Glucocorticoids rapidly inhibit cell migration through a novel, non-transcriptional HDAC6 pathway

ABSTRACT Glucocorticoids (GCs) act through the glucocorticoid receptor (GR, also known as NR3C1) to regulate immunity, energy metabolism and tissue repair. Upon ligand binding, activated GR mediates cellular effects by regulating gene expression, but some GR effects can occur rapidly without new transcription. Here, we show that GCs rapidly inhibit cell migration, in response to both GR agonist and antagonist ligand binding. The inhibitory effect on migration is prevented by GR knockdown with siRNA, confirming GR specificity, but not by actinomycin D treatment, suggesting a non-transcriptional mechanism. We identified a rapid onset increase in microtubule polymerisation following GC treatment, identifying cytoskeletal stabilisation as the likely mechanism of action. HDAC6 overexpression, but not knockdown of αTAT1, rescued the GC effect, implicating HDAC6 as the GR effector. Consistent with this hypothesis, ligand-dependent cytoplasmic interaction between GR and HDAC6 was demonstrated by quantitative imaging. Taken together, we propose that activated GR inhibits HDAC6 function, and thereby increases the stability of the microtubule network to reduce cell motility. We therefore report a novel, non-transcriptional mechanism whereby GCs impair cell motility through inhibition of HDAC6 and rapid reorganization of the cell architecture. This article has an associated First Person interview with the first author of the paper.

In separate experiments tracking cells using brightfield microscopy, cell migration is also significantly inhibited with the GR antagonist RU486 (vehicle-treated control median total displacement=235.5µm; median step length=2.68µm compared to RU486 median total displacement=210.3µm; median step length=2.63µm) (Fig.   1D,S1C), which was surprising, and indeed RU486 did not antagonise the inhibitory effect of dex (Fig. 1D,E).
RU486 is a competitive GR antagonist that binds and induces GR nuclear translocation, but then recruits corepressors including NCoR to block transcription (Fig. S1D). In A549 cells we did not detect statistically significant GR transactivation with RU486 treatment (Fig. S1D).
The overlapping actions of dex and RU486 on cell migration suggest a common mechanism of action, but one that requires the GR, and not the transcriptional regulatory actions of the GR. As the effects seen were so unexpected we also tested the requirement of the GR, using siRNA (Fig. 1F, S1E), which confirmed the need for GR (Fig. 1G), a conclusion strengthened by complementation assays using siRNA-resistant HaloTag-GR (Fig. 1H, S1E,F). In addition, scratch wound assays of A549 cells demonstrate the inhibitory effect of the endogenous glucocorticoid hydrocortisone in addition to dex and RU486 on cell migration, suggesting a ligand-dependent mechanism of action ( Fig. S1C).
GCs also inhibit the migration of many other cell types. To test the broader applicability of our findings we used primary peritoneal macrophages from GR f/f and LysM-creGR f/f mice (Fig. S1F). These cells show remarkable inhibition to GC, but this inhibitory effect is completely lost in the GR null cells.

Alpha stable distribution models A549 motion
The distribution of step lengths in vehicle and dex treated conditions (Fig. 1C) showed the distinctive walk pattern indicative of an alpha stable distribution, characterised by four parameters that describe the stability exponent (α), skewness (β), scale (γ), and location (δ) (Salas-Gonzalez et al., 2013;Burnecki et al., 2012). Glucocorticoid treatment reduces median step length, signified by a left-shift in the frequency distribution curve (measured by a reduction in δ parameter). Alpha stable parameters were derived using MATLAB, showing that A549 cell movement adopts an alpha stable distribution irrespective of glucocorticoid treatment (Fig. S2A, B). These changes in parameters show that the movement of vehicle-treated A549 cells primarily consists of small steps occasionally interspersed with larger relocating or searching steps. Glucocorticoid alters these parameters inhibiting the low frequency, large displacement searching movements.

Synthetic, selective GR ligands exhibit similar effects to conventional GC
In view of the inhibitory effect of RU486 on cell migration, the study was extended to further non-steroidal GR ligands with unique pharmacological properties (Trebble et al., 2013;Schiller et al., 2014), and we selected a panel based on high affinity and specificity ( Fig. 2A). For example, GRT7 extends into the meta channel of the GR LBD driving slower kinetics of activation, but more potent transcriptional induction (Fig. 2B,C,D) (Trebble et al., 2013). GW870086X (086X), is a selective GR modulator (SeGRM), deficient in transactivation function (Fig 2d). All the GR ligands tested similarly reduced A549 cell displacement (Fig. 2E), affecting both total displacement (086X=188.5µm; vehicle=208.3µm) (GRT7=208.3µm; vehicle=224.1µm) and median step length (086X=2.28µm; vehicle=2.58µm) (GRT7=1.95µm; vehicle=2.44µm). Cell walk properties were similarly inhibited (Fig. 2F,G,S2C,D,E,S3E,F). As with dex, RU486 did not antagonise the inhibition of cell migration with GRT7 or O86X (GRT7+RU486 median total displacement=192.6µm; GRT7+RU486 median step length=1.86µm; 086X+RU486 median total displacement=201.3µm; 086X+RU486 median step length=2.26µm). In the study with GRT7 we did see a slight potentiation of the inhibitory effect when RU486 was added to the GRT7 compound, but in this series of studies the RU486 effect was slightly less than seen in other repeats when we analyse total cell displacement.
However, the reduction in median step length was striking, and consistent throughout. Additional analysis was conducted using the high potency, steroidal GR agonist fluticasone propionate (FP), with similar effects on movement (Fig. S3A, B,C) and alpha stable distribution parameter changes confirming altered walk properties (Fig. S3D,E,F).

