Pro-Oncogenic Transcription Factors BACH1 and Nrf2 Associate with Cytoplasmic Biomolecular Condensates of GFP-MxA in Oral Cancer Cells

1. Introduction

Oral cancer is the sixth-most prevalent human malignancy worldwide, and has a significant 5-year mortality of 45-50% [1,2,3,4]. Insights into the pathogenesis of this cancer have been typically obtained from patients already diagnosed with cancer followed by the physician’s attempt to unravel the patient’s history [1,2,3,4,5,6,7,8]. Such studies have uncovered preventable causes such as tobacco or betel nut/pan leaf chewing, smoking and alcohol consumption using products of usually of local manufacture containing “addictive carcinogens” (e.g., arecoline in betel nut) [1,2,3,4,5,6,7,8,9]. In these instances the “exposome” includes the buccal mucosa of the cheek or the lower lip (where the quid is held) and the floor of the mouth along the path of liquid transit [1,3]. Overt causes also include chronic mechanical irritation of the mucosa due to dental prostheses (e.g., ill-fitting dentures)[1]. Long-term consumption of beverages at or above 150°F have also been associated with an increased cancer risk, as for the upper esophagus [10,11].

Importantly, a significant subset of patients with oral cancer reveal no overt causes [4]. In a pioneering study of such cases published 50 years ago, Mashberg and Myers enumerated development of oral cancer in approximately 60,000 asymptomatic patients presenting at a dental clinic over 12 years with a focus on the anatomic locations of their oral cancer [4]. They identified a U-shaped “high-risk” zone comprising the floor of mouth, especially around the papilla of the submandibular duct, side of tongue, anterior pillar of fauces and retromolar region. They reported that 97% of 207 intraoral malignant lesions detected “prospectively” in such asymptomatic patients occurred in this “high-risk” zone (especially see Figure 3a,b in reference [4]). The authors called for “increased scrutiny” of the underlying cell biology of the epithelium of this region [4]. They, and others, noted that this region comprised the pathway of liquid (and beverage) flow through the mouth, that the lining epithelium in these regions was thin, largely non-keratinized and relatively more permeable than other regions of the mouth [3,4,12,13]. The cellular basis for the anatomic localization of oral cancer occurrence in this high-risk zone of cancer development in the mouth in the absence of any overt causes remains incompletely understood [1,2,3,4]. Changes in the oral microbiome, including bacterial, fungal and viral agents (including papilloma and Epstein–Barr viruses) [2,3,7,8,9], and citations therein), enhanced heme oxygenase 1 (HO1) and molecular drivers such as pro-oncogenic p-STAT3, mutations in p53 and Rb, and activity of the E6 protein of human papilloma virus (reviewed in [2,3,14,15,16]) have received attention. The present study addresses novel discoveries in biomolecular condensate biology in cells likely applicable to this high-risk zone [17,18,19,20,21,22,23,24,25].

Liquid-liquid phase-separated (LLPS) biomolecular condensates in the cytoplasm and nucleus are now recognized to contribute to carcinogenesis through dysregulation of signaling pathways, and aberrant function of fusion oncoproteins and transcription factors [17,18,19,20,21,22,23,24,25]. Indeed, condensate disassembly and spontaneous reassembly during cellular stress responses, the regulation of transcription and translation by condensate droplets, cancer pathogenesis through aberrant prooncogenic signaling, involvement of condensates in mechanisms of innate and adaptive immunity, cytokine signaling, viral replication and antiviral mechanisms, and the targeting of condensates by cancer therapeutic agents have been highlighted by numerous investigators [17,18,19,20,21,22,23,24,25].

Our previous studies have focused on the dynamic alterations of the structure and biology of cytoplasmic condensates of interferon (IFN)-induced human antiviral “myxovirus resistance protein” (MxA alias Mx1), a dynamin-family large GTPase (approx. 60 kDa [19,24,25,26,27,28]. Cytoplasmic human MxA and nuclear murine Mx1 exhibit antiviral activity towards influenza A virus (IAV) and additional RNA and DNA viruses [27]. In our previous studies, membraneless cytoplasmic human MxA condensates showed rapid disassembly in 1–2 min in cells exposed to hypotonic medium, and rapid reassembly (also in 1–2 min) in cells shifted back to isotonic medium [24,25,27,28]. In as much as saliva exiting the submandibular duct (in the middle of the high-risk zone of oral cancer development, [4]) is hypotonic (approximately 100 mOsm, i.e., approximately one-third tonicity compared to plasma) ([3,4,12,13], and citations therein), we wondered whether hypotonicity- and beverage-like temperature driven changes in condensate cell biology in the mouth might drive pathogenesis of oral cancer. Since MxA is known to be constitutively present in normal human gingiva [29], we used GFP-MxA as a reporter of condensate dynamics in oral cells [25,28,30]. Based on our previous data about the rapid and reversible dynamics of MxA condensates in oral epithelial cells [25,28,30], we suggested a novel hypothesis for the anatomic localization of oral cancer in the U-shaped “high-risk” zone in the mouth: that dysfunction of diverse biomolecular condensates in oral cells along the beverage transit pathway through the mouth due to repetitive tonicity and temperature stresses that might underlie a prooncogenic progression leading to oral cancer [25].

