V-ATPase Inhibition Decreases Mutant Androgen Receptor Activity in Castrate-resistant Prostate Cancer

Prostate cancer was diagnosed in 1.28 million men and caused 359,000 deaths in 2018 (1). Initially, prostate cancer cell growth is dependent on androgen receptor (AR) activation. Advanced disease is therefore treated with permanent androgen deprivation therapy (ADT), to limit cancer cell growth (2). However, subsequent progression from a hormone-sensitive prostate cancer (HSPC) to a more aggressive castrate-resistant prostate cancer (CRPC) phenotype is almost inevitable. Metastatic CRPC treatment options include hormonal therapy (abiraterone, enzalutamide), chemotherapy, and radium-223. However, CRPC outcomes remain poor with a median survival of around 2 years (2).

CRPC prognosis is partly attributed to persistent AR signaling through adaptive alterations to the AR itself. AR alterations are relatively rare at prostate cancer diagnosis but become more prevalent through ADT/hormonal therapy exposure and represent a key mechanism for transition to CRPC (3). Gain-of-function point mutations, particularly in the ligand-binding domain (LBD), may provide growth advantage despite a castrate state. For example, the F877L AR mutation converts some second-generation antiandrogens (e.g., enzalutamide, apalutamide), to agonist function (4). The F877L mutation has been observed in enzalutamide-resistant prostate cancer xenograft models (5) and detected in plasma DNA of patients with CRPC previously treated with apalutamide (4). T878A is a further well-characterized AR point mutation (6) that converts AR antagonists (e.g., flutamide) to agonists (7), and broadens AR ligand binding specificity, for example to progesterone (8). This mutation was found in abiraterone-treated patients with CRPC producing a progesterone-activated abiraterone-resistant state (9).

AR splice variants (AR-Vs) are truncated AR forms, resulting in ligand-independent constitutive activation, through loss of an LBD portion and represent a further CRPC transition mechanism (10). The most common, AR-V7, is associated with enzalutamide and abiraterone resistance (11), and reduced survival (12). AR-V7 expression is heterogeneous, and may heterodimerise with full-length AR, or form homodimers, to bind androgen response elements (AREs) to facilitate a protumorigenic transcriptome (13). AR-V7 mRNA expression has been found 20 times higher in patients with CRPC than in patients with HSPC (14), and protein levels of AR-V7 have been shown to increase on development of CRPC (15).

Vacuolar-ATPase (V-ATPase) is a multiprotein complex that catalyzes ATP-dependent proton transport across intracellular and plasma membranes. The resulting acidification of organelle lumens or extracellular space influences a diverse range of cellular processes, many of which are dysregulated in cancers (16). V-ATPase is composed of two domains, the cytosolic V₁ domain and the integral membrane complex, V_O. V₁, which is responsible for ATP hydrolysis, is composed of eight subunits (A–H), whereas V_O, responsible for proton translocation across the endomembrane, is composed of five (a, c, c”, d, e; ref. 17). The V₁ domain is tethered to the V_O domain via the V₁ subunit V₁C, which is important in regulating enzyme disassembly (18). In addition to its proton pump function, V-ATPase is now understood to have a central role in sensing environmental signals, including pH (19), and is required for amino acid mediated mTORC1 signaling (20).

Limited data indicate a potential role of V-ATPase in prostate cancer and, in particular, on AR modulation. V-ATPase expression correlates to invasive and metastatic potential of prostate cancer cells as evidenced by: (i) its expression in the plasma membrane (PM), (ii) V-ATPase–dependent activation of proteases (e.g., MMP-2, -9, -14), and (iii) increased cell motility (21). Evidence suggests that the V_Oa3 isoform (and potentially V_Oa1, V_Oa2, and V_Oa4) plus the V-ATPase accessary subunit Ac45, are key in targeting V-ATPase to the PM in prostate cancer cells (22). In addition, individual subunit isoform knockdown resulted in reduced proton flux and V-ATPase function (22), and chemical inhibition caused alkalization of endolysosomal compartments (23). V-ATPase modulatory proteins such as LASS2/TMSG1 and PEDF can affect tumor cell growth and invasion via regulation of V-ATPase activity (24, 25). Chemical inhibition of V-ATPase in HSPC (LNCaP) and CRPC (C4-2B) cell lines reduced in vitro invasion of both (26). In addition, V-ATPase colocalized with PSA and V-ATPase inhibition resulted in a PSA relocalization to lysosome-like intracellular vesicles (26). Furthermore, V-ATPase inhibition in LNCaP and LAPC4 cells depleted both AR protein and mRNA (23).

