Structural, pharmacological, and biochemical characterization of JNJ-BJ, a novel TNKS/2 inhibitor
XAV939 is a pyrimidine derivative that inhibits TNKS/2 by binding to the nicotinamide pocket of the enzymes, with half-maximal inhibitory concentrations (IC50) of 0.011 μM and 0.004 μM, respectively [12, 18]. JNJ-BJ is the first eluted enantiomer of a 3-ethylquinolinone (1A) and, like XAV939, competes with nicotinamide binding to tankyrases. When tested in an auto-PARsylation assay against the recombinant, baculovirus-expressed PARP domain of TNKS2, JNJ-BJ displayed an IC50 of 0.13 μM (pIC50 6.88; Fig. 1b). Details on the compound synthesis scheme are provided in Additional file 1.
A characteristic readout of TNKS/2 inhibition is a reduction in β-catenin-dependent signaling in cells with a hyperactive Wnt pathway [12]. Coherent with the inhibitory activity towards purified TNKS2, treatment of adenomatous polyposis coli (APC)-mutant DLD1 colorectal cancer cells with JNJ-BJ impaired Wnt-driven transcriptional responses, as assessed by both a TOPflash luciferase reporter assay (Fig. 1c; raw data in Additional file 2) and reverse transcription quantitative polymerase chain reaction (RT-qPCR) analysis of the expression of established β-catenin target genes (Fig. 1d; raw data in Additional file 2). As expected, and in accordance with previous findings [12], similar results were obtained with XAV939 (Fig. 1c, d; raw data in Additional file 2).
TNKS/2 inhibition hampers lung cancer cell invasion and migration in response to hepatocyte growth factor
Although mutations of APC or β-catenin are infrequent in lung cancer, hyperactivation of the Wnt pathway, as evidenced by transcriptional overexpression of Wnt-responsive genes, has been documented in samples from aggressive lung adenocarcinomas [19]. Because TNKS/2 are accredited upstream regulators of the Wnt pathway [12], we initially pursued the idea that interception of TNKS/2 activity might prevent Wnt-induced lung cancer cell dissemination. As a first step, we explored the consequences of TNKS/2 blockade on cell motility in four lung adenocarcinoma cell lines—H322, HCC827, H460, and A549—using XAV939 and JNJ-BJ as tool compounds.
To provide proof of concept that TNKS/2 blockade was proficient in lung cancer, A549 cells were treated with increasing concentrations of XAV939 or JNJ-BJ for 24 h and assessed for expression of axin1, which is typically stabilized by TNKS/2 inhibition owing to impaired TNKS/2-mediated PARsylation and consequent protein degradation [12]. Western blot analysis of total cell extracts revealed that both compounds were able to induce a dose-dependent increase of axin1 protein content (Fig. 2a), indicating successful TNKS/2 inactivation. Remarkably, when challenged in Matrigel-coated Transwell systems using hepatocyte growth factor (HGF) as a chemoattractant [20], A549 cells exhibited a dose-dependent reduction in invasive ability following TNKS/2 inactivation by XAV939 or JNJ-BJ (Fig. 2b; raw data in Additional file 3).
Analyses were subsequently extended to the remaining lung cancer cell lines by applying the dose that yielded maximal invasion impairment in the setup experiments (10 μM). In the case of XAV939, 5–10 μM is the standard inhibitor concentration commonly used in biological studies [21, 22]. Consistent with that observed in A549, axin1 was invariably stabilized upon treatment with either compound (Fig. 2c). Similarly, TNKS/2 inactivation compromised HGF-induced chemotactic response to a comparable extent in all the cell lines tested, apart from a weaker activity of XAV939 in H322 (Fig. 2d; raw data in Additional file 3). A decrease in cell invasion was paralleled by reduced migration in wound healing (scratch) assays. With the exception of H460 cells (which proved unsuitable for production of a compact monolayer and were therefore excluded), abrogation of TNKS/2 activity markedly dampened HGF-induced wound closure competence (Fig. 2e; raw data in Additional file 3).
It is worth noting that in these cells we did not observe noticeable anti-proliferative effects following TNKS/2 inhibition, even after a 72 h exposure to drugs (Additional file 4: Figure S1; raw data in Additional file 5). This result is at odds with the established mitotic function of TNKS/2, but is congruent with previous observations showing that TNKS/2 pharmacological inhibition is much less detrimental to cell proliferation than RNA interference (RNAi)-based silencing [4, 5, 12, 21]. Whatever the explanation for this discrepancy, which remains a matter of debate [2, 12, 21], these findings suggest that mechanisms other than a mere growth disadvantage are implicated in the reduced cell motility observed in response to TNKS/2 blockade.
