpectedly however, kindlin2 did not bind the pseudo-phosphorylated integrin β1 variants T/D and TT/DD (Fig. 7 A), and the same results were obtained by solid phase binding assays (Fig. 7 B). Furthermore, pull-down assays with biotinylated, phosphorylated (pT788/pT789), or unphosphorylated integrin β1 peptides confirmed that phosphorylation of these threonine residues prevents kindlin2 binding (Fig. 7 C). Similar to the situation with talin, kindlin2 was readily displaced from the integrin β1 tail in vitro by the integrin-binding domain of filaminA (Fig. 7 D and Fig. S1 H). These puzzling results would imply that kindlin2 is outcompeted by filaminA, when the integrin β1 tail is unphosphorylated, but kindlin2 could also not bind on its own, when the T788/T789 motif is phosphorylated. As kindlin2 cooperates with talin to modulate integrin function (Theodosiou et al., 2016; Ye et al., 2013), we wondered whether the capability of talin to associate with the phosphorylated integrin tail might enable binding of kindlin2. Therefore, recombinant full-length human kindlin2 and His-tagged talin F3 domain were employed in pull-down assays with various Twin-StrepII-tag (strep)-tagged integrin β1 cytoplasmic domains (Fig. 7 E). As before, kindlin2 alone bound the unphosphorylated integrin β1 cytoplasmic domain, but failed to associate with the pseudo-phosphorylated TT/DD variant (Fig. 7 F). Importantly, the presence of the recombinant talin F3 domain allowed the association of kindlin2 with integrin β1 TT/DD (Fig. 7 F). Despite the presence of talin, binding of kindlin2 still required the intact membrane distal NPKY amino acid motif (Fig. 7 F), suggesting that talin and kindlin2 do not physically interact, as also observed by Bledzka et al. (2012). Rather, talin binding seems to reorient the integrin tail in a manner that allows kindlin2 association with the pseudo-phosphorylated integrin. Disruption of the talin binding site in the pseudo-phosphorylated integrin β1 (TT/DD + Y783A) concomitantly abolished kindlin association (Fig. 7 F), confirming that kindlin2 binding to pseudo-phosphorylated integrin depended on the presence of talin.

To analyze if this cooperation is also true for phospho-threonine residues and if the cooperative binding with talin might allow kindlin2 to withstand the presence of filaminA, pull-down experiments with phospho-integrin peptide β1-762-798 pTpT and with the unphosphorylated peptide (β1-762-798) were performed (Fig. 7 G). Clearly, in the absence of filaminA, kindlin2 together with talin associated with the unphosphorylated as well as the phosphorylated integrin β1 peptides (Fig. 7 G). Upon addition of filaminA, kindlin2 was readily displaced from the unphosphorylated integrin β1 peptide, even though the talin F3 domain was present (Fig. 7 G). Importantly, this displacement was completely prevented by phosphorylation of the T788/T789 motif, which allowed not only talin but also kindlin2 to remain associated with the integrin β1 cytoplasmic tail in the presence of filaminA (Fig. 7 G). These observations form the basis of a refined working model for integrin activity regulation by a phosphorylation-dependent displacement of the negative regulator filaminA and the cooperative binding of talin and kindlin2 under these circumstances (Fig. 7 H).

To follow the consequences of integrin β1 T788/T789 phosphorylation for kindlin2 recruitment in intact cells, we employed integrin β1 chimeras and their phospho-mimicking mutants as well as deletion of PPM1F. In agreement with the in vitro binding data, GFP-kindlin2 clustered at WT integrin β1 tails, but was not enriched at the integrin β1 TT/AA or the integrin β1 Y795A variant (Fig. 8 A). However, in contrast to the situation with purified components, GFP-kindlin2 was also strongly enriched at pseudo-phosphorylated integrin β1 in intact cells (Fig. 8 A). The recruitment of GFP-kindlin2 in this situation was critically dependent on talin binding, as mutation of the core talin binding site in the regular (Y783A) or in the pseudophosphorylated (TT/DD + Y783A) integrin β1 tail abolished the increased association (Fig. 8, A and B).

The capability of kindlin2 to associate with full-length phosphorylated integrin β1 was also reflected by the fact that PPM1F KO cells, which display elevated levels of pT788/pT789, showed not only a peripheral ring of active integrin but also accumulation of kindlin2 at these sites (Fig. 8 B). The accumulation of kindlin2 in these structures mimicked the accumulation seen for talin in the PPM1F KO cells (Fig. 8 B). Importantly, reexpression of active, WT PPM1F, which reduces the increased integrin β1 T788/T789 phosphorylation, also prevented kindlin2 and talin accumulation, while in PPM1F KO cells reexpressing inactive PPM1F D360A, the massive peripheral ring of active integrin and the exaggerated presence of kindlin2 and talin persisted (Fig. 8 B). These data illustrate how phosphorylation of the integrin β1 subunit enforces cooperation between the integrin activators talin and kindlin2 to overcome filamin-mediated inhibition of integrins. Moreover, these results point to the critical position of PPM1F during integrin activity regulation as the major protein phosphatase directed toward the conserved T788/T789 motif.

To test whether PPM1F activity toward the conserved threonine motif in the integrin cytoplasmic tail dictates integrin activity, we again employed the chimeric receptor proteins in the unphosphorylated (WT) or pseudo-phosphorylated (T788D/T789D) forms. These integrin β1 chimeras were coexpressed with PPM1F WT or PPM1F D360A and together with either GFP-talin (Fig. 8 C) or the GFP-tagged integrin binding domain of filamin (GFP-FLN19-21; Fig. 8 D). Upon PPM1F overexpression, talin recruitment to the clustered WT β1 integrin tail was strongly impaired, while overexpression of inactive PPM1F D360A did not compromise talin association (Fig. 8 C). As predicted from the in vitro binding assays, talin also associated with integrin β1 T788D/T789D, and this association could not be diminished by overexpression of PPM1F (Fig. 8 C). Vice versa, GFP-filaminA was only enriched around clustered WT integrin β1 in situations where WT PPM1F, but not the D360A mutant, was overexpressed (Fig. 8 D). These results indicate that continuous activity of PPM1F toward the integrin β1 T788/T789 motif is needed to enable filaminA association, which nicely explains the elevated integrin activity in PPM1F-deficient as well as filaminA-deficient cells. Together, these data illustrate the consequences of PPM1F activity on protein–protein interactions occurring at the integrin β1 subunit in intact cells and suggest that this phosphatase holds a key position to set integrin activity levels.

Integrin β1 and its key interaction partners, such as talin, kindlin, filaminA, paxillin, and focal adhesion kinase, are all essential for mammals (Fässler and Meyer, 1995; Ilić et al., 1995; Stephens et al., 1995; Monkley et al., 2000; Hagel et al., 2002; Feng et al., 2006; Montanez et al., 2008). Furthermore, compromising integrin activity regulation via targeted mutations in the integrin β1 cytoplasmic domain abrogates integrin function in the intact tissue (Czuchra et al., 2006; Meves et al., 2013). Accordingly, if PPM1F indeed serves a critical function in integrin activity regulation, one would expect a severe phenotype upon inactivation of the ppm1f gene.

To test this prediction, we employed mice containing a gene-trap insertion in exon 4 of the ppm1f gene resulting in disruption of the ppm1f gene (Fig. 9 A). The genotype of the mice containing the disrupted ppm1f allele was verified by PCR (Fig. 9 B). While crosses between WT (ppm1f^+/+) and heterozygous (ppm1f^+/−) mice led to the expected 50:50 ratio of ppm1f^+/+ and ppm1f^+/− offspring, crosses between heterozygous ppm1f^+/− animals yielded no homozygous ppm1f^−/− pups, suggesting that PPM1F-deficient embryos die in utero (Fig. 9 C). Timed matings showed that ppm1f^−/− embryos were not present at embryonic day (E) 12.5, E13.5, or E14.5 of development, but putative ppm1f^−/− embryos were detected at E10.5, when ∼20% (9 out of 50 total) embryos showed retarded development (Fig. 9 D). These embryos were about half the size of the other embryos, displayed malformed forebrain structures, and showed reduced development of the branchial arches (Fig. 9 D). This observation suggested that ppm1f^−/− embryos die before or around E10.5 in utero. To unambiguously determine their genotype, we isolated primary murine embryonic fibroblasts (MEFs) from E10.5 embryos. PCR on genomic DNA of isolated fibroblasts confirmed that the small malformed embryos harbor two dysfunctional PPM1F alleles (ppm1f^−/−), while the normal-sized embryos were either homozygous for the WT allele or heterozygous (ppm1f^+/−; Fig. 9 E). Primary fibroblasts from ppm1f^−/− mouse embryos and WT littermates were immortalized by retroviral transduction with SV40 largeT antigen, and WB with polyclonal anti-murine PPM1F (mPPM1F) antibodies demonstrated the absence of the full-length enzyme or any truncated ppm1f gene products in ppm1f^−/− KO fibroblasts (MEF PPM1F^−/−; Fig. 9 F). Already during regular cell culture, MEF PPM1F^−/− showed enhanced cell attachment to the culture dish, and it took longer to detach these cells during passaging. When placed onto fibronectin, the primary ppm1f^−/− cells showed pronounced accumulation of active integrin β1 and talin at peripheral focal adhesion sites (Fig. 9 G). Most importantly, phosphorylation of the threonine motif in the integrin β1 cytoplasmic tail was strongly elevated in the MEF PPM1F^−/− cells compared with the WT cells (Fig. 9 H). The enhanced integrin phosphorylation in MEF PPM1F^−/− cells correlated with elevated levels of active integrin β1 (Fig. 9 I) and translated into increased adhesiveness on low (0.4 µg/ml) and high (10 µg/ml) concentrations of the integrin ligand fibronectin (Fig. 9 J). Furthermore, the enhanced integrin-mediated cell adhesion directly translated into reduced cell migration velocity and migration distance of MEF PPM1F^−/− cells (Fig. 9 K). These results highlight the severe consequences of PPM1F deletion on integrin function in primary cells. Together, our data further confirm a critical and nonredundant role of PPM1F as a negative regulator of integrin activity during embryonic development and during physiological integrin-dependent processes such as cell adhesion, spreading, and migration.