These data support a potential link between AR and V-ATPase function in prostate cancer. We provide data to explore this interaction, including for altered AR forms relevant for clinical transition to CRPC.

HEK-293 and LNCaP/DuCaP/22Rv1 cells were maintained in DMEM (Sigma) and RPMI medium (Sigma), respectively, supplemented with 10% FBS with 4 or 2 mmol/L l-glutamine, respectively, and 1 mmol/L pyruvate. Where indicated, media were changed at 24 hours to phenol red-free RPMI1640 (Life Technologies) supplemented with 10% charcoal-stripped FBS (CSS-RPMI; Life Technologies). LNCaP F877L/T878A and LNCaP control empty plasmid cells were a gift from Novartis (5). All cell lines were regularly Mycoplasma tested and passaged between 15 and 30 times for all experiments. Enzalutamide was purchased from Selleckchem, bafilomycin-A1 (baf-A1) from Melford and DHT and concanamycin-A (con-A) from Sigma-Aldrich. Compounds were dissolved in DMSO. Cell viability was assessed using the MTS assay (Sigma) as per the manufacturer's instructions.

Briefly [as described previously (27)], HEK-293 cells were transfected with the ARE reporter plasmid p(ARE)₃Luc, wild type of mutant/variant AR expression plasmids (pEAR-WT/pEAR-V7/pEAR-Q641X/pEAR-F877L) and pRL-CMV (control Renilla luciferase expression plasmid) using Fugene HD transfection reagent (Promega). The AR-V7 and AR-Q641X expression plasmids were a gift Jocelyn Ceraline, University of Strasbourg (Strasbourg, France; ref. 28). The dual luciferase reporter assay (Promega) was performed using the Varioskan Flash Multimode Reader (Thermo Fisher Scientific). Results show firefly luciferase activity normalized to Renilla luciferase activity as a transfection efficiency control.

Total mRNA was extracted using the ReliaPrep RNA Cell Miniprep System (Promega) following the manufacturer's instructions and quantified using a NanoDrop (Thermo Fisher Scientific). For cDNA synthesis, M-MLV Reverse Transcriptase (Promega) followed manufacturer's instructions. For qRT-PCR, TaqMan Universal PCR Master Mix, No AmpErase UNG (Thermo Fisher Scientific) utilized commercially available TaqMan Gene Expression Assays probes for KLK3 (PSA), TMPRSS2 and the reference gene GAPDH. The 7500 Real-Time PCR thermal cycler machine (Applied Biosystems) was used for the qRT-PCR reaction via thermocycling and detection of subsequent fluorescence. The Real-Time qPCR software v2.0.6 (Applied Biosystems) then calculated C_t values and the comparative 2^−ΔΔC(t) method as described previously (27) to calculate fold changes in gene expression.

Whole-cell lysates were collected in 1× RIPA buffer and fragmented using sonication. Proteins were quantified using the Bradford assay (Promega), separated using SDS-PAGE, transferred to a nitrocellulose membrane (Amersham Protran 0.45 NC, GE Healthcare) and blocked in phosphate-buffered saline with 0.1% Tween-20 with 5% milk solution. Membranes were probed with primary antibodies to AR, PSA (Cell Signaling Technology), ATP6V1C2 (Santa Cruz Biotechnology), ATP6V1C1, ATP6V1A (Invitrogen), and β-Actin (Sigma) followed by horseradish peroxidase–conjugated anti-mouse or anti-rabbit IgG secondary antibodies (Sigma). Detection was with enhanced chemiluminescent reagents (Thermo Fisher Scientific). Immunoblot signals were quantified using ImageJ (http://imagej.nih.gov/ij/) and protein expression was normalized to β-Actin.

Experimental [ON-TARGETplus Human ATP6V1C1 (528) siRNA - SMARTpool and ON-TARGETplus Human ATP6V1A (523) siRNA - SMARTpool) and nontargeting siRNA (ON-TARGETplus Non-targeting Pool) were diluted with 1× siRNA buffer (Dharmacon). Cells were reverse transfected using Dharmafect reagent (Dharmacon) following manufacturer's instructions. Cells were incubated in siRNA containing media for either 48 or 72 hours prior to harvesting.