GC inhibit migration independent of gene transcription
Through dynamic cell tracking (Fig. S1C), we noted that the GC effect was of rapid onset, an observation not identified previously using fixed end-point assays (Fig. S1A,B). When we analysed displacement at earlier time points there was a significant reduction by 12 hours, and even by 4 hours a trend was seen, although this did not reach significance (Fig. S4A, B,C). To investigate the time course of glucocorticoid response in greater detail we used a non-parametric rank-sum test to determine the earliest time-point at which cell migration is significantly reduced following treatment with each GC and compared this to the dynamics of ligand-induced GR nuclear translocation (Fig. 3A, B,C). Dex and RU486 both inhibit migration within 60 minutes of administration; and GRT7 which does not translocate the GR until 3 hours post treatment also inhibits migration by 60 minutes. O86X, which induces the most rapid GR translocation only inhibits migration after 5 hours (indicated by the coloured arrows, Fig.   3C). The kinetics of nuclear translocation were inverse to those for migration inhibition. The rapid onset of action, with response preceding nuclear translocation as in the case of the GRT7 ligand, suggested a non-conventional mechanism of action, such as a cytoplasmic, non-genomic circuit. A non-conventional mechanism of action is also supported by the lack of antagonism seen with the use of RU486 in the presence of the agonists, despite the absolute requirement for the presence of the GR.
In order to test the importance of new gene transcription for the altered cell migration we used actinomycin D pretreatment. One hour actinomycin D was enough to block the transcriptional activation function of the GR (Fig.   S5A), and these studies again confirmed the complete lack of agonist activity seen with RU486 in these cells, under these conditions. Under these treatment conditions we were able to see the same change in cell migration with a four-hour incubation with Dex either with, or without actinomycin D blockade of gene transcription (Fig. 3D). Actinomycin D did not affect cell migration independently.
We also profiled changes in gene expression of a panel of genes known to control cell migration. In this study we again selected the four hour time point as the most discriminating, as the change in cell migration by this point does not require a change in gene expression. Here, we observed only two genes (PLAUR and BCAR1) to be downregulated by a 4 hour incubation with dex. This repression was opposed by RU486, as expected for a conventional GR antagonist. As the two genes were regulated as conventional glucocorticoid target genes this implies that they are not relevant to the migration phenotype we are observing, as we show that no new gene transcription is required for the effect, and also that the migration phenotype is both observed with RU486 treatment, and also that RU486 treatment does not oppose the actions of agonists such as Dex when the cell migration phenotype is observed ( Fig. 3E).

GC treatment rapidly stabilises microtubules
To investigate the mechanism explaining the early onset, non-transcriptional response of cell migration to glucocorticoid treatment we profiled the activation status of a panel of candidate proteins (Fig. 4A). We observed a reduction in phosphor-ERK, and an induction in phosphor-Ezrin/Radixin/Moesin, but these changes were not seen until 24h post glucocorticoid. This suggests that these molecules are not involved in mediating the very rapid responses we see in response to glucocorticoid treatment, and their change in status may follow on from rather than drive the change in cell phenotype. To analyse rapid responses, we turned to live-cell imaging of the cytoskeleton, to determine if changes in cytoskeletal architecture could be observed.
To investigate the actin and MT cytoskeletal networks we examined the effect of dex on MT dynamics using GFPtagged MT plus-end (+TIP) binding protein, EB3 (Fig. 4B). GC drives an increase in overall MT growth speed (vehicle median=17.2μm/min; dex median=17.7 μm/min), consistent with stabilisation of microtubules ( Fig. 4C; total). To analyse the impact of GCs on MT dynamics in more depth the growth speed data was subdivided into thirds, designated slow, medium and fast (Fig. 4C,D). Dex resulted in an increased median speed (vehicle=26.2μm/min; dex =27.6μm/min) in the fast speed events, with a decreased speed (vehicle=11.1μm/min, dex=10.8μm/min) observed in the slow speed events, while no significant difference was detected at medium speed ( Fig. 4C). This indicates a shift to more rapidly polymerising microtubules in the presence of dex. A concomitant increase in the frequency of fast growth speed events was also detected ( Fig. 4D; 40μm/min onwards). We attempted direct measurement of acetylation of α-tubulin, a marker of microtubule (MT) stability. We did see small increases at early time points (<10 minutes), but the effect size was small, and on further investigation the effect was not robust in replication. Therefore, we do not have convincing evidence that tubulin acetylation changes in response to glucocorticoid, but the changes in microtubule growth were very robust, and highly significant.

GC alter microtubule dynamics by inhibiting HDAC6
Acetylation of tubulin is tightly controlled by the α-tubulin acetyltransferase αTAT1 and the tubulin deacetylase HDAC6, making these two enzymes candidate effectors. We tested αTAT1 knockdown cells (Fig. 5A), but found no effect, suggesting an alternative mechanism of GR action (Fig. 5B). Tubacin, a selective HDAC6 inhibitor, not only mimicked the inhibitory effect of dex (Fig. 5C), but showed no additive effect in co-treatment protocols, suggesting convergent mechanism of action. Therefore, we analysed the effect of augmenting HDAC6 expression ( Fig. 5D), which increased both the displacement and median step length of cells and rendered cells resistant to GC (Fig. 5E,F). Alpha stable distribution parameters changed in response to HDAC6 overexpression with cells adopting a higher proportion of large walk steps indicating increased cell migration relative to the controls, which was unchanged following administration of dex (Fig. 5G,H). A pan-HDAC activator ITSA1 also reversed the GC migration phenotype, confirming the contributory role of HDAC6 in this mechanism (Fig. 5I).