Dynamic modulation of integrin activity is indispensable for the normal functioning of animal cells, especially with regard to cell attachment, spreading, and migration (Morse et al., 2014; Shattil et al., 2010). Integrin activity is controlled by protein–protein interactions occurring at the cytoplasmic tail of the integrin β subunit, and the fine-tuning of cell adhesion requires regular transitions from talin/kindlin-bound active to filaminA-associated inactive integrin. Here we identify the molecular machinery that executes this integrin activity switch: the cytoplasmic metal-dependent protein phosphatase PPM1F. This serine/threonine phosphatase acts on a conserved threonine motif located in the cytoplasmic tail of most integrin β subunits and thereby controls filamin binding. Accordingly, this phosphatase represents the first enzyme known to directly regulate integrin activity.

PPM1F belongs to the PPMs, which, in contrast to the family of phospho-protein phosphatases, do not rely on regulatory subunits for subcellular localization and substrate recognition (Moorhead et al., 2009; Stern et al., 2007). Instead, these phosphatases harbor additional domains involved in protein–protein interaction and phosphatase activity regulation (Ishida et al., 2018). The tertiary structure of PPM1F is not known, but a PPM family member from Caenorhabditis elegans termed FEM-2 (sharing ∼25% sequence identity with human PPM1F) has been crystallized (Zhang et al., 2013). Though the N-terminal domain of FEM-2 exhibits a large intramolecular binding interface with the C-terminal catalytic domain, in vitro experiments did not indicate a role of the N terminus in regulating phosphatase activity, but rather in mediating protein–protein interactions (Zhang et al., 2013). In this regard, PPM1F is known to bind the focal adhesion protein β-PIX and has been identified by unbiased mass spectrometry approaches as a constituent of focal adhesions (Schiller et al., 2011). PPM1F also functions as a negative regulator of the β-PIX–associated protein serine/threonine kinase PAK and of CaMKII (Harvey et al., 2004; Ishida et al., 1998; Koh et al., 2002). In both cases, PPM1F shows preference for phospho-threonine (pT) residues located in the activation loops of these kinases (Ishida et al., 2008). Interestingly, the amino acid sequences surrounding these known PPM1F target motifs in PAK and CaMKII (K/Rx_(1–2)TxV) are highly reminiscent of the sequence (Kx₂TTV) encompassing T788/T789 in integrin β1. As such a motif occurs also in the cytoplasmic domains of human integrins β2, β3, β5, β6, and β7, it can be envisioned that this ubiquitously expressed threonine phosphatase controls a larger set of integrin heterodimers. It is important to note that PPM1F, but not the related PPM family member ILKAP, readily dephosphorylated the phospho-T788/T789 motif. In myoblasts, the phosphatase PP2A has been shown to associate with and dephosphorylate the T788/T789 motif on integrin β1 (Kim et al., 2004). Striated muscle cells require stable, long-term adhesion for their physiological function. Therefore, PPM1F-mediated negative regulation of integrin activity might not be required to the same extent in this tissue, where PP2A could be involved in integrin regulation. Though additional enzymes might be able to dephosphorylate phospho-T788/T789 of integrin β1, the prominent gain in integrin function in normal fibroblasts and glioma cells upon PPM1F depletion and the severe phenotype of homozygous deletion of PPM1F in the mouse strongly argue for a unique role of PPM1F. The idea that the integrin β1 subunit is a major target of PPM1F is supported by the constitutive T788/T789 phosphorylation, the enhanced talin recruitment, and the exaggerated integrin β1 activity seen in all examined PPM1F-deficient cells. The biochemical consequence of PPM1F action, namely filaminA binding to the T788/T789 motif, nicely explains how this phosphatase can have a direct impact on integrin activity.

The proposed phospho-switch mechanism in the integrin β1 subunit and the role of filaminA in this context is in line with previous reports on the integrin β7 and integrin β2 subunits (Kiema et al., 2006; Takala et al., 2008). Structural modeling clearly indicates that sterical hindrance and charge repulsion will prohibit filaminA from binding to the phosphorylated T788/T789 motif in integrin β1, and we experimentally confirm this binding pattern with pseudo-phosphorylated recombinant proteins as well as synthetic phospho-peptides. Intriguingly, our in vitro competition assays demonstrate that filaminA readily displaces the talin band 4.1, ezrin, radixin, moesin (FERM) domain as well as kindlin2 from the integrin β1 tail, but can do so only in the absence of T788/T789 phosphorylation (Fig. 10 A). Phosphorylation of this conserved motif and the resulting dissociation of filamin grants talin access to the membrane proximal NPxY motif (Fig. 10, B and C). The binding of talin to the phosphorylated integrin tail unclasps the integrin α and β subunits and leads to extension of the bend extracellular domains (Fig. 10 C). In this situation, talin binding is required before kindlin2 can associate with the membrane distal NPxY motif, as the T788/T789 motif is phosphorylated (Fig. 10 D). Upon kindlin2 binding, the active conformation of integrins is further stabilized, and integrin-mediated adhesion is further promoted by receptor clustering (Fig. 10 D). The phosphorylation-guided cooperation between talin and kindlin, as suggested by our in vitro and in vivo data, could be a means to safeguard against a kindlin2-mediated inside-out activation of integrins in the absence of prior talin binding. In retrospect, these findings now provide a mechanistic explanation for the previously observed adhesion phenotype of integrin β1 T789D and of integrin β1 TT788/789AA–expressing cells, which appear locked in either the active or inactive situation (Nilsson et al., 2006; Wennerberg et al., 1998). Our novel observations also nicely combine with previous results in a coherent picture of integrin activity regulation from within the cell (inside-out signaling). It has been noted before that overexpression of the isolated talin FERM domain can trigger inside-out signaling of integrin β1 and β3, while kindlin overexpression on its own is not sufficient, but rather intensifies talin-initiated integrin activity (Harburger et al., 2009; Li et al., 2017; Ma et al., 2008; Ye et al., 2010). On the other hand, kindlin critically contributes to integrin activity when cells respond to an integrin ligand and reorganize their cytoskeleton during outside-in signaling (Böttcher et al., 2017; Montanez et al., 2008; Theodosiou et al., 2016).

Interestingly, clustering of the WT integrin β1 chimera in intact cells, a situation mimicking the unclasped integrin, does not lead to filaminA recruitment. This finding could indicate that the threonine motif is mainly phosphorylated when the integrin β subunit is separated from the integrin α subunit. Intriguingly, upon overexpression of PPM1F, but not PPM1F D360A, filaminA accumulates at the WT integrin β1 cytoplasmic tail. This observation suggests that the activity of the overexpressed phosphatase can override a potential default phosphorylation of the threonine motif in the unclasped integrin β subunit to allow filaminA binding. It is interesting to speculate that a default phosphorylation of the conserved threonine motif in the isolated WT integrin β1 tail would not only promote displacement of the negative regulator filamin but also prohibit kindlin2 from driving integrin inside-out signaling in the absence of talin. This scenario is in line with the observation that kindlin overexpression does not lead to integrin inside-out activation (Ma et al., 2008; Harburger et al., 2009; Ye et al., 2010; Li et al., 2017).