A stable ATP6V1C1 (ATP6V1C1-202; ENST00000518738.2) knockout was generated using the protocol of Zhang and colleagues (29). Briefly, the single-guide RNA (sgRNA) sequence: GGACTGCTTGATTGGATATT TGG was ligated into the plasmid pSpCas9(BB)-2A-Puro (PX459) V2.0 (Addgene plasmid #62988; http://n2t.net/addgene:62988; RRID:Addgene_62988). Respective oligonucleotide sequences (Sigma) were: 5′ CACC GGGACTGCTTGATTGGATATT 3′, and 5′ AAAC AATATCCAATCAAGCAGTCCC 3′. Cells were transfected with the sgRNA-expressing plasmid using Fugene HD (Promega) with puromycin (Sigma) selection (1.5 μg/mL) for 48–72 hours. A single-cell clone was isolated and validated using Western blotting and Sanger sequencing.

Where indicated either two-way ANOVA with Tukey multiple comparison post hoc test or two-tailed Student t tests were used to generate P values and determine statistical significance of the indicated differences. P values under 0.05 were considered statistically significant. Values were plotted using GraphPad Prism 7 and displayed as mean values ± SD. Biological triplicates were used unless otherwise stated.

Androgen sensitivity and dependence is a characteristic of HSPC. We first determined the noncytotoxic concentrations of the V-ATPase inhibitors baf-A1 (30) and con-A (31) in the HSPC LNCaP cell line (Fig. 1A), as well as in HEK-293 (Fig. 1B) and the androgen-insensitive 22Rv1 cell line (Fig. 1C). We then measured the effect of V-ATPase inhibition on wild-type AR (AR-WT) transactivation, using the HEK-293 cell line which does not express the AR. HEK-293 cells were transfected with an ARE reporter plasmid, an AR-WT expression plasmid, and a Renilla transfection control plasmid. Following 24-hour exposure to 1 nmol/L DHT, and a range of concentrations of the V-ATPase inhibitors baf-A1or con-A, levels of ARE gene activation were detected via luminescent quantification.

Consistent with previous research (26), nanomolar concentrations of baf-A1 and con-A significantly reduced AR-WT activity (Fig. 2A). We then tested the impact of V-ATPase inhibitors in the LNCaP HSPC cell line model. V-ATPase inhibition using con-A (Supplementary Fig. S1) or baf-A1 (Fig. 2B; Supplementary Fig. S2) reduced expression of downstream AR targets KLK3 (PSA) and TMPRSS2 at the transcript level.

At the protein level, V-ATPase inhibition also reduced AR expression (Fig. 2C) consistent with prior data in LNCaP and LAPC4 HSPC cells (23). Of note, despite reduced AR protein expression being a consistent finding, we found that V-ATPase inhibition had variable effects on PSA protein expression. In some experimental conditions, we observed PSA protein increase, despite consistent PSA depletion at the transcript level (see Supplementary Fig. S3).

To further investigate the potential link between V-ATPase and AR signaling, we used the AR antagonist enzalutamide (which binds the AR LBD) in LNCaP cells. When V-ATPase inhibition was combined with AR antagonism, there was a significant reduction of both AR downstream transcript expression (Fig. 2D) and AR protein expression (Fig. 2E).

To address the impact of V-ATPase function on altered AR forms, we first utilized an AR-F877L model system for CRPC. Following validation of relative enzalutamide resistance and AR-F877L protein expression (Supplementary Figs. S4 and S5), we found that at nanomolar concentrations, V-ATPase chemical inhibition reduced AR-F877L transactivation activity in a same manner to AR-WT (Fig. 3A).

We also investigated LNCaP cells containing the AR-F877L/T878A double mutant treated with both enzalutamide and baf-A1. In these cells, AR signaling was significantly reduced compared with controls at both the mRNA (Fig. 3B) and protein (Fig. 3C) level, indicating V-ATPase inhibition could overcome enzalutamide resistance in an endogenous AR system. To validate these results, the experiments were repeated using a partner expression plasmid LNCaP cell line containing only the endogenous T878A mutation (Supplementary Fig. S6). Results using this control cell line aligned as expected with what we had previously observed in nonmodified LNCaP cells (as shown in Fig. 2E), confirming that F877L was driving the enzalutamide resistance.