GR and HDAC6 are complexed together in the cytoplasm
There was no evidence of altered HDAC6 sub-cellular trafficking in response to dex with the enzyme remaining predominantly cytoplasmic (Fig. S6A). Co-immunoprecipitation studies also failed to identify GR and HDAC6 in complex together (Fig. S6B), despite previous reports of co-binding and interactive function on gene repression in the nucleus (Rimando et al., 2016;Govindan, 2010). However, we did detect a change in HDAC6 interactions with actin components of the cytoskeleton in response to GC (Fig. S6C,D), although these correlative studies do not provide evidence for physical, or functional interaction, they do suggest molecular proximity.
To study the GR-HDAC6 interaction in further detail, we employed real-time fluorescence cross-correlation spectroscopy (FCCS) able to discriminate between cellular compartments (marked by crosses, Fig. 6A). We identify a cytoplasmic but not nuclear interaction between GR and HDAC6 as compared to an empty fluorophore negative control (Fig. 6B, C,D,E,F). An approximate five-fold increase in interaction strength post-dex after GR ligand activation was estimated through determination of in vivo dissociation constant (Kd) values (Fig. 6G,H,I,J).

Discussion
Although therapeutic GCs are widely used, their diverse actions limit long-term safety. Multiple candidate mechanisms of action have been advanced, with the major focus on how the same activated receptor can both repress and activate different genes in a cell-type specific context. Gene repression has been a focus of study, as this pathway appears to mediate the beneficial anti-inflammatory and immune suppressive actions of GCs. To this end, new partial agonist GR ligands have been developed and tested in the clinic. Such selective GR modulators (SeGRMs) differentiate GR function mainly by affecting the ligand-bound GR conformation, and thereby recruitment of co-modulators. However, GR can also affect other cellular processes through a non-transcriptional pathway, for example mitotic spindle function (Matthews et al., 2011). One major and consistent effect of GC treatment is loss of tissue integrity, and impaired wound healing. In part, this results from reduced epithelial, macrophage, and fibroblast migration (McDougall et al., 2006;Hardman et al., 2005). This programme has not received much attention but may serve as a model to understand the distinct actions of GCs. Therefore, we used an epithelial cell model to measure migratory responses to GC.
Our initial studies sought to mathematically model the walk properties of cells under basal conditions, to provide a solid baseline for GC comparison. Our cell walk characteristics fitted an alpha stable distribution, and the impact of GC altered the parameters in such a way that longer steps were selectively reduced in favour of shorter steps, thereby impairing the searching behaviour of cells. This real-time, individual cell tracking permitted the kinetics of response to be measured, and here the surprising finding was the very rapid onset of action with significant deviation from control cells within 40 minutes of treatment. This rapid onset of action was similarly seen with the GR antagonist RU486 and with further non-steroidal ligands. That RU486 failed to oppose the agonist effects raised a question of specificity, which was addressed in siRNA studies in epithelial cells, and by replicating the migration assay in macrophages, which permitted genetic loss of GR to be tested. The rapid onset of effect, and paradoxical full agonist phenotype seen with RU486, suggested an unconventional mechanism of action, which was supported by showing that no new mRNA synthesis is required. Although we have previously detected inhibition of Rac1 activity soon after GC treatment in podocytes, in that study we were unable to establish that the Rac1 effect was required for the GC effect, and the time for a Rac1 inhibitor to reduce cell movement was prolonged (>10 hours), and also strikingly different to that seen with GC exposure (McCaffrey et al., 2017). For these reasons we did not pursue a role for Rac1 in the current work. Here, we also employed selective GR ligands with well-characterised differences in GR nuclear translocation kinetics in order to gain further insights into mechanism of action. Here we showed that rapid nuclear translocation did not associate with rapid inhibition of migration, but rather a GR ligand with a markedly slow GR translocation rate was still able to affect cell migration rapidly, even while predominantly residing within the cytoplasm. This is consequently a well-documented model of a truly non-genomic mechanism of GR action.