We also employed RNAi as an alternative means of inactivating TNKS/2. In agreement with pharmacologic experiments, productive co-depletion of TNKS and TNKS2 in A549 cells (Additional file 6: Figure S2A and S2B; raw data in Additional file 7) resulted in axin1 stabilization (Additional file 6: Figure S2B) and reduced cell invasion (Additional file 6: Figure S2C; raw data in Additional file 7). Likewise, wound closure ability was lessened by RNAi-mediated TNKS/2 silencing in A549 cells (Additional file 6: Figure S2D; raw data in Additional file 7). Although genetic knockdown of TNKS/2 has been shown to affect cell proliferation, the time frame of Transwell and scratch assays (24 h) was likely sufficiently short not to bias the anti-invasive outcome of TNKS/2 abrogation. In summary, impaired cell invasion proved to be a direct function of increasing compound concentrations and was achieved by two structurally different inhibitors; moreover, TNKS/2 genetic silencing recapitulated the biochemical and biological effects of pharmacologic inhibition. These findings indicate that the impaired chemotactic response is a specific consequence of TNKS/2 disruption.
The anti-invasive outcome of TNKS/2 inhibition is independent of the Wnt pathway
The anti-invasive and anti-migratory effects produced by TNKS/2 neutralization were consistent with the working hypothesis that blockade of TNKS/2 activity would blunt Wnt-mediated pro-invasive cues. We therefore analyzed whether this weakened chemotactic response was in fact ascribable to interception of Wnt signaling.
First, the TOPflash reporter system was employed to gauge Wnt-dependent transcriptional responses after cell exposure to TNKS/2 inhibitors (for this purpose, we used H322 cells owing to their high amenability to transfection procedures). We found that TNKS/2 inhibition did not affect Wnt transcriptional activity, either basally or upon addition of the canonical Wnt ligand Wnt3a (Additional file 8: Figure S3A; raw data in Additional file 9). As a complementary approach, expression of established Wnt target genes was assessed by RT-qPCR in the whole panel of lung adenocarcinoma cell lines tested in the cell invasion experiments. As shown in Additional file 8: Figure S3B (raw data in Additional file 9), stimulation with Wnt3a led to increased expression of at least some of the target genes, with variable levels of induction in the various cell lines (likely due to cell type-specific differences). Also in this experimental setting, and consistent with the TOPflash assay, cell line-dependent transcription of Wnt3a target genes was not detectably influenced by treatment with TNKS/2 inhibitors (Additional file 8: Figure S3B; raw data in Additional file 9). Finally, Transwell assays demonstrated that cell invasion was not evidently fostered by Wnt3a (Additional file 8: Figure S3C; raw data in Additional file 9), further supporting the irrelevance of Wnt signaling to TNKS/2-related migratory phenotypes in our cellular models.
All in all, Wnt-dependent activities did not substantially enhance cell motility in lung cancer cell lines, nor were they clearly impacted by TNKS/2 inactivation. This implies that the observed effects on migration and invasion likely rely on alternative mechanisms.
TNKS/2 inhibition impacts the dynamics of formation of cell membrane protrusions
To get better insight into how cell motogenic responses were impacted by TNKS/2 abrogation, wounded A549 monolayers were treated with HGF and monitored for 1 h using time-lapse videomicroscopy in the presence or absence of JNJ-BJ, which was selected as the best performing compound in preliminary experiments (see Fig. 2e). Single images were captured every 12 s to allow for distinct visualization of the migration process (Additional file 10: Movie M1). Figure 3a shows representative time-lapse snapshots, captured 15, 30, and 60 min after HGF stimulation. We noticed that the dynamics of protrusion formation at the wound edge were discernibly slowed by JNJ-BJ. In particular, control cells displayed pronounced membrane ruffling, characterized by repetitive and vigorous bursts of incipient projections at the leading edge (Fig. 3a, white arrows); by contrast, protrusive activity appeared flimsier and membrane extensions subsided quickly in cells treated with JNJ-BJ. Besides the stronger propulsive flows, control cells also developed numerous circular dorsal ruffles (Fig. 3a, yellow arrows), whose function generally implies transition from a static to a motile phenotype [23]; conversely, the formation of such structures was almost completely prevented by JNJ-BJ (Fig. 3a and Additional file 10: Movie M1).