A major remaining question is where and when the conserved threonine motif in integrin β subunits is phosphorylated and which kinase(s) are involved in this regulatory step. Phosphorylation of the conserved integrin threonine motif has been most intensely studied for the integrin β2 subunit, where T758/T759 form part of the Kx₂TTTV motif in the cytoplasmic tail. In this case, stimulation of G-protein–coupled receptors or the T cell receptor leads to phosphorylation of T758 in integrin β2 and enhanced integrin-mediated cell attachment (Chatila et al., 1989; Fagerholm et al., 2005; Takala et al., 2008; Uotila et al., 2014; Valmu et al., 1991). Application of PKC inhibitors abrogates T758 phosphorylation, and the corresponding synthetic peptides of the integrin cytoplasmic domain are phosphorylated in vitro by conventional and unconventional PKC enzymes (Fagerholm et al., 2002). However, additional kinases such as CaMKII have been shown to associate with the integrin β1 subunit in breast tumor cells (Takahashi, 2001), and inhibitors of CaMKII prevent the increase in T789 phosphorylation driven by constitutive active Ndr1 kinase, an abundant kinase in differentiating neurons (Rehberg et al., 2014). Our kinase assays with the purified integrin β1 cytoplasmic domain now confirm that CaMKII is a bona fide integrin kinase. These studies indicate that multiple serine/threonine kinases can relay signaling inputs, eventually originating from different extracellular and/or intracellular cues, toward the integrin β1 cytoplasmic domain. Our finding of a strong elevation of integrin phosphorylation upon deletion of a single phosphatase was therefore unexpected. However, in several distinct human cell types and in different scenarios, such as matrix-attached or suspended cells, PPM1F deficiency leads to constitutive T788/T789 phosphorylation. It has to be noted that apart from the integrin β1 subunit, PPM1F acts on additional substrates such as kinases, cytoskeletal proteins, and apoptosis regulators (Zhang et al., 2013; Ishida et al., 2018). Therefore, it remains to be determined whether the abortion of embryonic development seen in PPM1F KO mice is a direct consequence of alterations in integrin activity and to what extent deregulation of other PPM1F substrates may play a role. However, the early embryonic lethality observed upon disruption of genes encoding integrin β1, talin, kindlin, filaminA, and PPM1F in mammals and the functional interplay of these proteins in intact cells strongly argue for a critical role of PPM1F-mediated integrin activity regulation in vivo.

It is easily conceivable that PPM1F is ideally suited to serve as key control for the T788/T789 phospho-switch, as it dephosphorylates the integrin β cytoplasmic domain, and this phosphatase is also able to reverse the auto-phosphorylation of CaMKII at Thr286 (Harvey et al., 2004; Ishida et al., 1998). Thus, PPM1F could shift the balance toward the unphosphorylated, inactive integrin by acting on both an integrin-directed serine/threonine kinase and the integrin T788/T789 motif itself. Taken together, our study identifies PPM1F as the enigmatic integrin phosphatase that acts on the highly conserved threonine motif in the integrin β cytoplasmic domain (Gahmberg et al., 2014). Thereby, this widely expressed protein phosphatase functions as an essential constituent of the integrin off-switch. In contrast to other negative regulators of integrins, such as ICAP-1α, Sharpin, or Dok1, PPM1F has a defined enzymatic activity that can serve as a target for small molecule modulators. Given the important role of fine-tuning integrin activity in thrombus formation, immune cell motility, or wound healing, agonists as well as antagonists of PPM1F could provide novel access points to adjust integrin activation thresholds.

The following antibodies were used with the corresponding dilutions for WB, immunofluorescence (IF), immunohistochemistry, or integrin activity assay (IA): α-actinin (BM75.2; mouse anti-human/anti-mouse; Abcam; 1:1,000 WB), α₁-integrin (TS2/7; mouse anti-human/anti-mouse; Abcam; 1:50 IF), α₂-integrin (6F1; mouse anti-human/anti-mouse; DSHB; 1:60 IF), α₃-integrin (P1B5; mouse anti-human/anti-mouse; DSHB; 1:60 IF), α₅-integrin (BIIG2; rat anti-human; DSHB; 1:10 IF; and MFR5; rat anti-mouse; BD Pharmingen; 1:300 IF), α_v-integrin (PE-P2W7; mouse anti-human/anti-mouse; sc-9969; 1:300 IF), β₁-integrin (AIIB2; IgG₁, rat anti-human; DSHB; 1:600 IA; and HMβ1-1, armenian hamster anti-mouse; BioLegend; 1:300 IF/IA; and D2E5; rabbit anti-human; Cell Signaling; 1:1,000 WB), active β₁-integrin (9EG7; G₁, rat anti-human/anti-mouse; generous gift of D. Vestweber, Max-Planck-Institut for Molecular Medicine, Münster, Germany; 1:300 IA/IF), pT788/789 β₁-integrin (44-872G; rabbit anti-human; Thermo Fisher Scientific; 1:1,000 WB), β₃-integrin (2C9.G3; Armenian hamster anti-human/anti-mouse; eBioscience; 1:300 IF), β₅-integrin (AST-3T; mouse anti-human; BioLegend; 1:150 IF; and KN-52; mouse anti-mouse/human; eBioscience; 1:300 IF), carcinoembryonic antigen-related cell adhesion molecule (CEACAM) 1,3,4,5,6 (D14HD11; Aldevron; 1:6,000 WB, 1:200 IF), Ezrin (MAB3822; mouse anti-human; Millipore; 1:200 WB), FAK (77; mouse anti-human; BD Biosciences; 1:250 WB), ILK (EP1593Y; rabbit anti-human; Epitomics; 1:800 WB), Kindlin-2 (3A3; mouse anti-human; Millipore; 1:200 WB, 1:250 IF), paxillin (5H11; mouse monoclonal; Thermo Fisher Scientific; 1:1,000 WB), hPPM1F (17020–1-AP; rabbit anti-human; Protein-Tech; 1:1,000 WB), mPPM1F (1147; rabbit anti-mouse PPM1F; generated at the Tierforschungsanlage, University of Konstanz [see below]; 1:200 WB), filaminA (EP2405Y; IgG, rabbit anti-human; Epitomics; 1:125,000 WB; and PM6/317; mouse anti-human, Millipore; 1:100 IF), Tubulin (E7, IgG1, mouse anti-human, DSHB; 1:1000), Talin (8d4, mouse anti-human; Thermo Fisher Scientific; 1:800 WB, 1:40 IF), vinculin (hVIN-1; mouse anti-human; Sigma-Aldrich; 1:2,000 WB, 1:200 IF), Zyxin (Zol301; mouse anti-human; Abcam; 1:1,000 WB), GFP (Jl-8; Clontech; 1:3,000 WB), GST (B-14 sc-138; Santa Cruz; 1:1,000 WB, 1:500 ELISA), 6x His (H8; Thermo Fisher Scientific; 1:1,000 WB and ELISA), Dylight488-conjugated goat anti-mouse IgG (Jackson; 1:200), Cy3-conjugated goat anti-rabbit IgG (Jackson; 1:200), Cy3-conjugated goat anti-mouse IgG (Jackson; 1:200), Cy5-conjugated goat anti-mouse IgG (Jackson; 1:200), RhodamineRed-conjugated goat anti-rat IgG (Jackson; 1:200), RhodamineRed-conjugated goat anti-Armenian hamster IgG (Jackson; 1:200), Cy5-conjugated phalloidin (A22287; Molecular Probes; 1:100), HRP-conjugated goat anti-mouse IgG (Jackson; 1:10,000 WB), HRP-conjugated goat anti-rat IgG (Santa Cruz; 1:250), HRP-conjugated goat anti-rabbit IgG (Jackson; 1:3,000 WB), streptactin-HRP (IBA Lifesciences; 1:10,000 WB, 1:1,000 ELISA), murine endoglin (CD105; rat anti-mouse, MJ7/18; DSHB), and unspecific control IgG (anti-mouse 96/1; generated at the Tierforschungsanlage, University of Konstanz).

For the construction of His-Sumo–tagged talin F3 subdomain of the FERM domain (residues 309–411), the murine talin1 cDNA (kindly provided by R. Fässler, Max-Planck-Institut Biochemistry, Martinsried, Germany) was amplified with the following primers: mTalin1 F3-forward: 5′-GAC​GGT​CTC​AAG​GTG​GTG​TCT​CCT​TCT​TCC​TAG​TC-3′ and mTalin1 F3-reverse: 5′-GTC​TCT​AGA​TTA​CAG​CCC​AAA​ATG​GTC​CTT​GCT​G-3′. The resulting PCR fragment was cloned into pET24a His-Sumo bacterial expression vector (Andréasson et al., 2008) via BsaI and XhoI restriction sites. For construction of GFP-tagged talin, cDNA of human talin-1 (kindly provided by R. Fässler) was cloned into pDNR Dual (Clontech) using the following primers and restriction digest with SalI/AgeI: hTalin1_SalI_sense: 5′-GAA​GTT​ATC​AGT​CGA​CAC​CAT​GGT​TGC​ACT​TTC​ACT​GAA​G-3′ and hTalin1_AgeI_anti: 5′-GTC​ATA​CCG​GTT​TAG​TGC​TCA​TCT​CGA​AGC​TCT​G-3′. Cre/LoxP recombination was used to move sequences from pDNR Dual into pEGFP-C1-loxP for eukaryotic expression.