In pursuit of the mechanism of action for rapid changes in cell motility induced by glucocorticoid receptor activation we analysed changes in the cytoskeletal architecture, using real-time imaging of microtubule growth. Microtubule kinetics are inverse to cell migration velocity, with increased microtubule polymerisation making cells less able to change conformation, and to migrate. We selected a four hour time exposure to analyse microtubule polymerisation, as we had shown that by this time point significant changes in cell migration were seen, and there were no changes in gene transcription that could plausibly affect cell migration, also that blockade of new gene transcription had no impact on the glucocorticoid effect. Microtubule polymerisation is driven by increased tubulin acetylation (Hubbert et al., 2002;Boggs et al., 2015;Zhang et al., 2003). Tubulin acetylation is tightly controlled by the opposing actions of the α-tubulin acetyltransferase-1 (αTAT1) and the deacetylase HDAC6 (Castro-Castro et al., 2012;Liu et al., 2012). HDAC6 is localised predominantly in the cytoplasm where it directly interacts with microtubules and catalyses tubulin deacetylation along the length of the microtubule track (Asthana et al., 2013;Miyake et al., 2016). Moreover, HDAC6 has shown to prefer deacetylation of tubulin dimers over polymerized microtubules (Skultetyova et al., 2017).
Here, we show that a non-conventional GR mechanism of action is responsible to the rapid onset of cell migration inhibition. We also see the same mechanism at play at later time points, with RU486 failing to inhibit the migratory phenotype even out to 24 hours. It is hard to exclude the involvement of additional mechanisms of action at later time points, as with all studies of glucocorticoid action the longer exposures result in a greater number of responses, as secondary, and tertiary events come into play. We cannot block gene transcription for long periods of time as this is toxic to the cells, and therefore makes the cellular phenotype hard to interpret, but we can use the RU486 to block transcriptional activity of the GR. In these cells and under these conditions RU486 is essentially without any agonist activity. RU486 alone exerts a similar impact on cell migration compared to the full agonists, identifying a non-transcriptional mechanism of action.
The GR rapidly translocated to the nucleus after addition of glucocorticoid, but despite its mainly nuclear location there is evidence for very rapid cycling on and off recognition sites in DNA, and shuttling between the cytoplasm and nucleus, so providing an explanation for persisting engagement with the HDAC6 enzyme. We observed GR:HDAC6 interaction in our FCCS studies which showed cytoplasmic interaction one hour after ligand addition, when bulk GR is mainly nuclear (Paakinaho et al., 2019). Therefore, taken together we have evidence of an unconventional mechanism of GR action which does not involve target gene transactivation. We do, however, find a novel interaction between the activated GR and HDAC6, and show that the HDAC6 is required to mediate the glucocorticoid effect.
Our studies identify a previous unrecognised mechanism of GR action with involvement of a protein-protein interaction circuit targeting HDAC6. We were able to show the HDAC6 dependence of the GC loss of long-step length migration and the rapid cellular response, coupled with a lack of requirement for new gene transcription pointed to a direct mechanism of action with a pathway connecting activated GR and the HDAC6 protein. We were not able to show HDAC6-GR interaction by co-immunoprecipitation, but FCCS studies identified a fraction of the cytoplasmic HDAC6 pool as interacting with GR, with resultant changes in movement kinetics, implying a change in molecular complex formation. The unique cytoplasmic preference for GR amongst the nuclear receptors may explain its capacity to interact with cytoplasmic enzymes such as HDAC6. Our data support a GR-driven change in HDAC6 behaviour as the mechanism explaining rapid-kinetic loss of cell movement in response to GC exposure.
Defective cell migration in response to GC has widespread consequences including defective tissue repair, and loss of barrier function. Identification of a new mechanism of GC action has implications for attempts to design novel GR ligands, with reduced off-target effects, but also the screening for potent GR ligands capable to engaging this pathway to treat exuberant wound-healing, such as keloid.
The identification of a coherent non-genomic GR mechanism of action leading to a clinically relevant cell migratory phenotype offers new insight into the diversity of GC action. This pathway underlines the difficulty in developing specific anti-inflammatory GR ligands, exemplified by identical action of GR antagonists and agonists on cell migration. In addition, the inhibition of epithelial cell migration is also observed in macrophages, providing a new insight into the anti-inflammatory and immune suppressive functions of GCs, which have largely been focused on chemokine production, adhesion molecule expression, cell survival, and enzyme production. Taken together we elucidate a newly discovered non-genomic pathway of GC action affecting cell migration, with proximal impacts on tissue integrity, repair, and innate immune function.
siRNA: AllStars Negative Control siRNA (SI03650318), GR siRNA (SI02654764), and αTAT1 siRNA (S104145162) were purchased from Qiagen. Promega. GRT7 and GW870086X were developed by GlaxoSmithKline. Unique materials used are available from the authors or from standard commercial sources outlined above.