Using the HEK-293 reporter assay system, V-ATPase chemical inhibition led to significant reduction in AR-V7 mediated ARE activation (Fig. 4A) despite substantially higher baseline activation than for AR-WT. Similar results were obtained (Fig. 4B) with a plasmid expressing AR-Q641X [p.(Gln641Xaa)] which contains a nonsense mutation in the hinge region of the AR, resulting in a constitutively active protein (32). We also found that AR expression was reduced with V-ATPase inhibition in the AR-V7–expressing 22Rv1 cell line at both the mRNA (Fig. 4C) and protein level (Fig. 4D).

Many V-ATPase subunits have tissue-specific isoforms, with some functioning independently of the complete complex to regulate various signaling pathways (33). Various subunit isoforms are overexpressed in different cancer types compared with healthy tissue (16). This led us to question whether targeting one of these isoforms could lead to reduced V-ATPase activity, and consequently a depletion of AR activity.

The regulatory V₁C1 subunit, encoded by ATP6V1C1, was selected for further investigation as its overexpression in different cancer types has been linked to functional defects (34, 35). Structurally, the V₁C subunit connects the V₁ and V_O domains, with detachment otherwise. In yeast, this detachment results in a loss of V-ATPase activity in a regulatory process known as reversible dissociation (36). In addition, the V₁C1 isoform was selected over V₁C2; because V₁C1 is ubiquitously expressed, whereas V₁C2 is tissue specific (37). Interestingly, data from The Human Protein Atlas indicates that the protein levels of the V₁C1 isoform is low in prostate tissue, whereas the V₁C2 isoform is expressed at high levels (38).

First, we depleted V₁C1 by siRNA knockdown to determine the effect on AR protein in 22Rv1 cells. As earlier experiments (described in Fig. 4) demonstrated that AR-V activity is correlated to expression, only endogenous AR protein expression was investigated. To assess the specificity of effects of V₁C1 knockdown, the V₁A subunit was also depleted. The V₁A subunit has no other isoforms, and thus is an essential catalytic component of the V-ATPase complex (18). Therefore, V₁A depletion should reduce overall V-ATPase complex expression and activity. In addition to this, the expression of the alternative isoform V₁C2 was also measured in response to V₁C1 knockdown.

After 48 hours of ATP6V1C1 siRNA transfection, AR-WT protein levels were reduced (Fig. 5A). Moreover, after 72 hours, AR-WT and AR-V7 levels were substantially reduced compared with the nonspecific siRNA-transfected controls. In addition, AR-WT depletion was accompanied by a significant increase of V₁C2 expression, suggesting a possible compensatory role.

To investigate this further, CRISPR-Cas9 was used to eliminate V₁C1 expression. A single 22Rv1 CRISPR-Cas9 V₁C1 knockout clone (V₁C1 k/o) was validated by Sanger sequencing. In the V₁C1-depleted cells, AR-WT and AR-V7 expression was comparable with plasmid only control cells (Fig. 5B). In addition, V₁C2 protein levels were increased in the knockout cells compared with plasmid-only control cells.

Treatment options for HSPC, based on AR inhibition, are usually initially highly effective. However, once the cancer becomes castrate resistant, with AR activity independent of systemic androgenic stimulation, treatment options are limited (39), with cross-resistance against AR-directed drugs (40). We therefore investigated whether V-ATPase inhibition would impact AR signaling in prostate cancer driven through altered AR expression. Previous data had established a link between V-ATPase and AR signaling in HSPC (23). Here, we show that the effects of V-ATPase inhibition extend beyond HSPC models, and that V-ATPase inhibition can alter aberrant AR signaling in the form of AR-Vs and mutant AR forms.