Human kindlin2 cDNA (ID4547604 in pOTB7; Source BioScience) was used as a template for PCR amplification with the following primers: hKindlin-2_SalI_sense: 5′-ATC​AGT​CGA​CGC​TCT​GGA​CGG​GAT​AAG​GAT​G-3′ and hKindlin-2_BamHI_anti: 5′-TAC​CGG​ATC​CTC​ACA​CCC​AAC​CAC​TGG​TAA​G-3′. The product cloned into pDNR Dual via SalI/BamHI restriction digest. The kindlin2 cDNA was transferred to pEGFP-C1-loxP vector by Cre-recombination. Mouse kindlin2 cDNA (pCMV-SPORT6; Source Bioscience) was used as a template for PCR amplification with the following primers: mKindlin2_LIC_sense: 5′-ACT​CCT​CCC​CCG​CCA​TGG​CTC​TGG​ACG​GGA​TAA​GGA​TG-3′ and mKindlin2_LIC_anti: 5′-CCC​CAC​TAA​CCC​GTC​ACA​CCC​AAC​CAC​TGG​TGA​G-3′. The product was cloned into pDNR Dual-LIC vector by ligation-independent cloning (LIC; Adrian et al., 2019). The mKindlin-2 insert was cloned into pET24a-His-SUMO by PCR amplification with the following primers: mKindlin-2_BamHI_sense: 5′-TAT​AGG​ATC​CCA​TGG​CTC​TGG​ACG​GGA​TAA​G-3′ and mKindlin-2_XhoI_anti: 5′-TAT​ACT​CGA​GTC​ACA​CCC​AAC​CAC​TGG​TGA​G-3′ and by restriction digest with BamHI and XhoI enzymes.

The integrin-binding part of filaminA (filaminA Ig19-21 [FLNIg19-21]) was generated by amplifying human filaminA (plasmid 8982; Addgene) with the following primers: FLNIg19-forward 5′-ACT​CCT​CCC​CCG​CCA​TGA​TCA​GCC​AGT​CGG​AAA​TTG-3′ and FLNIg21-reverse 5′-CCC​CAC​TAA​CCC​GAG​CCA​CAG​GCA​CCA​CGA​AG–3′. It was cloned into pDNR-Dual-LIC. FLNIg19-21 cDNA was subsequently transferred to pEGFP-C1-loxP by Cre-mediated recombination, generating GFP-FLNIg19-21 for mammalian expression, and to pGEX4T1 loxP (Schmitter et al., 2007) for bacterial expression of GST-FLNIg19-21.

For His-Sumo–tagged FLNIg19-21, the respective filaminA sequence was amplified by PCR with the primer pair FLNIg19-forward: 5′-GAC​GGT​CTC​AGG​TGG​GGA​TGC​CAG​TCG​TGT​TCG​CGT​CTC​TG-3′ and FLNIg21-reverse: 5′-CTG​ATG​CTC​GAG​TTA​AGA​CGG​AGA​AGC​CAC​AG-3′. The resulting PCR fragment was cloned into pET24aHis-Sumo via BsaI and XhoI restriction sites.

cDNA of control protein yeast enolase-1 was obtained by genomic DNA extraction (PureLink Genomic DNA Mini Kit; Thermo Fisher Scientific) from Candida albicans SC5314 (generous gift of J. Morschhäuser, University of Würzburg, Würzburg, Germany) and PCR amplification using the following primers: enolase-1 BamHI sense: 5′-ATA​GGA​TCC​GAT​GTC​TTA​CGC​CAC​TAA​AAT​CC-3′ and enolase-1 XhoI NEF anti: 5′-ATA​CTC​GAG​TTA​TTT​AAA​GTA​TTC​GGG​CCC​CAA​TTG​AGA​AGC​CTT​TTG​G-3′. The resulting PCR fragment was cloned into pET24aHis-Sumo via BamHI and XhoI restriction sites.

For the production of recombinant GST-tagged cytoplasmic tail of integrin β1, the human integrin β1 cDNA (plasmid 54129; Addgene) was amplified by PCR with the following primer pair: hβ1Integrin-cyto-EcoRI-sense: 5′-ATA​GAA​TTC​TGG​AAG​CTT​TTA​ATG​ATA​ATT​C-3′ and hβ1Integrin-cyto-XhoI-anti: 5′-ATA​CTC​GAG​TCA​TTT​TCC​CTC​ATA​CTT​CG-3′. The cDNAs of human integrin β1 T788A/T789A, T788D, and T788D/T789D cytoplasmic domains were custom synthesized (Eurofins Genomics GmbH) and amplified using the same primer pair as for WT β1 integrin. PCR products of integrin β1 WT and T788A/T789A were cloned into pGEX-4T1 via EcoRI/XhoI restriction sites to yield an N-terminal GST-tagged fusion protein, respectively. The Y795A mutation was inserted by using integrin β1 WT cDNA as a template and the following primers: integrin β1 Y795A BamHI sense: 5′-GCT​ATG​GAT​CCC​ATG​ACA​GAA​GGG​AGT​TTG-3′ and integrin β1 Y795A XhoI anti: 5′-ATA​TCT​CGA​GTC​ATT​TTC​CCT​CCG​CCT​TCG​GAT​TG-3′.

The Strep-tag vector for bacterial expression of Strep-tagged integrin cytoplasmic domains was produced by custom gene synthesis (pEX-A128 Strep-integrin β; Eurofins Genomics GmbH) according to Pfaff et al. (1998). Briefly, the E. coli codon-optimized cDNA encompasses an N-terminal Twin-StrepII-tag and a coiled-coil domain encoding a heptad-repeat coiled-coil–forming region followed by in-frame BamHI/XhoI restriction sites for insertion of integrin cytoplasmic tails. Using NcoI/XhoI restriction sites, the Strep-tag and coiled-coil sequence of pEX-A128 Strep-integrin β was ligated into the pET28a(+) vector (Novagen; Merck Millipore). Strep-tagged integrin β1 WT, T788D, T788D/T789D, T788A/T789A, and Y795A constructs were produced by BamHI/XhoI restriction digest of corresponding pGEX-4T1 vectors and subcloning of the integrin β1 tail cDNA into the Strep-tag vector. For the generation of Strep-tagged integrin β1 Y783A, Strep-integrin β1 WT cDNA containing vector was used as a template for site-directed mutagenesis with the following primers: integrin β1 Y783A sense: 5′-CAC​GGG​TGA​AAA​TCC​TAT​TGC​TAA​GAG​TGC​CGT​AAC​AAC​TG-3′ and integrin β1 Y783A anti: 5′-CAC​AGT​TGT​TAC​GGC​ACT​CTT​AGC​AAT​AGG​ATT​TTC​ACC​CGT​GTC​CC-3′. To generate Strep-tagged β1 integrin Y783A + T788D/T789D, Strep-tagged integrin β1 T788D/T789D vector was used as a template for site-directed mutagenesis with the following primers: integrin β1 Y783A DD sense: 5′-GGT​GAA​AAT​CCT​ATT​GCT​AAG​AGT​GCC​GTA​GAC-3′ and integrin β1 Y783A DD anti: 5′-GTC​TAC​GGC​ACT​CTT​AGC​AAT​AGG​ATT​TTC​ACC-3′.

CEACAM3-integrin β1 fusion constructs were generated via PCR amplification using the cDNA of human integrin β1 WT, T788/789D, or T788/789A synthetic sequences as a template with following primers: integrin β1_E762-K798_BamHI sense: 5′-GCG​GCT​ATG​GAT​CCG​AGT​TTG​CTA​AAT​TTG​AAA​AGG​AGA​AAA​TG-3′ and integrin β1_E762-K798_XhoI anti: 5′-TAT​CTC​GAG​TCA​TTT​TCC​CTC​ATA​CTT​CGG-3′. PCR products were cloned into pcDNA3.1 CEACAM3ΔCT using BamHI/XhoI restriction sites (Baade et al., 2019; Schmitter et al., 2004). For the β1 integrin Y783A construct, pcDNA3.1 CEA3-β1 WT integrin-containing plasmid was used to perform site-directed mutagenesis using the following primers: integrin β1 Y783A sense: 5′-CAC​GGG​TGA​AAA​TCC​TAT​TGC​TAA​GAG​TGC​CGT​AAC​AAC​TG-3′ and integrin β1 Y783A anti: 5′-CAC​AGT​TGT​TAC​GGC​ACT​CTT​AGC​AAT​AGG​ATT​TTC​ACC​CGT​GTC​CC-3′. CEACAM3-integrin β1 Y783A + T788D/T789D fusion construct was generated via PCR amplification using Strep-tagged β1 integrin Y783A + T788D/T789D as a template with the following primers: integrin β1_E762-K798_BamHI sense: 5′-GCG​GCT​ATG​GAT​CCG​AGT​TTG​CTA​AAT​TTG​AAA​AGG​AGA​AAA​TG-3′ and integrin β1_E762-K798_XhoI anti: 5′-TAT​CTC​GAG​TCA​TTT​TCC​CTC​ATA​CTT​CGG-3′. The PCR product was cloned into pcDNA3.1 CEACAM3ΔCT using BamHI/XhoI restriction sites (Baade et al., 2019; Schmitter et al., 2004).