Chemotaxis Migration Assay
Chemotaxis migration assay was performed in 24-well Millicell hanging cell culture inserts (Millipore, MCEP24H48) with an 8µm polyethylene terephthalate membrane pore. A549 cells were pre-conditioned to 100nM dex or vehicle control (DMSO) for 48 hours (37°C/5% CO2). Cells were suspended in serum-free DMEM and seeded into the upper chamber of the Transwell insert (2.5x10 4 cells/well). The lower chamber was filled with FBS to act as the chemoattractant. 100nM dex or vehicle control was added to the upper and lower compartments of the transwell.
The cells are incubated for 24 hours (37°C/5% CO2) to allow chemotaxis to occur. Following incubation the cells were fixed in 4% paraformaldehyde (PFA) for 15 minutes at RT. Any cells that did not migrate were removed from the upper side of the membrane with a cotton swab. Cells are stained with crystal violet (5mg/ml in 2% ethanol) for 30 minutes at RT. The inserts were washed twice in dH2O and excess stain was removed mechanically from the upper side of the membrane. The migrated cells were solubilised in 2% SDS overnight at room temperature and absorbance was read at 560nm using a Glomax plate reader (Promega). Chemotaxis was quantified as a percentage relative to the vehicle control.

Cell Stopper Migration Assay
Migration assay was performed using an Oris 96-well plate with Oris Cell Seeding Stoppers (Platypus Technologies, CMA1.101) according to the manufacturer's instructions. A549 cells were seeded in DMEM + 10% FBS (1x10 5 cells/well) into an Oris 96-well plate containing Oris Cell Seeding Stoppers and incubated for 18 hours (37°C/5% CO2) to allow attachment. Following incubation, stoppers were removed and cells washed with 1xPBS.
The media was replaced with DMEM + 10% cFBS and cells were treated with 100nM dex or vehicle control (DMSO). The cells were incubated for 48 hours to allow migration into the detection zone to occur (37°C/5% CO2).
Reference wells had Oris Cell Seeding Stoppers left in place to act as the no migration controls. Cells were washed with 1xPBS to remove any debris/unattached cells, fixed in 4% PFA for 40 minutes at 4⁰C, and stained with Hoeschst (2µg/ml; #14533; Sigma) for 5 minutes at room temperature to label DNA. Images were collected on an Axio Observer A1 (Axiovision) inverted microscope using a 2.5x/0.07 Plan-Apochromat objective and captured using a Axio Cam HRc (Axiovision) through MetaVue Software (Molecular Devices). Specific band pass filter sets for DAPI were used to prevent bleed through from one channel to the next. Images were processed and quantification of migration was achieved using ImageJ (http://rsb.info.nih.gov/ij) (Schneider et al., 2012). Images were thresholded for high intensities, converted to binary, then analysed as particles to determine the area of the detection zone covered with cells, using the no migration controls for reference. Migration was then quantified as the percentage of the detection zone covered with cells, relative to the vehicle control.

Scratch Wound Healing Assay
A549 cells were seeded in DMEM + 10% cFBS (2x10 4 cells/well) into a 96-well ImageLock plate (Essen Bioscience, #4379) and allowed to adhere for 24 hours (37°C/5% CO2). Simultaneous, uniform scratch wounds were induced in each well with the WoundMaker tool (Essen Bioscience) and wells were washed twice in DMEM + 10% cFBS to remove debris. Cell migration into the wound was acquired immediately following administration of a dose response of dex, RU486, and hydrocortisone (0.1nM, 1nM, 10nM, 100nM, 1µM, 10µM) along with the vehicle control (DMSO). Images were taken at 30 minute intervals over 24 hours (37°C/5% CO2) on an Incucyte Zoom Live-Cell Analysis system using a 10x/0.3 Plan Fluor OFN25 (DIC L/N1) objective in brightfield. Cell migration was analysed and quantified using Incucyte Zoom software.

Transfection
Transfections were performed with Fugene 6 reagent (E2691; Promega) used at a ratio of 3:1 volume/weight ratio with DNA. Fugene 6 was pre-mixed with RPMI media (serum-free) for 5 minutes prior to incubation with DNA for 15 minutes at room temperature. Transfections were performed over 24 hours at 37°C/5% CO2. Small interfering RNA (siRNA) transfections were performed with Lipofectamine RNAiMAX reagent (#13778150; ThermoFisher Scientific) as described in the manufacturer's instructions and performed over 48 hours at 37°C/5% CO2.

Live-cell Immunofluorescent Migration
A549 cells were seeded onto glass-bottomed 24-well plates (#662892; Greiner) at 25,000 cells per well in CSM and left to adhere overnight. Cells were transiently transfected with 0.5µg pBOS-H2B-GFP and left to incubate for 24 hours at 37°C/5% CO2. Cells were treated with vehicle (DMSO) or dex (100nM) and live-cell imaging was performed for 48 hours using a Nikon TE2000 PFS microscope. Images were acquired every 5 minutes using a 20x Plan Apo objective and the Sedat filter set (Chroma 89000). Cells were maintained at 37°C/5% CO2 throughout imaging. The images were collected using a Coolsnap HQ (Photometrics, USA) camera and raw images were processed using ImageJ.

RT 2 Profiler-PCR Array
A human cell motility RT 2 Profiler PCR array (384-well plate) was used to assay gene expression changes following glucocorticoid treatment (PAHS-128Z; Qiagen). Cells were treated as required, then lysed and RNA extracted using an RNeasy kit including the on-column DNase digestion step (#74104; Qiagen). 400ng RNA was reverse transcribed (#330401; Qiagen) and cDNA samples were added to the reaction plate and real-time PCR acquired using an ABI qPCR machine (Applied Biosystems, CA, USA). Cycle threshold (CT) values were exported and analysed according to manufactures instructions by RT 2 profiler PCR array data analysis software (http://dataanalysis.qiagen.com/pcr/arrayanalysis.php ; Qiagen). 5 housekeeping genes were assessed and the most suitable, RPLP0, was selected for normalisation of gene expression. A complete data set is provided in the supplementary data, presented as a fold change over the vehicle control and 95% confidence interval for each gene analysed.

Quantitative RT-PCR
A549 cells were treated as required, then lysed and RNA extracted using an RNeasy kit including the on-column DNase digestion step to remove genomic DNA (#74104; Qiagen). 1µg of RNA was reverse transcribed (#4387406; Applied Biosystems) and analysed by qPCR using Sybr Green detection. Gene expression was assessed using the following primer pairs Glyceraldehyde-3-phosphate dehydrogenase (GAPDH), Forward Primer, 5'-