On the basis of such observations, we assumed that TNKS/2 inhibition impaired cell movement by negatively impacting migration dynamics at the leading edge. To complement the time-lapse qualitative information, we quantified membrane extensions in HGF-stimulated A549 cells with or without TNKS/2 inhibitors. As shown in Fig. 3b (raw data in Additional file 11), the proportion of protruding cells was significantly decreased by either compound after 15 and 30 min of HGF exposure. Remarkably, the curves related to TNKS/2-inhibited cells tended to re-align with the curve of control cells after 1 h, suggesting that TNKS/2 blockade hindered early rather than late events of cell migration.
TNKS/2 inhibition enhances microtubule stability in interphase cells
TNKS/2 couple with the mitotic microtubule circuitry to affect spindle structure and function [4]. As specified earlier, this is accomplished through interaction with various microtubule-related proteins as well as with other spindle-associated targets [1, 2, 4, 15]. We reasoned that analogous functional connections might be extended to interphase microtubule-dependent activities, whose dynamics are intimately related to polarized cell migration [24, 25].
Inception of oriented cell movement entails microtubule-dependent reorganization of the cellular architecture to establish a rear–front axis. This asymmetric pattern is supported by the inherent instability of microtubules, which constantly undergo rounds of shrinkage and regrowth [26]. To investigate whether TNKS/2 neutralization interfered with microtubule dynamic instability, we deconstructed the microtubule network in A549 cells by cold treatment (4 °C, 6 min) or nocodazole (1 μM, 5 min). Under basal conditions, microtubules were intact and their organization was similar in both untreated and TNKS/2-inhibited cells (Fig. 4a, b and Additional file 12: Figure S4; raw data for Fig. 4b in Additional file 13). Notably, cold-induced or nocodazole-induced microtubule disassembly was widespread in control cells whereas it was markedly prevented in cells treated with XAV939 or JNJ-BJ (Fig. 4a, b and Additional file 12: Figure S4; raw data for Fig. 4b in Additional file 13). This indicates that microtubules were rendered more stable by TNKS/2 inactivation. RNAi-mediated TNKS/2 depletion recapitulated the phenotype produced by TNKS/2 pharmacologic inactivation (Additional file 14: Figure S5).
The finding that TNKS/2 blockade increased the proportion of stable microtubules suggests that TNKS/2 inhibition might obstruct microtubule-dependent activities implicated in cell polarity and directional migration.
TNKS/2 inhibition affects centrosome reorientation in migrating cells
Microtubule-related activities are central to polarized cell migration through mechanisms that involve protein targeting to cortical sites and the generation of pulling forces that help reorganize cell architecture in response to chemotactic cues [24, 27, 28]. One hallmark of cell polarization is the relative orientation of the nuclear-centrosome axis with respect to the rear–front axis (which defines the direction of cell migration); in general, this alignment is thought to correlate with the onset of cell migration and to contribute to the establishment of cell polarity by facilitating membrane trafficking from both the Golgi and the endocytic recycling compartments towards the leading edge [29]. Centrosome positioning is largely influenced by microtubule dynamics; regardless of context-dependent idiosyncratic differences, it is apparent that during productive cell locomotion the centrosome relocates in front of the nucleus facing the direction of cell migration [29, 30].
On such premises, we explored whether TNKS/2 inhibition could perturb microtubule-dependent establishment of cell asymmetry by measuring the amount of reoriented centrosomes in wound-edge A549 cells following induction of migration by HGF. Based on previously published studies [31], centrosomes were scored as “fully polarized” if the angle between the nuclear-centrosome axis (Fig. 5, red arrows) and the front-back axis (Fig. 5, white arrows) was less than 30°, which vouches for centrosome dwelling within the forward-facing quadrant (raw data in Additional file 15). In HGF-treated cells without TNKS/2 inhibitors, centrosome location shifted from an essentially random distribution around the nucleus to a biased rearrangement along the migration axis, with polarization angles reflecting a progressive degree of cell orientation over time (Fig. 5). The increase of fully polarized centrosomes was accompanied by a concomitant reduction in the number of untailored centrosomes displaying angles greater than 60° (Fig. 5). Remarkably, HGF-induced acquisition of the polarized phenotype was antagonized by treatment with either TNKS/2 inhibitor. At 1 h after cell wounding, almost 50 % of control cells were fully polarized, compared to only 34 % and 29 % of XAV939-treated and JNJ-BJ-treated cells, respectively; also the decline of non-polarized centrosomes was not as manifest as that observed in control cells (Fig. 5). After 4 h, 60 % of control cells showed fully oriented centrosomes, as opposed to nearly 40 % of TNKS/2-inhibited cells (Fig. 5). Akin to TNKS/2 pharmacologic blockade, genetic knockdown of both TNKS and TNKS2 in A549 cells disturbed centrosome repositioning upon HGF stimulus (Additional file 16: Figure S6). Collectively, these results indicate that TNKS/2 inactivation results in perturbed centrosome reorientation in migrating cells, with delayed alignment along the rear–front axis.
