Posts Tagged ‘Curve fitting’

Deciphering Mode of Action of Functionally Important Regions

Larry H. Bernstein, MD, FCAP, Curator



Deciphering Mode of Action of Functionally Important Regions in the Intrinsically Disordered Paxillin (Residues 1-313) Using Its Interaction with FAT (Focal Adhesion Targeting Domain of Focal Adhesion Kinase)

Intrinsically disordered proteins (IDPs) play a major role in various cellular functions ranging from transcription to cell migration. Mutations/modifications in such IDPs are shown to be associated with various diseases. Current strategies to study the mode of action and regulatory mechanisms of disordered proteins at the structural level are time consuming and challenging. Therefore, using simple and swift strategies for identifying functionally important regions in unstructured segments and understanding their underlying mechanisms is critical for many applications. Here we propose a simple strategy that employs dissection of human paxillin (residues 1–313) that comprises intrinsically disordered regions, followed by its interaction study using FAT (Focal adhesion targeting domain of focal adhesion kinase) as its binding partner to retrace structural behavior. Our findings show that the paxillin interaction with FAT exhibits a masking and unmasking effect by a putative intra-molecular regulatory region. This phenomenon suggests how cancer associated mutations in paxillin affect its interactions with Focal Adhesion Kinase (FAK). The strategy could be used to decipher the mode of regulations and identify functionally relevant constructs for other studies.
Neerathilingam M, Bairy SG, Mysore S (2016) Deciphering Mode of Action of Functionally Important Regions in the Intrinsically Disordered Paxillin (Residues 1-313) Using Its Interaction with FAT (Focal Adhesion Targeting Domain of Focal Adhesion Kinase). PLoS ONE 11(2): e0150153. doi:10.1371/journal.pone.0150153

Genomic data suggests that a large proportion of eukaryotic proteins appear to adopt disordered structures in physiological conditions [1, 2]. Mutations/modifications in such IDPs are shown to be associated with various diseases (like cancer) [3]; therefore, understanding their structural behavior is critical for various applications like drug-targeting, mapping protein interactions, deciphering mode of action and finding functional relevance. However, deciphering mode of action in IDPs has been challenging given that unstructured segments render poor chemical shift dispersions and electron density in major techniques like NMR and X-ray, respectively [4]. For example, it took almost 10 years to decipher the mode of action of Sic1, a disordered protein involved in inhibition of a cyclin-dependent kinase [5]. One way to map and study the functional regions is to make truncated constructs by dissecting the whole construct rationally. A limited number of dissection constructs are usually generated; this is due to the time-consuming and challenging process of generating soluble and functionally relevant constructs when studies are performed in-vivo and constructs are prepared and tested sequentially. Here we present a simple high throughput (HTP) screening strategy (Fig 1a), which focuses on finding functionally relevant regions in IDPs based upon its interaction with a binding partner. Close to thirty dissection constructs of the IDP were generated and studied in parallel to understand the importance and functionality of the various regions of the protein. We perform cell-free expression followed by solubility check and GST pull-down interaction study in HTP format. Though both cell-free expression and GST pull-down assay have been individually performed in HTP format [6, 7], we did not find previous studies that combine the two methods in HTP format. Although the nature of interaction of IDPs with respective binding partners may vary, our strategy may be used to derive crucial insights into “structural behavior” of the unstructured segments in modulating the interaction. The strategy can also be used to identify functionally important regions in the IDP that would be suitable for further structural studies.


Fig 1. Dissection of paxillin constructs (residues 1–313) followed by expression and interaction studies.

(a) Timeline for overall-strategy. (b) Illustration of solubility and activity level of linear dissected human paxillin (residues 1–313). (c) Phosphor screen image of filter assay for optimization of temperature for paxillin constructs (left). Tabular representation of paxillin constructs, negative and positive controls corresponding to each well in filter assay [1].(d) Phosphor screen image of 10% SDS PAGE of 35S labeled cell-free expressed samples after GST pull-down assay of the paxillin constructs A1–E1; The right panel shows fraction of interaction of each construct with respect to B2 (since B2 showed maximum level of interaction) (e) Illustration of solubility and activity of dissected C3 constructs. All experiments were performed in triplicates and averaged. To rule out non-specific interactions that might occur with GST tagged FAT, GFP that was expressed in cell-free system and a reaction without DNA template were used as negative controls.    http://dx.doi.org:/10.1371/journal.pone.0150153.g001