HaloTag pulldown assay
A549 cells were seeded onto 15cm 2 dishes at 300,000 cells per mL in CSM and left to incubate for 24 hours at 37°C/5% CO2. Cells were transiently transfected with either HaloTag-GR (10µg) or HaloTag-HDAC6 (10µg) and left to incubate for 24 hours. Cells were treated with vehicle (DMSO) or dex (100nM) for 1 hour and washed 2x with ice-cold 1xPBS. Cells were gently scraped into conical tubes and centrifuged at 2000xg for 10 minutes at 4°C.
Supernatant was discarded and cell pellets were stored overnight at -80°C prior to lysis. Before cell lysis, HaloTag resin (G1912; Promega) was mixed to obtain uniform suspension and 200µL resin was dispensed into 1.5mL Eppendorf tubes per treatment condition. Tubes were centrifuged at 800xg for 1 minute, supernatant discarded, and resuspended in 800µL resin equilibration buffer (100mM Tris-HCl, 150mM NaCl, 0.005% IGEPAL CA-630).
Tubes were centrifuged at 800xg for 2 minutes and supernatant discarded. Resin was washed an additional 3x in equilibration buffer. Cell pellets were thawed on ice and resuspended in 300µL lysis buffer (50mM Tris-HCl pH7.5, 150mM NaCl, 1% Triton X-100, 0.1% sodium deoxycholate) supplemented with 6µL of 50x protease inhibitor cocktail (800µ/mL benzamidine HCl, 500µg/mL phenanthroline, 500µg/mL aprotinin, 500µg/mL leupeptin, 500µg/mL pepstatin A, 50mM PMSF). Cells were passed 5x through a 25G needle to complete lysis and centrifuged at 14000xg for 5 minutes at 4°C. Cleared lysates (300µL) were transferred to new 1.5mL tubes and diluted in 700µL 1xTBS (100mM Tris-HCl pH7.5 and 150mM NaCl). 1mL diluted lysates were mixed with the washed HaloTag resin and left to incubate overnight on a tube rotator at 4°C. Tubes were centrifuged at 800xg for 2 minutes and supernatant discarded. Pellets were washed 4x in resin equilibration buffer. After the last wash, resin was resuspended Luciferase reporter gene assay HeLa cells were seeded onto 10cm 2 dishes at 50,000 cells per well in CSM and left to adhere overnight. Cells were transiently transfected with luciferase tagged-mouse mammary tumour virus (MMTV-Luc; 2µg) or luciferase tagged-nuclear factor-κB response element (NRE-Luc; 2µg) using Fugene 6 reagent (3:1 volume/weight ratio with DNA) for 24 hours. Cells were re-seeded onto 24-well plates at 50,000 cells per well in CSM and left to adhere overnight at 37°C/5% CO2. Cells were treated as specified in the results and 18 hours later each well was washed twice with 1xPBS. 100µL of Bright Glo lysis buffer (Promega, E2620) was added to each well and left to lyse on ice for 30 minutes. Cell lysates were transferred to a white, flat-bottomed 96-well plate and luciferase absorbance was read using a luminometer (Glomax, Promega). 10x 1-second reads were taken of each well and the average relative light unit (RLU) determined. Background wells were included that only contained lysis buffer.
IC50 and EC50 values were extrapolated from the resulting dose response curves using non-linear regression analysis in GraphPad Prism software, with the following equation: Y=Bottom + (Top-Bottom) / (1+10^((LogIC50-X)*HillSlope)). Where X: log of dose or concentration; Y: Response; Top and Bottom: Plateaus; LogIC50 interchangeable with LogEC50; HillSlope: Slope factor or Hill slope, unitless.

Fixed-Cell Immunofluorescent Imaging
A549 cells were seeded in DMEM + 10% cFBS (5x10 4 cells per coverslip) and allowed to adhere for 24 hours (37°C/5% CO2). Cells were treated with vehicle or dex (100nM) for 48 hours. Cells were fixed with 4 % PFA for 40 minutes at 4˚C and blocked (0.1% Triton X-100, 1% FBS in PBS) for 1 hour at room temperature. Remaining incubations were performed at room temperature unless stated otherwise. Coverslips were incubated with primary antibody diluted in blocking buffer overnight at 4°C. After three 5 minute washes in PBS, coverslips were incubated with flourophore-conjugated secondary antibody for 2 hours. Cells were subsequently stained with rhodaminephalloidin in PBS (2 µg/ml) for 15 minutes and then Hoeschst in PBS (2 µg/ml) for 5 minutes. Following four 5 minute washes in PBS, coverslips were mounted using Vectamount AQ (Vector Labs, H-5501). Images were acquired on a Delta Vision RT (Applied Precision, GE Healthcare) restoration microscope using either a 40X/0.85 Uplan Apo objective or a 60X/1.42 Plan Apo N objective and the Sedat Quad filter set (Chroma 86000v2, VT, USA).
The images were collected using a Coolsnap HQ (Photometrics, AZ, USA) camera with a Z optical spacing of 0.5μm. Raw images were then deconvolved using the Softworx software (GE Healthcare) and maximum intensity projections of these deconvolved images processed using Image J.
Fluorescence cross-correlation spectroscopy: A549 cells were seeded onto glass bottomed 35mm dishes (#627965; Greiner) at 25,000 cells per well in CSM and left to adhere overnight. Cells were transiently cotransfected with HaloTag-GR (500ng) and HDAC6-eGFP (500ng) using Fugene 6 reagent (3:1 v/w ratio with DNA) for 24 hours. Six hours post-transfection, cells were treated with 100nM HaloTMR Direct ligand (G2991; Promega) overnight to visualise HaloTag-GR. The following morning, cells were washed once with CSM before being treated with GC. FCCS was performed using either a Zeiss LSM780 or LSM880 equipped with GaAsP detectors using a plan-apochromat 63x/1.4 NA oil objective. EGFP was excited with 488nm (0.3%) laser light and emission collected between 500 and 530nm. Rhodamine/mCherry was excited with 561nm (0.3%) laser light and emission collected between 580 and 630nm. Single-point 5x5 second runs FCCS measurements were taken within the cytoplasm and nucleus of individual cells. Zen 2.3 software was used for data collection and correlation-curve fitting. A two-component 3D diffusion model using a fixed structural parameter (S=4) with or without triplet states were fit for autocorrelation and cross-correlation curves respectively. FCCS measurements were rejected if counts per molecule (CPM) were <1 kHz, or if they exhibited >10% photobleaching in either green or red channels. Structural parameter values of 4 and an effective confocal volume size of 0.57 fL were previously measured allowing for determination of molar concentrations [55]. The 'fit' and 'confint' functions in MATLAB 2018a were used to determine in vivo dissociation constants and confidence intervals according to the