For the generation of mCherry- and GST-tagged PPM1F, the cDNA of human PPM1F (I.M.A.G.E. cDNA clone IRAUp969F10111D) was obtained from Source BioScience. The following primers were used for amplification: hPPM1F-IF-sense: 5′-GAA​GTT​ATC​AGT​CGA​CAC​CAT​GTC​CTC​TGG​AGC​CCC-3′ and hPPM1F-IF-anti: 5′-ATG​GTC​TAG​AAA​GCT​TGC​CTA​GCT​TCT​TGG​TGG​AGC-3′. The resulting PCR fragment was cloned into pDNR-CMV using the In-Fusion dry-down PCR Cloning Kit (Clontech). The sequence-verified pDNR-CMV hPPM1F was used as donor vector, and the insert was transferred by Cre-mediated recombination into the acceptor vector pmCherry-C1-loxP (mammalian expression) and pGEX-4T1-loxP (bacterial expression) generating N-terminal–tagged pmCherry-hPPM1F and pGEX GST-hPPM1F. For the generation of mCherry- or GST-tagged hPPM1F D360A, the phosphatase dead mutant of PPM1F (Harvey et al., 2004), site-directed mutagenesis was performed with the hPPM1F-containing vectors using primers PPM1F-D360A-forward: 5′-GAC​TAC​CTG​CTG​CTA​GCC​TGT​GCT​GGC​TTC​TTT​GAC​GTC​G-3′ and PPM1F-D360A-reverse: 5′-GTC​AAA​GAA​GCC​AGC​ACA​GGC​TAG​CAG​CAG​GTA​GTC​CTC-3′.

For expression of PTP1B (a kind gift from W. Hofer, University of Konstanz, Konstanz, Germany), the cDNA was cloned into pDNR-Dual-LIC after PCR amplification using the primers PTP1B LIC sense: 5′-ACT​CCT​CCC​CCG​CCA​TGG​AGA​TGG​AAA​AGG​AGT​TCG-3′ and PTP1B LIC anti: 5′-CCC​CAC​TAA​CCC​GCT​ATG​TGT​TGC​TGT​TGA​ACA​GG-3′, followed by Cre-lox recombination into pGEX4T1 loxP vector for bacterial expression. All recombinant constructs were verified by sequencing.

Strep-tagged or GST-tagged integrin β1 cytoplasmic domains and 6xHis-SUMO-tagged or GST-tagged TalinF3, Kindlin-2, FLNIg19-21, enolase-1, human 7xHis-TEV ILKAP (plasmid 34817; Addgene; pQTEV vector by K. Buessow, Protein Structure Factory, Berlin, Germany), GST-tagged human PTP1B and human Trx-His-S-PP5 (a kind gift of D. Dietrich, University of Konstanz, Konstanz, Germany), and GST-tagged fibronectin type III repeats 9–11 (FN_III9-11; gift of M.A. Schwartz, Yale University, New Haven, CT) were expressed in E. coli as described for PPM1F. Proteins were purified using His- or GST-Trap FF column (GE Healthcare) or Streptactin Superflow column (IBA Lifesciences) depending on the protein and dialyzed against appropriate buffers.

2.5 µg of Strep-tagged integrins or 10 µg of biotin-integrin phospho-peptide β1-762-798 pTpT (Biotin-EFAKFEKEKMNAKWDTGENPIYKSAV[pT][pT]VVNPKYEGK-OH) or biotin-integrin peptide β1-762-798 (Biotin-EFAKFEKEKMNAKWDTGENPIYKSAVTTVVNPKYEGK-OH; Novopep Limited) were loaded onto Strep-Tactin Sepharose beads (50% suspension; IBA Lifesciences) or streptavidin agarose beads (50% suspension; 16-126; Merck) in immunoprecipitation (IP) buffer (50 mM Tris, pH 7.5, 100 mM NaCl, 10% glycerol, and 0.05% Tween) with freshly added Na₃VO₄ (200 mM) and Na₄P₂O₇ (200 mM) for 30 min at RT under continuous rotation. After centrifugation (2,700 g, 2 min, 4°C), samples were washed three times with IP buffer. Then integrin-loaded beads were suspended in bait protein solution (corresponding amount of protein diluted in IP buffer) and incubated 2 h at 4°C under constant rotation. Samples were centrifuged (2,700 g, 2 min, 4°C) and washed again three times with IP buffer. Strep-Tactin samples were eluted under native conditions by adding 30 µl of buffer BXT (50 mM Tris, pH 8, 150 mM NaCl, 50 mM biotin; IBA Lifesciences). After 10 min incubation at RT under constant rotation, samples were centrifuged. Supernatants were mixed with 4× SDS and boiled for 5 min at 95°C before they were subjected to WB. Streptavidin agarose beads were directly mixed with 2× SDS and boiled for 10 min at 95°C to elute proteins from biotin-integrin peptides before they were subjected to WB.

96 wells (high-binding; 655061; Greiner Bio-One) were coated with 1 µM of His-Sumo–tagged or GST-tagged proteins in PBS (1.37 M NaCl, 26.8 mM KCl, 14.7 mM KH₂PO₄, and 78.1 mM Na₂PO₄) overnight at 4°C in triplicate. Wells were washed three times with PBS and blocked with 200 µl/well PBS plus 2% BSA for 1 h at RT before incubation with 0.5 µM of Strep-tagged integrin β1 proteins in PBS plus 0.05% Tween overnight at 4°C. For competitive assays with FLN19-21 and TalinF3 domain, Strep-tagged integrin β1 cytoplasmic domains were immobilized first before incubation with FLN19-21 and/or TalinF3. Wells were washed three times with PBS plus 0.05% Tween and blocked again for 1 h at RT. Streptactin-HRP (IBA Lifesciences) or mouse-α-His antibodies followed by HRP-conjugated goat α-mouse IgG antibodies for competitive assays were added in blocking buffer for 1 h at RT, respectively. Finally, wells were intensively washed with PBS plus 0.05% Tween, and 100 µl/well of tetramethylbenzidine (TMB) solution was added (9.5 ml of 2.4 mg/ml TMB in 1:9 aceton:ethanol plus 20 µl/10 ml H₂O₂ plus 0.5 ml 30mM potassium citrate, pH 4.1). The reaction was stopped after 20 min by adding 100 µl/well 2 M H₂SO₄, and samples were measured at 450 nm in a microplate reader (Varioscan; Thermo Fisher Scientific).

Human embryonic kidney 293T cells (293T; American Type Culture Collection CRL-3216) and A172 glioblastoma cells (American Type Culture Collection CRL-1620) were grown in DMEM supplemented with 10% calf serum. NHDFs were obtained from PromoCell and cultured in PromoCell fibroblast growth medium. All cells were maintained at 37°C, 5% CO₂, and subcultured every 2–3 d.

For transient transfection of 293T cells, cells were seeded at 25% confluence the day before and transfected using the standard calcium phosphate method with a total amount of 5 µg plasmid DNA/dish. For transient transfection of A172 cells, Lipofectamine 3000 was used (Thermo Fisher Scientific) according to the provided protocol after seeding 2.5 × 10⁵ cells for 2 h into six-well plates.

The detailed OPTIC protocol, fluorescence microscopy, and evaluation of protein recruitment were described elsewhere (Baade et al., 2019). Briefly, HEK293T cells were transfected with pcDNA3.1 CEACAM3-integrin β1 fusion constructs together with GFP, GFP-FLNIg19-21, GFP-talin-1, or GFP-kindlin-2 and optionally mCherry-PPM1F or mCherry-PPM1FD360A. 48 h after transfection, cells were seeded onto 10 µg/ml poly-L-lysine–coated coverslips in suspension medium (DMEM plus 0.25% BSA). After 2 h, adherent cells were infected with Pacific Blue–stained Ngo (Opa₅₂-expressing, nonpiliated Neisseria gonorrhoeae MS11; Baade et al., 2019) at MOI 20 for 1 h in DMEM plus 0.25% BSA. Infection medium was aspirated, and cells were immediately fixed with 4% PFA for 15 min, permeabilized with 0.1% Triton X-100 in PBS for 5 min, and stained with α-CEACAM antibody (clone D14HD11; Aldevron) in blocking solution (10% calf serum [CS] in PBS) after washing with PBS and blocking. Wells were again washed with PBS and blocked followed by a second antibody staining. Coverslips were mounted on glass slides using Dako mounting medium (Dako). Samples were imaged at RT on a Leica SP5 confocal microscope equipped with a 63.0×/1.40 NA oil HCX PL APO CS UV objective and analyzed using LAS AF Lite software. Protein recruitment at one bacterial infection site was quantified per cell (n = 60 cells per sample).