Disorder/Intrinsic disorder seems to be a common feature of hub proteins in eukaryotes [2], thus highlighting the need for studying the mode of action of unstructured segments in such proteins. Here we used paxillin (residues 1–313), an intrinsically disordered construct, for demonstrating this approach. Paxillin (residues 1–313) consists of multiple protein interaction sites that are connected by flexible disordered sequences [8]. The disordered regions in paxillin have been detrimental in efforts to study the complete structure of the protein due to the demerits mentioned previously. This explains the lack of structural details of regulation of paxillin binding. Residues 1–313 of paxillin consist of five leucine-rich sequences LD1-LD5 (with consensus sequence: LDXLLXXL), termed LD motifs, which are highly conserved between species and other family members such as Hic-5, leupaxin and PAXB [8]. Paxillin interacts with multiple proteins involved in cell migration, actin rearrangements and cell proliferation [9]. Mutations in paxillin are shown to be associated with lung cancer [3, 10]; and the differential expression of paxillin is associated with various forms of cancer and other diseases such as Alzheimer’s and inflammation [1113]. This implies the importance of studying the structural and functional characteristics of paxillin. Most paxillin studies focus on interactions of LD motifs with proteins such as focal adhesion kinase (FAK), vinculin and v-crk, providing clues towards their importance in deciphering the functionality of paxillin [8, 14, 15]. Though regions of paxillin that bind to various partners were deciphered through previous studies, the basis of effect of mutations in paxillin on binding its partners was not explained. Mutations in paxillin, some that were observed to be associated with cancer were positioned in the intrinsically disordered regions between the LD motifs and not on the motifs themselves [3, 10]. For example, P30S, G105A and A127T mutations lie between LD1 and LD2 motif; P233L and T255I mutations lie between LD3 and LD4 motifs. This shows that the LD motifs alone do not govern the functionality, but unstructured regions linking the LD motifs could play a major role. In normal conditions, FAT (Focal adhesion targeting domain of FAK) binds hydrophobically through its HP1 (Hydrophobic patch 1) and HP2 (Hydrophobic patch 2) sites to paxillin LD motifs—LD2 and LD4 [16, 17], which lead to activation of binding sites for other proteins on paxillin. LD2 preferentially binds to the HP2 site, whereas LD4 preferentially binds to the HP1 site [18]. In a state of cancer caused by mutations in paxillin, the LD interactions could be hindered, as mutations in the unstructured segments result in abnormal binding of FAK to either of the LD motifs [9]. Here we wanted to locate the region involved in the structural modulation of paxillin-FAT interaction by adopting a simple approach (Fig 1) that involves dissected proteins generated using cell-free protein expression coupled with protein-protein interaction study. We map the disordered proteins’ structural importance to understand the function and modulation of paxillin-FAT interaction in days rather than months (Fig 1a).

Dissection and identification of fragments of paxillin (residues 1–313) with functional relevance

We dissected paxillin (residues 1–313) (Fig 1b) into nested sets using PCR such that each of the constructs had either or both LD2 and LD4 motifs (S1 Fig and S1 Table). Further, these constructs were expressed in soluble form using small-scale cell-free expression system in a 96 well format (Fig 1c). However, all constructs except A6, B6, C4 and C5 expressed detectable amounts of protein (S2a Fig and S2 Table). The failure in expression of the above constructs could be due to the instability of the smaller peptide fragments that might be susceptible to proteolytic cleavage [19]. Soluble protein from small-scale expression of the dissected constructs namely A1, A2, A3, A4, A5, B1, B2, B3, B4, B5, C1, C2, C3, D1, D2, and E1 were pulled down and analysed (Fig 1d). Although constructs A1–A5, B1–B5, C1, C2, D1 and E1 interacted successfully, C3 (containing LD2) and D2 (containing LD4) failed to interact (Fig 2c) despite containing LD motifs. However, based on previous reports [8, 16, 17], we expected all constructs containing either LD2 and/or LD4 to interact with the FAT domain. Therefore, this led us to suspect that intra-molecular auto-inhibition in unstructured segments modulated binding of FAT to LD motifs in paxillin.


Fig 2. Regulatory and masking regions around paxillin’s LD2 and LD4 and their circular dichroism spectra.

(a) CD spectra of paxillin LD peptides (LD1-LD5) and constructs: B2, C3, C35 and D2. CD spectra of LD2, LD4, C35 and D2 constructs showed negative bands at 222nm and 206nm and a positive band at 192nm that confirms the presence of alpha helical content thus may behave as folded effector binding sites. However, LD1, LD3, LD5, B2 and C3 do not show the characteristic peaks of secondary structures, thus may behave as unfolded effector binding sites. (b) LD2 regulatory region (54–130) and masking region (167–224) evidenced by constructs B3, B4 and B5. (c) LD4 regulatory region (216–257) and masking region (280–313) evidenced by constructs D1, D2 and E1.