Live-cell Microtubule Dynamics
Microtubule dynamics were monitored in A549 cells transiently transfected with 0.25µg of EB3-GFP and 0.75µg pcDNA3 using Fugene 6 reagent and treating with vehicle or dex (100nM) for 4 hours prior to imaging. Cells were selected at random based on the expression of EB3-GFP and images were captured every 0.5 seconds for 1 minute on a Nikon TE2000 PFS microscope using an apo-TIRF 100x/1.49 NA oil objective. The images were collected using a Cascade II EMCCD camera (Photometrics) with a Z optical spacing of 0.2μm. Raw images were then processed using Image J (Schneider et al., 2012).

Image Analysis
Live-cell pBOS-H2B-GFP tracking: A549 cell movement was tracked using an ImageJ plugin, Mosaic, based on pBOS-H2B-GFP expression. Tracking was performed following manufactures instructions. Co-ordinates of the tracks and the corresponding movies including tracks were exported, from which step length, the distance moved between each image acquisition, and total displacement, the overall distance each cell moved where calculated.
For step length, the median step length for each cell was calculated over the duration of tracking and then displayed graphically or a frequency distribution curve of all step lengths from every cell was generated.
Rose-plots: 20 cells were chosen at random to be displayed as a rose plot. Co-ordinates were transformed so every cells track originates from the X Y co-ordinates (0, 0). Rose plots were generated in GraphPad Prism.

Mathematical analysis
Frequency distributions of step lengths were parameterised and fitted to alpha stable distribution using the STBL: Alpha stable distributions functions package for MATLAB (available at https://www.mathworks.com/matlabcentral/fileexchange/37514-stbl-alpha-stable-distributions-formatlab). The parameters estimated were for media step length (µm), stability exponent (α), skewness (β), scale (γ), and location (δ). A standard resampling strategy was used to validate the parameters. Random sampling with replacement from the original data sets was performed to generate 100 subsets of 15,000 values. These subsets were then parameterised according to an alpha stable distribution to derive robust estimates of the standard deviations. The functions package was used to calculate probability distribution functions (PDF) of the empirical data and from the determined parameters, and plot them on the same axis.

Live-cell brightfield tracking
A549 cell movement was tracked using Imaris 8.0 software (Media Cybernetics, ltd). Migration was tracked using an autoregressive motion algorithm from cells filtered by size (25 µm) and from tracks filtered by minimum movement speed (above 2.5 µm/minute) to discount stationary debris. Cell migration was depicted as step length between each time point, which was determined using Pythagoras: SQRT of a 2 + b 2 = c, where a is the position along the x axis, b is the position along the y axis, and c is the step length.
The plusTipGetTracks tool was used to detect and track fluorescently-labelled MT plus end binding proteins (+TIPs), with only MT growth events being detected in both vehicle and dex treated cells. Overlay images showing the tracks for MT growth speed were generated with the plusTipSeeTracks tool. Raw data was collected from the gs_fs_bs_gl_fl_bl_gd_fd_bd.txt generated by the plusTipGetTracks tool and combined; frequency of comet growth speeds were determined. Data was sub-divided into slow, medium, fast and very fast comets and histograms were generated in GraphPad Prism.

Statistical analysis of cell movement data
Cumulative distance data for varying-length trajectories (~1000 cells in each experiment, measurements every 10 minutes up to 24 hours) for cells treated with GW870086X, dexamethasone, GRT7, and RU486 were used to perform statistical test on the median reduction in cumulative distance travelled against corresponding vehicletreated cells.
For any pair, e.g., dexamethasone vs vehicle, and at each time point, two distributions of cumulative distances are compared using a non-parametric rank-sum test (using MATLAB), reporting significance (at α=0.0001) of median reduction under treatment (e.g. dexamethasone) w.r.t vehicle against the null hypothesis that medians are equal in both distributions.

Acknowledgements
Special thanks to Peter March, Roger Meadows and Steven Marsden for their help with the microscopy.

Competing Interests
The authors declare no competing interests.

Funding
Stephen Kershaw is supported by an MRC studentship. Bioimaging Facility microscopes used in this study were purchased with grants from BBSRC, Wellcome Trust and the University of Manchester Strategic Fund.
DWR is supported by MRC programme grant MR/P023576/1 and is also a Wellcome Investigator: Wellcome Trust (107849/Z/15/Z).

Data Availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.  (D)- (E) Cell migration data for A549 cells that were tracked using brightfield microscopy. Cells were treated with vehicle (DMSO) and dexamethasone (dex; 100nM) and images acquired every 10 minutes for 24 hours, with data displaying all 24 hours of tracking.