To obtain WCLs, equal cell numbers were lysed by treatment with radioimmunoprecipitation assay buffer (1% Triton X-100, 50 mM Hepes, 150 mM NaCl, 10% glycerol, 1.5 mM MgCl₂, 1 mM EGTA, 0.1% wt/vol SDS, and 1% vol/vol deoxycholic acid) supplemented with freshly added protease and phosphatase inhibitors (10 mM sodium pyrophosphate, 100 mM NaF, 1 mM sodium orthovanadate, 5 µg/ml leupeptin, 10 µg/ml aprotinin, 10 µg/ml Pefabloc, 5 µg/ml pepstatin, and 10 µM benzamidine) and phosphatase saturating substrate (para-nitrophenolphosphate [pNPP]; Sigma-Aldrich; 10 mM). Chromosomal DNA and cell debris were pelleted by addition of Sepharose beads and centrifugation (13,000 rpm, 30 min, 4°C). Supernatant was supplemented with 2× or 4× SDS sample buffer (2 or 4% wt/vol SDS, 20% wt/vol glycerol, 125 mM Tris-HCl, 10/20% vol/vol β-mercaptoethanol, and 1% wt/vol Bromophenol blue, pH 6.8) and boiled for 5 min at 95°C. The protein amount was adjusted via the bicinchoninic acid protein assay kit (Pierce; Thermo Fisher Scientific) according to the manufacturer’s protocol and analyzed by WB using a prestained marker as protein size control (26619; Thermo Fisher Scientific). Briefly, proteins were resolved on 8–18% SDS-PAGE. After separation, the proteins were transferred to a polyvinylidene fluoride membrane (Merck Millipore), followed by blocking in 2% BSA containing 50 mM Tris-HCl, 150 mM NaCl, and 0.05% Tween 20, pH 7.5 (TBS-T) buffer. The membrane was incubated with primary antibody in blocking buffer overnight at 4°C, washed three times with TBS-T, and incubated with HRP-conjugated secondary antibody in TBS-T for 1 h at RT. The chemiluminescent signal of each blot was detected with ECL substrate (Thermo Fisher Scientific) on the Chemidoc Touch Imaging System (Bio-Rad) in signal accumulation mode. Acquired images were processed in Adobe Photoshop CS4 by adjusting illumination levels of the whole image.

For the generation of recombinant, shRNA-expressing lentiviral particles, the shRNA vector system pLKO.1 developed by Stewart et al. (2003) was applied. The different shRNAs were designed by using the AAN19 algorithm and siRNA selection program of the Whitehead Institute for Biomedical Research (http://sirna.wi.mit.edu/). According to the prediction of the siRNA selection program, two complementary oligos were synthesized. The sequences of all shRNA-oligos used in this study are provided in Table 1. The oligos were annealed and cloned via AgeI and EcoRI restriction sites into plasmid pLKO.1 puro (plasmid 8453; Addgene), which provides puromycin resistance for selection of stable knock-down cells. The correct insertion of the shRNA cassettes was verified by sequencing.

sgRNAs against Cerulean and human PPM1F were designed according to an appropriate gRNA target site selected by CHOPCHOP (http://chopchop.cbu.uib.no/; Montague et al., 2014). The following sgRNA encoding complementary oligos were synthesized and annealed: hPPM1F-sgRNA_sense: 5′-CAC​CGG​CAC​CGA​CCA​GAT​GTT​TCT​C-3′ and hPPM1F-sgRNA_anti: 5′-AAA​CGA​GAA​ACA​TCT​GGT​CGG​TGC​C-3′ as well as Cerulean-sgRNA_sense: 5′-CAC​CGC​CGT​CCA​GCT​CGA​CCA​GGA-3′ and Cerulean-sgRNA_anti: 5′-AAA​CTC​CTG​GTC​GAG​CTG​GAC​GGC-3′. The resulting sgRNA-Cerulean oligo was inserted into the BbsI restriction site of vector pSpCas9(BB)-2A-GFP (plasmid 48138; Addgene; a gift from F. Zhang, Broad Institute, Cambridge, MA). Using the primer pair CRISPR-U6gRNA_KpnI_sense: 5′-ATA​GGT​ACC​GTG​AGG​GCC​TAT​TTC​CC-3′ and CRISPR-U6gRNA_KpnI_anti: 5′-ATA​GGT​ACC​GTC​TGC​AGA​ATT​GGC​GC-3′, the complete U6-promoter–driven sgRNA-Cerulean–containing cassette was amplified and inserted into the KpnI site of plentiCRISPR v2 vector (plasmid 52961; Addgene; a gift from F. Zhang), resulting in plentiCRISPR v2 sgRNA-Cerulean. Using the BsmBI/Esp3I restriction sites, the sgRNA-PPM1F oligo was inserted in plentiCRISPR v2 sgRNA-Cerulean, resulting in plentiCRISPR v2 sgRNA-Cerulean-sgRNA-PPM1F. The correct insertion of the Cerulean sgRNA cassette and of the PPM1F sgRNA was verified by sequencing.

Lentiviral particles were produced as described previously (Muenzner et al., 2010). Briefly, 293T cells were transfected by standard calcium phosphate coprecipitation using 3.5 µg pMD2.G (packaging cassette), 5 µg psPAX2 (viral envelope expression cassette), and 6.5 µg pLKO.1, plentiCRISPR v2, or pLL3.7 containing the desired shRNA, sgRNA, or cDNA, respectively. After 72 h, virus-containing culture supernatant was collected and ultra-centrifuged, and target cells were infected with virus concentrate by spinfection (1 h, 800 g, RT) with 8 µg/ml polybrene following incubation for 24 h at 37°C. Control cells were generated by transducing cells with virus harboring empty pLKO.1, pLL3.7-LIC-mKate, or pLentiCRISPRv2 containing sgRNA against Cerulean. After 48 h recovery time, cells transduced with pLKO.1 or pLentiCRISPRv2 were selected with puromycin (0.8 µg/ml, 6 d). A172 cells transduced with pLL3.7 mCerluean or pLL3.7-LIC-mKate–derived vectors were selected for fluorescence protein expression by flow cytometry. Single cell clones were expanded by adding 20% conditioned medium to the regular growth medium supplied with 20% FCS and penicillin/streptomycin.

For the generation of A172 PPM1F KO cells, A172 cells were first stably transduced with a lentiviral vector encoding mCerulean cDNA. Using AgeI/BsrGI restriction sites, the GFP cDNA in vector pLL3.7 (plasmid 11795; Addgene) was exchanged for mCerulean cDNA derived from plasmid pmCerulean-C1 (kind gift of D. Piston, Vanderbilt University Medical Center, Nashville, TN). A172 cells stably expressing mCerulean were used as a basis for CRISPR/Cas9-mediated disruption of the Cerulean cDNA (as a sign for successful Cas9 expression and activity) and of the PPM1F gene. Accordingly, the cells were transduced with plentiCRISPR v2 sgRNA-Cerulean (A172 control cells) or plentiCRISPR v2 sgRNA-Cerulean-sgRNA-PPM1F. After puromycin selection, single Cerulean-negative cells were sorted by flow cytometry, and clonal cell lines were grown up using addition of 20% conditioned medium to the regular growth medium supplied with 20% FCS and penicillin/streptomycin. The derived cells lack Cerulean expression (A172 control) or lack Cerulean and PPM1F expression (A172 PPM1F KO cells) as verified by WB and sequencing of the Cas9-disrupted genomic Ppm1f locus.

For the reexpression of mKate2-PPM1F and mKate2-PPM1F D360A in A172 PPM1F KO cells, the pLL3.7 LIC mKate2 vector was created. First, mKate2 cDNA (pmKate2-c vector; plasmid FP181; Evrogen) was amplified using the primers pLL3.7_mKate2_AgeI_sense: 5′-ACT​ACC​GGT​CAT​GGT​GAG​CGA​GCT​GAT​TAA​G-3′ and pLL3.7_mKate2_EcoRI_anti: 5′-GTC​GAA​TTC​CTA​TCT​GTG​CCC​CAG​TTT​GCT​AGG-3′. The resulting mKate-encoding PCR fragment was cloned into pLL3.7 (plasmid 11795; Addgene) via AgeI/EcoRI restriction digest to replace the GFP cDNA to yield the pLL3.7-mKate2 vector. Subsequently, pLL3.7-LIC-mKate2 vector was created by replacing the multiple cloning site with a LIC sequence (primers 5′-CTA​GCG​ACT​CTC​CCC​CGG​GTT​AGT​GGG​GCA-3′ and 5′-CCG​GTG​CCC​CAC​TAA​CCC​GGG​GGA​GAG​TCG-3′) via NheI/AgeI. Then, hPPM1F and hPPM1F D360A cDNA in pmCherry were modified by site-directed mutagenesis with primers Rescue_sgRNA-hPPM1F_for: 5′-CGA​ACA​GAT​CAA​ATG​TTC​CTG​AGG​AAA​GCC​AAG​CGA​GAG​CG-3′ and Rescue_sgRNA-hPPM1F_rev: 5′-CCT​CAG​GAA​CAT​TTG​ATC​TGT​TCG​CCG​GAA​GGC​TTC​TCT​GAG​GG-3′ to inactivate the sgRNA-hPPM1F target sequence by silent mutations. The resulting sgRNA-resistant cDNAs of hPPM1F and hPPM1F D360A were amplified with primers PPM1F_LIC_for: 5′-CGA​ACA​GAT​CAA​ATG​TTC​CTG​AGG​AAA​GCC​AAG​CGA​GAG​CG-3′ and PPM1F_LIC_rev: 5′-CCT​CAG​GAA​CAT​TTG​ATC​TGT​TCG​CCG​GAA​GGC​TTC​TCT​GAG​GG-3′ and inserted via LIC (Aslanidis and de Jong, 1990) into pLL3.7-LIC-mKate2 to result in pLL3.7 mKate2-hPPM1F and pLL3.7 mKate2-hPPM1F D360A.