AR-Vs, such as AR-V7 and AR-Q641X, are associated with progression to CRPC through constitutively active AR function. Moreover, functional AR-activating mutations such as F877L and T878A may result in a broadening of the LBD and resistance to clinically relevant antiandrogens including enzalutamide. We have shown that V-ATPase inhibition reduced AR-V7, AR-Q641X, and AR-F877L transactivation. In addition, V-ATPase inhibition resulted in a reduction in AR protein expression in 22Rv1 cells and LNCaP cells expressing the inducible F877L/T878A double mutant. Together, these data strengthen evidence of a link between V-ATPase and AR signaling. Most importantly, it shows that V-ATPase inhibition results in a reduction of AR signaling, which is independent of AR aberration expression. Therefore, the potential exists for V-ATPase to be developed as a therapeutic target for CRPC.

The mechanism for reduced AR function resulting from V-ATPase inhibition is complex. It had been shown that inhibiting V-ATPase in prostate cancer cell lines resulted in reduced mRNA levels of downstream AR target genes such as KLK3 (the gene for PSA). Paradoxically, this was accompanied by an increase in PSA protein levels (26). Our data validate this in both the hormone-sensitive LNCaP and DuCaP cell lines. A primary function of V-ATPase is to acidify intracellular vesicles to enable functions such as receptor endocytosis and vesicular transport (41). The increase in PSA protein may potentially be explained therefore by inhibition of these key processes although the possibility of other mechanisms is accepted. Moreover, as PSA is secreted and TMPRSS2 is not, it is probable that this effect is linked to protein secretion, which occurs via the ER/Golgi pathway. It was previously suggested that V-ATPase inhibition would disrupt this secretory pathway and causes alkalinization of transport vesicles leading to an accumulation of intracellular PSA. This accumulation of PSA might then mask downregulation of PSA mRNA resulting from potentially therapeutic V-ATPase inhibition (26). This might imply that PSA level would not be reliable as a means of therapeutic monitoring for V-ATPase inhibition in prostate cancer.

It has also been suggested that the direct effect of V-ATPase inhibition on AR levels involves hypoxia-inducible factor-1 α (HIF1α). The proposed mechanism suggests V-ATPase inhibition results in defective transferrin receptor recycling due to alkalization of endolysosomal compartments. This in turn blocks iron uptake and reduces HIF1α hydroxylation, resulting in an increase in HIF1α stability. HIF1α is then free to translocate to the nucleus and downregulate AR expression (23). It is also of note that knockdown of specific V-ATPase subunits (Ac45, V_Oa1, and V_Oa3) reduced transferrin receptor recycling to the plasma membrane (22). Hence targeting these subunits might reduce AR activity via a decrease in intracellular iron concentration and a reduction in HIF1α hydroxylation. However, it was also found that coincubation of con-A with iron significantly reduced HIF1α protein levels, thus conflicting the suggested mechanism (23). Therefore, it is likely that the interactions between V-ATPase and the AR are multifaceted, involving several signaling pathways, and further research is required to conclusively determine the mechanism behind such interactions.

Transient versus sustained loss of the V₁C1 subunit had substantially different effects on AR expression in 22Rv1 CRPC cells. A transient reduction (using siRNA) resulted in reduced AR expression, whereas permanent loss of V₁C1 (using CRISPR) resulted in no substantial changes in AR expression. One possible explanation, is that V₁C2 compensates for permanent loss of V₁C1 and can maintain AR signaling in 22Rv1 cells. In support of this, it was recently found that ATP6V1C1 and ATP6V1C2, which encode the two V₁C isoforms, are differentially expressed in tumors from patients with esophageal squamous cell carcinoma (ESCC) (42). The expression levels of the isoforms were comparable in normal esophageal tissues but in ESCC tumors ATP6V1C1 was overexpressed and displayed tissue dominance over ATP6V1C2 (42). Therefore, despite being described as having low protein expression in normal prostate tissue (38), it is possible that the V₁C1 subunit is overexpressed in prostate cancer cells and loss of this isoform leads to a compensatory upregulation of the V₁C2 isoform to maintain V-ATPase function. As with chemical intervention, the compensatory increase in V₁C2 expression may result in an increase in cell survival, which is of potential concern and is an important consideration for the future development of this strategy.