Identification of regulatory regions and their mechanisms

To investigate the non-interaction of C3, a series of C3 deleted constructs (C31 –C310) (Fig 1e,S1 Table) were generated to determine the internal region that influenced the non-functioning of C3. C36 linear template could not be amplified for expression. As solubility of C3 could play a critical role in determining interaction, the homogeneity of the sample was confirmed by capillary electrophoresis under non-reducing conditions [20] (See S3 Fig). The linear templates—C31, C32, C33, C34 and C35 were successfully expressed in soluble form, The other C3 deleted constructs did not express due to issues related to small size as described earlier. Surprisingly, none of the C3 deleted constructs interacted with FAT despite the presence of the LD2 motif, although constructs such as B3, B4 and B5 that contain regions overlapping with C3 showed interaction (S2b and S2c Fig, Fig 1d and 1e). Here B3 that included the whole of C3 and unstructured segment 54–130 showed interaction (Fig 1b). Constructs B4 and B5 also containing residues 54–130 showed interaction despite differing from B3 by lacking regions 167–224 and 155–224, respectively. Interestingly, the non-interacting constructs C3 and C35 do not contain 54–130 residues, but include the regions 167–224 and 167–189, respectively (Fig 1b). Here constructs containing region 167–189 but lacking 54–130 did not interact with FAT despite LD motif alone showing interaction (switch off) (Fig 3a). Whereas, if 54–130 was included, interaction was reinstated (switch on) (Fig 3a). This clearly shows that interaction of LD2 in construct C35 is masked by residues 167–189 (masking region) (Fig 2b). The constructs B3 and B4 binding to FAT despite the presence of the masking region led us to conclude that the region 54–130 (regulatory region) acts to remove the masking effect (Fig 2b).
Fig 3. Binding studies of paxillin constructs using Bio-layer Interferometry on OctetRed96.

(a) Switch off in C3 and D2 on LD2 and LD4 respectively; Hypothesis of partial switch on when regulatory region of LD2 is absent, as evidenced in C2. (b) Concentration calibration curves depicting binding of constructs B2, C35, ‘54–189’, ‘79–189’, ‘105–189’ with GST-FAT. The data is representative of a single experiment. Each experiment was performed at-least thrice. (c) Illustrations of C35, C35_1, C35_2 and C35_3.  http://dx.doi.org:/10.1371/journal.pone.0150153.g003

Similar to LD2, LD4 in construct D2 containing 216–257 (masking region) requires additional residues of paxillin 280–313 (regulatory region) for FAT binding (Fig 2c), which was demonstrated by showing the interaction with constructs D1 (spanning region 216–313) (Fig 1b) and E1 (spanning region 258–313). To visualize the non-binding of FAT to C35, in-silico methods were employed to model the C35 construct and docked with the crystal structure of FAT (1K05, residues 916–1050 [21]) (Fig 4). The docking results showed a clear masking effect in the C35 construct by the 167–189 (masking region) residues. The constructs B2, C3 and C35 were also structurally characterized using CD analysis (Large scale cell-free expression was performed for this purpose, see S3 Fig). The percentage of alpha helical content was found to be much higher in C35 (95.32%) as compared to B2 (12.43%) (Fig 2a, S3 Table). Therefore, the dissection(s) of B2 to C35 allowed the identification of structured regions (C35) as compared to the disordered B2. Further, it showed that the LD2 peptide and C35 have significant alpha-helical structures that do not translate into functional similarity as evidenced by the inability of C3, C35 and D2 to bind to FAT. Moreover, LD2 peptide binds to FAT while C35 does not (Fig 1e and S2b and S2c Fig). A similar observation was made when comparing the ability of LD4 peptide and the inability of D2 to bind to FAT despite both having detectable α-helical content (Fig 1b). Thus, these results confirm the existence of masking and regulatory regions (Fig 2b and 2c) that determine switch on and off and in turn, intra-molecular auto-inhibition. C2 showed activity despite missing regulatory regions for both LD2 and LD4 (similar activity observed in C1). This could be because the unfolded nature of LD3 effector binding site that is located between LD2 and LD4 is flexible to mask only a single LD motif but not both (Partial switch on, Fig 3a).


Fig 4. In-silico analysis of non-binding of C35.

(a) LD2 crystal structure from PDB id: 1K05 (left) being compared with the LD2 structure in the side view and top view of C35 structure showing the masking of the hydrophobic binding region predicted through HMM based SAM-T08 software. The LD2 binding region and the masking regions are depicted by the bracketed region. (b) Docking control showing FAT (co-ordinates from PDB id: 1K05) and LD2 (co-ordinates from PDB id: 2L6F, NMR model # 1) interaction using Hex 6.3 software. (c) Docking of C35 with FAT showing non-interaction due to masking effect. The sidechains of the active residues are shown as red sticks. The hydrophobic patch—HP2 in FAT molecule, which preferentially binds to LD2 is shown as a space filling model in orange (part of helix 1 of FAT) and grey (part of helix 4 of FAT) colors.   http://dx.doi.org:/10.1371/journal.pone.0150153.g004

To predict the influence of this structural modulation, the state of LD motifs structurally before and after binding to FAT had to be understood. CD spectra of LD1, LD3 and LD5 peptides showed characteristics of random coil (Fig 2a, S3 Table) thus validating that the LD1, LD3 and LD5 motifs could exist as unfolded effector binding sites (not available for interaction) in our study and could fold upon undergoing allosteric changes after binding to their respective targets.