Contributions
(G) -Cell migration data for A549 cells that were tracked using brightfield microscopy. Cells were transiently transfected with siRNA targeting GR or a non-targeting siRNA negative control for 48 hours. GR knockdown cells were then treated with vehicle (DMSO) or dex (100nM) and images acquired every 10 minutes for 6 hours, with data displaying all 6 hours of tracking.
(H) -Cell migration data for A549 cells that were tracked using fluorescence microscopy based on rhodamine expression. GR knockdown cells were transiently transfected with HaloTag empty vector (2µg)  (G) Violin plots of total displacement (µm) and median step length (µm) of control and GR knockdown A549 cells in response to 6 hours of vehicle and dex (100nM) treatment. Migration data shown as median ± IQD and represents all cells analysed over three independent experiments (Kruskal-Wallis non-parametric test followed by Dunn's multiple comparisons test; ****p<0.0001, **p=0.0012, *p=0.0342).
(C) Non-parametric rank-sum test of A549 cell migration (cumulative distance) signifying the earliest timepoint at which migration is statistically different in response to dex (100nM), RU486 (100nM), GRT7 (3nM), and 086X (100nM) compared to vehicle-treated controls (D) Violin plots of total displacement (µm) and median step length (µm) of A549 cells in response to 1 hour pretreatment with vehicle or actinomycin D (1µg/mL) and subsequent 4 hour treatment with vehicle or dex (100nM). Migration data shown as median ± IQD and represents all cells analysed over two independent experiments (Kruskal-Wallis non-parametric test followed by Dunn's multiple comparisons test; ****p<0.0001, **p=0.003, *p=0.01).  (C) Violin plots of microtubule growth speed (µm/min) in A549 cells following a 4 hour vehicle and dex (100nM) treatment. Growth speed data is shown entirely (upper left panel; total) or split into tertiles according to speed (designated slow, medium or fast). Each bin size corresponds to one-third of all growth speed events following vehicle treatment, and was applied to the dex treated data set. Data is representative of 3 independent experiments. 9149 (vehicle-treated) and 9669 (dex-treated) growth events, from 10 cells per condition were tracked and analysed.
Median ± IQD (Mann Whitney non-parametric test; ***p=0.0002, ****<p=0.0001). (D) Histograms displaying the frequency of microtubule growth speeds (µm/min) in A549 cells following a 4 hour vehicle and dex (100nM) treatment Growth speed data is shown entirely (upper left panel; total) or split into tertiles as described in (C). Data for vehicle treated cells is indicated by red bars and for dex treated cells by blue bars. (C) -Cell migration data for A549 cells that were tracked using brightfield microscopy. Cells were treated with vehicle (DMSO), dex (100nM), or tubacin (100nM) and tracked over 24 hours. Images were acquired every 10 minutes, with data displaying all 24 hours of tracking.
(E) -Cell migration data for A549 cells that were tracked using fluorescence microscopy based on GFP expression. Cells were transiently co-transfected with H2B-GFP (0.25µg) and HDAC6-FLAG (0.25µg) for 24 hours. Cells were then treated with vehicle (DMSO) or dex (100nM), and images were acquired every 10 minutes for 24 hours, with data displaying 24 hours of tracking.
(I) -Cell migration data for A549 cells that were tracked using fluorescence microscopy based on GFP expression. Cells were treated with vehicle (DMSO) or ITSA1 (5µM) in combination with vehicle (DMSO), dex (100nM), or RU486 (100nM) for 24 hours. Images were acquired every 10 minutes for 24 hours, with data displaying 24 hours of tracking.
(A) Representative images of control and αTAT1-knockdown A549 cells stained with immunofluorescence for αTAT1 (green) and F-actin (red). Nuclei were stained with DAPI (blue). αTAT1 knockdown was siRNA- (H) Probability density function (PDF) plots of experimental step length data (black line) and PDF plots generated from estimated alpha stable distribution parameters (dark grey) in response to vehicle and dex (100nM) treatment. Plots generated in MATLAB.
(I) Violin plots of total displacement (µm) and median step length (µm) of A549 cells in response to 24 hours vehicle, dex (100nM), RU486 (100nM), ITSA1 (5µM)+vehicle co-treatment, ITSA1 (5µM)+dex (100nM) cotreatment, and ITSA1 (5µM)+RU486 (100nM) co-treatment. Migration data shown as median ± IQD and represents all cells analysed over two independent experiments (Kruskal-Wallis non-parametric test followed by Dunn's multiple comparisons test; ****p<0.0001, ***p=0.0009). Cell migration data for cells that were transiently transfected with 0.5µg pBOS-H2B-GFP and incubated for 24 hours at 37°C/5% CO2 prior to imaging. Cells were treated with vehicle and dexamethasone (dex; 100nM) and images acquired every 5 minutes for 48 hours. Data displayed represents subsampling of migration data from Expression of the GR target genes FKPB5, TSC22D3 and PDK4 was analysed by RT-qPCR. Data is normalised to the housekeeping gene GAPDH and displayed as a fold change over the vehicle control. Graphs (mean ±SD) combine data from at three independent experiments. Statistical analysis was performed by oneway ANOVA followed by Holm-Sidak's multiple comparison (**p<0.005; ****p<0.0001) ImarisColoc software (Bitplane), which utilises the Costes algorithm to automate co-localisation. Table S1: RT 2 Profiler PCR array data (A) A549 cells were treated with a vehicle control, dex (100nM), RU486 (100nM) or a combination of dex and RU486 for 4 hours. RNA was extracted and 400ng was reverse transcribed. Expression of 84 genes associated with cell migration was assessed via a human cell motility RT 2 Profiler PCR array. Data is normalised to the housekeeping gene RPLP0 and is displayed as a fold change over the vehicle control, and includes a 95% confidence interval. Fold change values greater than 2 (upregulated) are in red and fold change values less than 0.5 (downregulated) are indicated in blue. Table S1 J. Cell Sci.: doi:10.1242/jcs.242842: Supplementary information