Sterile coverslips were coated with 0.4–10 µg/ml FN_III9-11 in PBS overnight at 4°C in a 24-well plate, and cells were starved with DMEM plus 0.5% FCS for 15 h. The next day, coating solution was removed, and wells were blocked with suspension medium (DMEM plus 0.25% BSA). In parallel, cells were trypsinized, trypsin-inactivated with soybean trypsin inhibitor (12.5 mg in 50 ml DMEM, sterile filtered; AppliChem), counted, and kept in suspension medium for 45 min at 37°C. 2.5 × 10⁴ cells were seeded onto coverslips and allowed to adhere for corresponding time periods. Cells were fixed with 4% PFA supplemented with 0.1% Triton X-100 for 5 min at RT and again without Triton X-100 for 20 min. Coverslips were washed thrice with PBS²⁺ (0.9 mM CaCl₂ and 0.5 mM MgCl₂ in 1× PBS) and blocked for 20 min with blocking solution (10% CS in PBS). Primary antibody solution was added for 1 h at RT. After washing thrice with PBS²⁺ and blocking for another 20 min, coverslips were treated with secondary antibody in blocking solution and optionally with phalloidin-Cy5 or DAPI for 1 h at RT in the dark. Finally, cells were washed thrice with PBS²⁺ and mounted with Dako fluorescent mounting medium (Dako). Samples were imaged on a Leica SP5 confocal microscope equipped with a 63.0×/1.40 NA oil HCX PL APO CS UV objective and acquired in xyz mode with a 1,024 × 1,024 pixel format and 100 Hz scanning speed at 8-bit resolution. All images were analyzed in ImageJ Software. For spreading assays, a macro was set up together with the Bioimaging Center at the University of Konstanz and used for quantitative picture analysis. Unrecognized cells were analyzed manually in the Leica LAS AF Lite software.

Cells were detached with trypsin/EDTA, pelleted, and suspended in FACS buffer (0.1% NaAzide and 5% FCS in PBS). After centrifugation (800 rpm, 3 min), cells were washed with FACS buffer, and 3 × 10⁵ cells were transferred into Eppendorf tubes, centrifuged (2,500 rpm, 2 min, 4°C), and incubated with primary antibody in FACS buffer at the desired concentration for 1 h at 4°C under constant rotation. Cells were washed thrice with FACS buffer and incubated with secondary antibody in FACS buffer at desired concentrations for 30 min at 4°C under constant rotation in the dark. After washing thrice with FACS buffer, cells were suspended in 1 ml of FACS buffer containing 2 mM EDTA. Finally, cells were analyzed by flow cytometry (BD LSRII, FACSDiva software; BD Biosciences).

10-cm dishes were coated with 2 µg/ml FN_III9-11 in PBS overnight at 4°C. Cells were starved with DMEM plus 0.5% FCS overnight at 37°C. The next day, dishes were blocked with DMEM plus 0.25% BSA for 1 h at 37°C. In parallel, cells were trypsinized, trypsin-inactivated with soybean inhibitor, counted, and kept in suspension medium for 45 min at 37°C. Afterward, the same cell numbers were seeded onto coated dishes and allowed to adhere for corresponding time periods at 37°C. Dishes were washed with PBS. WCLs were prepared and subjected to WB analysis. Phosphorylation of proteins was detected by phospho-specific antibodies. Densitometric analysis was performed by ImageJ software. The amount of phosphorylated β1 integrin was normalized to total β1 integrin expression levels.

96-well plates were coated with PBS containing corresponding concentrations of collagen I (Sigma-Aldrich), FN_III9-11, or poly-L lysine (SERVA) as an integrin-independent control overnight at 4°C. Wells were blocked with DMEM plus 0.25% BSA for 1 h at 37°C. In parallel, cells were trypsinized and kept in suspension medium for 45 min. 2 × 10⁴ cells/well were seeded and allowed to adhere for the indicated time periods at 37°C. After incubation, nonadherent cells were removed by gently washing with PBS²⁺ thrice. Adherent cells were fixed with 4% PFA in PBS for 15 min, washed with PBS, and stained with 0.1% crystal violet in 0.2 M borate buffer (pH 8.5) for 30 min. After intense washing, the color was unhinged from cells with 10 mM acetic acid, and the absorption was measured at 590 nm using a spectrophotometer.

For ELISA, 96-well plates were coated with 0.1 µg/ml FN_III9-11 in PBS overnight at 4°C. The next day, wells were blocked with 0.25% BSA in DMEM. In parallel, cells were trypsinized and kept in suspension for 1 h. Then, 5 × 10⁴ cells were seeded in triplicate and allowed to adhere for 40 min. Then cells were transferred onto ice, fixed with 4% PFA, washed with PBS, and permeabilized with 0.15% Triton X-100 for 15 min. Afterward, cells were blocked with 2% BSA in PBS for 20 min and stained with 9EG7 or AIIB2 antibodies. After incubation with the primary antibody, cells were washed with PBS, blocked for 20 min, and incubated with the secondary HRP-conjugated goat anti-rat IgG antibody for 1 h at RT. Finally, cells were intensively washed, and 100 µl of substrate solution was added (10 ml of 2.4 mg/ml TMB in 10% acetone/90% ethanol with 0.5 ml of 30 mM potassium citrate, pH 4.1). The enzymatic color reaction was stopped by adding 100 µl/well 2 M H₂SO₄, and the absorption was detected via spectrophotometric measurement at 450 nm. Controls were stained with secondary antibody only.

For FACS, starved cells were trypsinized and kept in suspension (2% BSA and 5 mM glucose in PBS) for 45 min before they were stimulated with 10 µg/ml FN_III9-12 for 15 min at 37°C or kept unstimulated by adding double-distilled H₂O. Cells were put on ice and split into two fractions, which were stained for either active integrin β1 (9EG7; 1:600) or total integrin β1 (HMβ1-1; 1:300, or AIIB2; 1:600) for 1 h on ice in PBS plus 2% BSA. Cells were washed thrice with PBS and incubated with Rhodamine-Red conjugated secondary antibody for 45 min on ice in the dark. Cells were washed, and fluorescence intensity was measured by flow cytometry (BD LSRII, FACSDiva software; BD Biosciences).

293T cells were transfected by standard calcium phosphate coprecipitation using plasmid DNA encoding for either GST-hPPM1F or GST-hPPM1F D360A. 48 h after transfection, cells were lysed in lysis buffer (50 mM Tris, pH 8, 1% Triton X-100, 1 mM EDTA, and 0.1% β-mercaptoethanol). Cleared lysates were incubated with glutathione-Sepharose beads for 3 h under constant rotation at 4°C. Beads were pelleted via centrifugation and washed three times in lysis buffer and once in GST buffer (50 mM Tris-HCl, pH 8, 150 mM NaCl, 1 mM DTT, and 5 mM MgCl₂). For the elution of the GST-fusion proteins, beads were incubated in GST buffer supplemented with 10 mM reduced L-glutathione twice for 20 min at 4°C under constant rotation. Aliquots were transferred to liquid nitrogen, and long-term storage occurred at −80°C.

For the production of recombinant GST-tagged hPPM1F or GST-hPPM1F D360A, the corresponding sequence-verified construct was transformed into competent E. coli BL21 Rosetta (DE3). Bacteria were cultured in lysogeny broth medium containing appropriate antibiotics at 37°C with constant shaking at 200 rpm. Expression was induced at OD₅₈₀ = 0.6 with 0.5 mM IPTG at 30°C and 220 rpm for 3 h, and bacteria were pelleted by centrifugation at 4,700 rpm, 20 min, RT. Afterward, bacteria were resuspended in PBS (pH 7.4) supplemented with 5 mM EDTA, protease inhibitors (10 µg/ml Pefabloc, 10 µg/ml aprotinin, 1 µM PMSF, and 5 µg/ml leupeptin) and 2.5 mM DTT and sonicated at 4°C three times for 2 min. Cleared lysates were put onto a GST-Trap FF column (GE Healthcare) and washed with PBS, pH 7.4. Finally, GST-tagged proteins were eluted with 50 mM Tris, pH 8.0, supplemented with 10 mM reduced L-glutathione and dialyzed against 25 mM sodium phosphate, pH 8.0, supplemented with 150 mM NaCl and 1 mM EDTA or against 50 mM Tris pH 8.0 supplemented with 150 mM NaCl and 1 mM Tris(2-carboxyethyl)phosphine (TCEP). The amount and the purity of proteins were analyzed via SDS-gel electrophoresis and aliquots supplemented with 10% glycerol frozen at −80°C.