TNKS/2 inhibition retards APC recruitment at the leading edge
Adenomatous polyposis coli (APC) is a strategic cytoskeletal coordinator of migration directionality owing to its recruitment to microtubule-dependent clusters at protrusive areas. This is thought to regulate microtubule anchoring at the cell cortex and consequent centrosome rearrangement [32, 33]. With this in mind, we assessed HGF-induced redistribution of APC in wounded A549 cell monolayers, either left untreated or exposed to TNKS/2 inhibitors. Immunofluorescence staining showed a diffuse APC cytoplasmic distribution along with an intense nuclear signal (Fig. 6a), consistent with the notion that APC shuttles between the nucleus and the cytoplasm to assist β-catenin nuclear export [34]. Immediately after wounding, APC was evenly distributed throughout the cell monolayer and in cells closely adjacent to the wound area (Fig. 6a). APC redistribution in cortical clusters at the wound edge was evident in control cells as early as 15 min after HGF stimulation and persisted up to 1 h (Fig. 6a). By contrast, treatment with TNKS/2 inhibitors led to a marked attenuation of APC membrane targeting at the 15 min time point after HGF stimulation. Morphometric quantitation revealed that the number of APC-positive protrusions was reduced by TNKS/2 inactivation also when normalized against the number of total protrusions (Fig. 6b; raw data in Additional file 17). This indicates that the impairment of APC membrane relocalization was not a mere consequence of lessened membrane ruffling secondary to TNKS/2 blockade. Of note, the representative curves of control and TNKS/2-inhibited cells tended to readjust over time: in fact, APC-decorated lamellipodia were equally represented at 30 min and 1 h post-wounding in all the conditions tested (Fig. 6a, b). Thus, TNKS/2 inhibition deferred APC membrane redistribution in response to HGF, which integrates with our previous data about retarded centrosome repositioning. The fact that APC localization in protrusions was only transiently retarded, whereas centrosome reorientation was impaired for longer times, is congruent with the notion that, once cell polarization is established, positive feedback loops initiate between the actin-rich cortex and the microtubule cytoskeleton to maintain and reinforce the existing polarity axis [35]. Therefore, while APC relocation is required for centrosome rearrangement, the persistence of such a tailored phenotype is likely sustained by processes that engage proteins and network interactions other than APC, together with the microtubule cytoskeleton. Remarkably, impaired accumulation of APC at the protrusive front was even exacerbated by RNAi-mediated TNKS/2 knockdown; indeed, HGF-induced recruitment of APC at cortical areas was prevented at all the time points in A549 wound-edge cells transduced with TNKS- and TNKS2- short hairpin RNAs (Additional file 18: Figure S7).
Finally, to get further insight into how TNKS/2 may regulate APC-dependent microtubule dynamics, we analyzed the subcellular distribution of TNKS and APC in migrating A549 cells. We found that both APC and TNKS were recruited at the leading edge upon HGF stimulation (Additional file 19: Figure S8). Similar to APC, TNKS enrichment at membrane protrusions was impaired in the presence of TNKS/2 inhibitors (Additional file 19: Figure S8). TNKS staining was specific, because TNKS/2 blockade also induced the formation of TNKS-enriched puncta, a reported phenotype of TNKS/2-inhibited cells [36] (Additional file 19: Figure S8).
The fact that TNKS follows subcellular dynamics similar to those experienced by APC and the experimental observation that persistent obliteration of TNKS/2 by genetic silencing durably precludes APC relocation at the leading edge reinforce the hypothesis that TNKS/2 are implicated in the establishment of cell polarization during oriented locomotion. Taken together, these observations allow us to draw a coherent scenario whereby TNKS/2 blockade appears to interfere sequentially with regulated events that, in space and time, orchestrate the establishment of cellular polarity. This perturbation of cell polarity is facilitated by the enhanced microtubule stability produced by TNKS/2 inactivation.