Validation of protein-protein interaction study using bio-layer interferometry studies

Bio-layer interferometry studies were performed to further validate the interaction studies and also to get insights into the binding affinities. Here apart from constructs B2 and C35, three other constructs that include different lengths of the regulatory region along with the C35 region were used for the studies, namely—Construct C35_1(54–189); Construct C35_2 (79–189) and Construct C35_3 (105–189) (See Fig 3b and S4 Fig). As seen in Fig 3c and Table 1, B2 shows maximum binding with KD value in the nano-molar range and the curves fit into a 1:1 binding model. C35 shows negligible binding and the rest of the constructs show binding lower than B2 with KD values in micro-molar range and the curves fit into a 2:1 binding model (See S5 Fig).

According to previous reports, FAK has to bind to both LD2 and LD4, failing which phosphorylation during signalling is reduced [8], which is observed in case of cancer [3], thus resulting in abnormal functioning of paxillin. We investigated this by analysing B2, which showed higher interaction than B1, despite missing the regulatory region of LD4 (Fig 1b). Similarly, C2 showed activity despite missing regulatory regions for both LD2 and LD4 and the presence of masking regions (similar activity observed in C1). This suggests that the masking region that is located between LD2 and LD4 is flexible to mask only a single LD motif but not both (Fig 3a). Interestingly, paxillin mutations associated with lung cancer were observed in the unstructured segments, particularly the regulatory region of LD2 and masking region of LD4 [3]. We hypothesize that these mutations prevent proper functioning of the regulatory regions, thus resulting in masking of either of the LD motifs causing abnormal functioning of paxillin. Evidence that these regions regulate FAT-paxillin binding was further provided in our study in the form of the bio-layer interferometry results; where C35 did not show any binding, but the constructs that included different lengths of the regulatory region along with the C35 region showed binding with KD values in the micro-molar range. This suggests that the LD2 region in these constructs is not masked, since it is seen in previous studies that the KD value for FAT binding to a single LD motif of paxillin is in micro-molar range. It also suggests that the region between residues 105–131 is sufficient for preventing the masking of LD2 region, thus allowing interaction with FAT (See illustrations in Fig 3b). Except B2 (that had a 1:1 binding stoichiometry and higher binding affinity), all other constructs (C35_1, C35_2, C35_3) showed a 2:1 binding stoichiometry. This suggests that both LD motifs of B2 engage both the FAT HP sites thus resulting in higher affinity; whereas in the other 3 constructs (C35_1, C35_2, C35_3), each FAT HP site (HP1 and HP2) interacts with individual molecules thus giving a 2:1 stoichiometry. This is in agreement with previous studies where both the LD motifs were found to interact with both HP1 and HP2 hydrophobic patches of FAT [16]. The higher affinity of B2 to FAT could be due to presence of both LD2 and LD4; the proposed intra-molecular regulatory regions could also play a role in the increased affinity. Therefore, we understand that the abnormal modulation in cancer involves redirection of FAK to a single LD motif; and targeting drugs for re-establishing the function at regulatory regions could be critical.

Unlike many existing techniques like array based yeast two hybrid assay, phage display method and tandem affinity purification; the strategy used here (combination of cell-free expression, filter based solubility assay and interaction study in HTP format) facilitated quick identification of the role of unstructured regions involved in paxillin-FAT interaction in HTP format. Particularly, in paxillin-FAK interactions, which determine focal adhesion and cellular signalling, we understood the structural masking and unmasking behaviour of unstructured segments in paxillin to determine FAK interaction. The structure of paxillin is not yet elucidated due to difficulties with respect to its disordered nature. In this study, the templates that we generated using the high throughput dissection strategy allowed us to analyze various regions of paxillin, with respect to structure, solubility and function. To our knowledge, this study is the first report of switch on and off mechanisms working together in controlling allosteric modulation/auto-inhibition in a human hub protein. As many eukaryotic proteins are disordered, our study opens avenues for analyzing novel modulations at allosteric sites using appropriate interaction studies, which could lead to identification of new drug target sites. In this regard, we hope the above strategy will be instrumental in understanding mechanisms of other disordered proteins as well, in days rather than months. This strategy could also be used as an initial screening method for techniques like SAXS, smFRET and others.

Read Full Post »