Human recombinant CaMKIIβ (PV4205; Thermo Fisher Scientific) was resuspended in kinase buffer (50 mM Tris-HCl, pH 7.7, 10 mM MgCl₂, 5 mM MnCl₂, 1 mM TCEP, and 0.05% Triton X-100) either supplemented with 200 µM ATP, 1.2 µM calmodulin (Sigma-Aldrich), and 2 mM CaCl₂ or without supplementation and incubated for 10 min at 30°C. The kinase assay was started by adding 2 µg of purified GST-fusion protein and incubated for 60 min at 30°C under constant shaking at 750 rpm. The reaction was stopped via the addition of SDS sample buffer.

For the in vitro phosphatase assay, the CaMKIIβ phosphorylated cytoplasmic tail of β1 integrin was incubated with 2 µg of recombinant GST-PPM1F or PPM1F D360A or corresponding amounts of ILKAP in phosphatase buffer (50 mM Tris-HCl, pH 8, 10 mM MnCl₂, and 0.01% Tween 20) for 1 h at 30°C under constant shaking at 750 rpm. The reaction was stopped by the addition of either SDS sample buffer or the same volume of malachite green solution (54 mM NH₄Mo and 0.9 mM malachite green in 1 M HCl) and analyzed by WB or photometric measurement with OD_615nm.

Phosphatase assays with phospho-peptides were conducted using peptides synthesized by Pepscan: β1-22pT788pT789: (Biotin-Ahx-TGENPIYKSAV[pT][pT]VVNPKYEGK-OH), β1-22pT788: (H-TGENPIYKSAV[pT]TVVNPKYEGK-OH), β1-22pT789: (H-TGENPIYKSAVT[pT]VVNPKYEGK-OH), and MLC2-20pT10: (H-MSSKRAKAK[pT]TKKRPQRATS-OH). Depending on the assay, recombinant, E. coli–expressed GST-tagged PPM1F, PPM1FD360A, PTP1B, 7xHis-TEV ILKAP, Trx-His-S-PP5, or GST-PPM1F and GST-PPM1F D360A expressed in 293 cells were incubated with 100 µM phospho-peptides in phosphatase buffer for 1 h at 30°C. The reaction was stopped by adding the same volume of malachite green solution, and OD_615nm was measured. To determine PPM1F kinetics, GST-tagged PPM1F or PPM1FD360A was expressed in 293 cells and purified by glutathione agarose. Proteins were incubated with different concentrations of the double phosphorylated β1 integrin TpTp788/789 peptide or with 4-methylumbelliferone phosphate (Sigma-Aldrich). A standard curve for either K₂HPO₄ or 4-methylumbelliferone was measured in parallel and used to determine the kinetic parameters K_m and V_max by directly fitting the data to the Michaelis–Menten equation v = V_max [S]/K_m + [S] with reaction rate v, maximum velocity V_max, substrate concentration [S], and Michaelis constant K_m. Activity of ILKAP, PP5, and PTP1B was determined by 4-MUP assay using same molar amounts of each phosphatase. As negative controls, phosphatases were inhibited by 100 mM EDTA, 20 mM NaF, or 250 µM PTP1B specific inhibitor (CAS765317-72-4; Merck Millipore), and activity was determined by measuring fluorescence over 30 min (excitation/emission 386/448 nm). PPM1F WT was used as the reference.

Mice were kept in accordance with relevant institutional and national guidelines and regulations in the central animal care facility of University of Konstanz. The B6.129P2-PPM1Ftm1Dgen/J (PPM1F^+/−) mouse strain was obtained from The Jackson Laboratory. The targeted ppm1f gene was created by Deltagen by inserting a Lac0-SA-IRES-lacZ-Neo555G/Kan cassette via homologous recombination into the ppm1f locus, allowing the endogenous promoter to drive expression of β-galactosidase. The PPM1F^+/− mice have been backcrossed at least 20 generations to C57BL/6 mice. 3-wk-old mice or embryos were genotyped by amplification of DNA extracted from tissue biopsies or isolated from mouse embryonic fibroblast. The following PCR primers were used: primer 1, WT forward: 5′-CAA​CTC​TCC​ATC​ATG​CCC​ATC​AG-3′; primer 2, common reverse: 5′-AAG​CAG​GAA​GGG​ACA​CGT​GTC​GGT​C-3′; and primer 3, targeted allele forward: 5′-GGG​TGG​GAT​TAG​ATA​AAT​GCC​TGC​TCT-3′.

For genotyping, a PCR with 32 cycles was performed with an annealing temperature of 59°C and an elongation time of 40 s at 72°C, yielding a 200-bps and 450-bps PCR fragment for the WT and the targeted allele, respectively (see also Fig. 9 A).

For the generation of ppm1f^−/− MEFs, heterozygous mice were mated, and 10.5 d after coitus, the female mice were anesthetized and sacrificed. The uterus was dissected and cut between the implantation sites along the uterine horns into pieces containing single embryos. The embryos were isolated by removing the enveloping tissue, washed in sterile fibroblast growth medium (Promocell) supplemented with penicillin, streptomycin, and ciprofloxacin, and minced via up and down pipetting. The tissue homogenates were plated onto gelatin (0.1%) and human fibronectin (2 µg/ml)–coated dishes. After the second passage, primary fibroblasts were immortalized via transduction with pBabeZeo SV40 largeT (plasmid 1779; Addgene; Hahn et al., 2002). MEFs were cultured in DMEM supplemented with 10% FCS, nonessential amino acids, and sodium pyruvate.

The cDNA of mPPM1F was obtained from Source BioScience (I.M.A.G.E. Full Length cDNA clone IRAVp968A0987D; sequence accession BC042570) and was amplified with primers mPPM1F-BamHI-forward: 5′-GCT​TTA​GGA​TCC​AAT​GGC​CTC​TGG​AGC​CGC​ACA​GAA​C-3′ and mPPM1F-HindIII-reverse: 5′-CGC​CCG​TCA​AGC​TTC​TTA​GCT​TCT​CTG​TGA​GGT​ATT​AC-3′. The resulting PCR fragment was cloned into the pET24aHis-Sumo bacterial expression vector (Andréasson et al., 2008) via BamHI and HindIII restriction sites. The sequence-verified construct was transformed into competent E. coli BL21 (DE3), and expression of the recombinant protein was induced at OD₅₈₀ = 0.67 with 0.5 mM IPTG at 30°C for 4.5 h. His-Sumo–tagged mPPM1F was purified on a HisTrap FF column and eluted with 50 mM sodium phosphate buffer, pH 8, 0.5 M NaCl, and 0.5 M imidazole before removal of the His-Sumo tag by Ulp1 protease (Andréasson et al., 2008). 100 µg of purified recombinant mPPM1F were used for immunization of a New Zealand White rabbit in accordance with relevant institutional and national guidelines and regulations in the central animal care facility of University of Konstanz.

MEFs were seeded in 24-well plates and incubated for 24 h. Cells were starved in DMEM supplemented with 0.5% BSA for 12 h, and afterward stimulated with serum-containing growth medium and imaged for 12 h (30 min/frame). Single cells were tracked manually using the ImageJ particle tracking plugin and analyzed using the chemotaxis and migration tool (Ibidi GmbH).

All data are presented as mean ± SEM or mean ± SD as indicated. All statistical significances were determined using a two-tailed Student’s t test or one-way ANOVA followed by Bonferroni post hoc test with Prism5 (GraphPad). Significance is indicated with *, P < 0.05; **, P < 0.01; ***, P < 0.001; or ns, not significant.

Fig. S1 shows that the T788/T789 motif in the integrin β1 cytoplasmic tail is evolutionary conserved and that its pseudo-phosphorylation regulates the association with talin and filamin. Fig. S2 demonstrates that knock-down of PPM1F in 293T cells or NHDFs does not alter expression of core focal adhesion proteins and does not affect integrin surface levels. Fig. S3 shows that PPM1F KO in A172 cells does not alter integrin surface levels or expression of core focal adhesion proteins, but strongly affects integrin-dependent processes. Fig. S4 shows that filaminA knock-down pheno-copies integrin-dependent effects of PPM1F depletion in A172 cells. Fig. S5 shows that PPM1F purified from 293T cells dephosphorylates the conserved T788/T789 motif in the integrin β1 cytoplasmic domain.

The authors thank W. Hofer, D. Dietrich, R. Fässler, J. Morschhäuser, D.W. Piston, M.A. Schwartz, and D. Vestweber for valuable reagents. We are indebted to S. Feindler-Boeckh and the Flow Cytometry Facility Konstanz (FlowKon) for expert technical support. We also thank S. Helfrich and M. Stöckl for help in image analysis (Bioimaging Center, University of Konstanz).

T.M. Grimm and N.I. Dierdorf acknowledge support by the Konstanz Research School Chemical Biology. This study was supported by funds from the Deutsche Forschungsgemeinschaft (CRC969 project B06) to C.R. Hauck.

The authors declare no competing financial interests.

Author contributions: T.M. Grimm, N.I. Dierdorf, and C.R. Hauck conceived the study and designed the experiments; T.M. Grimm and N.I. Dierdorf conducted the experiments and evaluated the data; C. Paone conducted cell migration assays; K. Betz performed structure modeling; T.M. Grimm and C.R. Hauck wrote the paper. All authors read and approved the final manuscript.