In addition to this, prior data exist to support the concept of V-ATPase subunit isoform compensation for another. Kawamura and colleagues generated a Cre-lox genetically modified mouse model lacking the neuronal specific V₁G2 isoform with no obvious detriment to brain architecture or phenotype (43). In this model, the V₁G1 ubiquitously expressed isoform was found to accumulate to larger amounts than in a wild-type mouse model. Interestingly, despite an increase in V₁G1 protein, there was no increase in V₁G1 mRNA, indicating that a loss of function of V₁G1 was compensated by V₁G2 without mRNA upregulation (43). In addition, another group demonstrated that the ubiquitously expressed V₁B2 isoform can compensate for the loss of the kidney-specific V₁B1 isoform in medullary A intercalated cells. Apical V₁B2 immunostaining was 2-fold higher in a V₁B1 null mouse model compared with one positively expressing V₁B1. The compensated V₁B2 complexes were also able to maintain 28%–40% of normal V-ATPase activity, which was sufficient to maintain acid-base homeostasis in V₁B1-deficient mice (44). Another study investigating the role of V_Oa3 in phagosome acidification found that mice deficient in V_Oa3 still exhibit V-ATPase–dependent acidification, albeit to a lesser degree than in wild-type mice (45), indicating that V_Oa1 and V_Oa2 isoforms could at least partially rescue V-ATPase function.

Alternatively, as V₁C1 is involved in V-ATPase regulation (41), it is plausible that a transient reduction of the V₁C1 subunit causes aberrant V-ATPase regulation, consequently impacting upon other signaling pathways such as mTORC1. This transcriptional cross-talk with AR remains poorly understood but might represent another way the V-ATPase is linked to AR signaling. Although not investigated in this article, due to the interactions between V-ATPase and mTORC1, it also cannot be ruled out that the V-ATPase complex, or at least its components, contribute directly to AR signaling. For example, McConnell and colleagues proposed that reducing V₁C1 expression may reduce V-ATPase assembly and prevent mTORC1-mediated cancer cell growth due to a failure of mTORC1 to receive amino acid signals (35).

Because of ubiquitous expression, targeting the V-ATPase complex as a whole would likely be toxic with off target effects. However, these data support that targeting specific subunit isoforms might be an efficient method of reducing AR signaling. For example, subunits such as Ac45 and the V_Oa isoforms may represent optimal targets to prevent V-ATPase–mediated trafficking to downstream effector proteins (22), whereas perhaps transiently reducing the V₁C1 subunit would reduce AR signaling via alternative pathways such as mTORC1 (35). Further research is required to ascertain which subunit isoforms are of consequence in CRPC.

In this study, we have shown that combining AR inhibition (with enzalutamide) with chemical inhibition of V-ATPase has an additive effect on reducing AR functional activity. Combinatorial therapy might represent an approach to overcome acquired resistance observed in CRPC to existing AR-directed treatments. A key area for future development would be in V-ATPase inhibitor candidates with drug-like properties.

To summarize, V-ATPase inhibition can reduce AR signaling in prostate cancer cells. Of significance, we have demonstrated, for the first time, that this extends to both point mutant and splice-variant AR forms that are clinically relevant for transition to CRPC. This suggests that V-ATPase could be targeted as a way to overcome AR signaling in patients with AR aberrations, including the significant unmet clinical need that exists for patients with CRPC.

G. Packham reports grants from The Urology Foundation, Wessex Medical Research, The Gerald Kerkut Charitable Trust, and Cancer Research UK during the conduct of the study. S.J. Crabb reports personal fees from Roche, Janssen Cilag, MSD, Pfizer, and Bayer; grants and personal fees from AstraZeneca; and grants from Astex Pharmaceuticals and Clovis Oncology outside the submitted work. No potential conflicts of interest were disclosed by the other authors.

B. Whitton: Conceptualization, data curation, software, formal analysis, investigation, methodology, writing–original draft, writing–review and editing. H. Okamoto: Conceptualization, resources, formal analysis, supervision, funding acquisition, investigation, methodology, project administration, writing–review and editing. M. Rose-Zerilli: Formal analysis, supervision, investigation, writing–review and editing. G. Packham: Conceptualization, resources, formal analysis, supervision, writing–original draft, writing–review and editing. S.J. Crabb: Conceptualization, resources, formal analysis, supervision, funding acquisition, investigation, writing–original draft, project administration, writing–review and editing.

B. Whitton was funded by a doctoral studentship from The Urology Foundation, Wessex Medical Research, and The Gerald Kerkut Charitable Trust. This work was also supported by grants from Cancer Research UK (C2750/A23669).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.