In the present study we identified some of the constituents of GFP-MxA cytoplasmic condensates in oral cancer cells by isolating GFP-MxA structures from prefixed interferon (IFN)-λ1-treated GFP-MxA expressing OECM1 human oral cancer cells using magnetic-bead based immunoisolation, followed by unbiased peptide/protein family identification. The new data highlight the unexpected discovery that MxA condensates contain the transcription factors BACH1 (BTB and CNC homology1) and Nrf2 (nuclear factor erythroid 2-related factor 2), and also their downstream responder protein heme oxygenase 1 (HO1). Taken together, these proteins are well known to be involved in cellular responses to environmental and oxidative stresses, antiviral defense, oral epithelial dysplasia, and in cancer progression and metastases [2,31,32,33,34,35,36,37,38,39,40,41,42]. The new data provide support for the hypothesis that the tonicity- and temperature-driven dynamic repetitive changes (disassembly/reassembly) of condensates involving prooncogenic transcription factors in oral cells located along the liquid transit pathway through the mouth might contribute to carcinogenesis.

2. Materials and Methods

2.1. Cells and Cell Culture

Human oral carcinoma cell line OECM1 was purchased from Millipore-Sigma (St. Louis, MO, USA), while the human lung adenocarcinoma cell line A549 was obtained from the American Type Culture Collection (Manassus, VA, USA). The respective cell lines were grown in DMEM (Corning Cat. No. 10-013-CV, with glutamine, Na-pyruvate and high glucose) supplemented with 10% v/v fetal bovine serum (FBS; Gibco, Grand Island, NY, USA) in T25 flasks, 35 mm dishes without or with cover-slip bottoms [24,28,30].

2.2. Plasmids and Transient Transfection

The GFP (1-248)-tagged full-length human MxA and GFP-murine Mx1 constructs were gifts from Dr. Jovan Pavlovic (University of Zurich, Zurich, Switzerland) [24,43]. HA=tagged MxA was a gift from Dr. Otto Haller (University of Freiburg, Germany) [24]. EGFP-tagged human BACH1 construct was purchased from Genescript, Piscataway, NJ, USA) while the EGFP-tagged human Nrf2 construct was obtained from Addgene, Watertown, MA, USA. Transient transfections were carried out using the Polyfect reagent (Qiagen, Germantown, MD, USA) and the manufacturer’s protocol [24,28,30].

2.3. Fluorescence Imaging

Live-cell imaging of GFP-MxA structures in transiently transfected cells was carried out in cells grown in 35 mm plates using the upright the Zeiss AxioImager 2 equipped with a temperature-regulated stage (range: 5–50 °C) and a 40× water immersion objective with data capture in a manual time-lapse mode (using Axiovision 4.8.1 software) as described earlier [24,28,30]. Additionally, imaging was also carried out by placing a coverslip on the sheet of cells or droplets containing GFP-MxA condensates and imaging using a 100× oil immersion objective [24,28,30]. Moreover, confocal imaging, including z-stack imaging, was carried out by growing cells in 35 mm plates with glass cover-slip-bottoms, formaldehyde fixation followed by imaging with a 63x oil immersion objective and Zeiss LSM980 confocal microscope.

2.4. Phase Transition Experiments

GFP-MxA expressing cells two days after transient transfection were exposed to fresh full growth medium or phosphate-buffered saline (PBS) at 37 °C and imaged (the “0 min” time). Subsequently the cultures were shifted to 1:3 or 1:4 dilution of the medium (adjusted with water) or hypotonic ELB (50 mOsm) for 4- 5 min and then either fixed using 4% paraformaldehyde (PFA) (for 1 hr at room temperature in the hypotonic buffer0, or live cultures switched to full strength medium or to PBS for 5 min, then imaged, and then fixed using 4% PFA in full strength medium. Fluorescence was imaged as summarized above [24,28,30].

2.5. Magnetic-Bead Based Immunoisolation of GFP-MxA Condensates and Unbiased Proteomics

We showed earlier that, as with stress granules [44], the integrity of GFP-MxA condensates requires that the cells remain intact [24,28,30]. Thus, we used the methods developed by Parker and colleagues to isolate stress granules from PFA prefixed cells using magnetic-bead-based immunoisolation followed by trypsin digestion, mass-spectroscopy and unbiased peptide identification [44]. Briefly GFP-MxA transfected cells in 35 or 90 mm culture dishes were additionally treated with IFN-λ1 (50 ng/mL) for 2 days, fixed at room temperature using 2% PFA in PBS for 10 min, washed with PBS, the PFA quenched with 1 M glycine in ELB for 10 min, the cultures washed in cold PBS, then harvested by scraping into cold “GG” buffer (1 mL/plate)(20 mM TrisHCl, pH 7.4, 100 mM NaCl, 1 mM EDTA, 0.1% SDS, 0.5% Triton X-100, 1 mM PMSF). The suspension was sonicated for 45-60 sec, the debris centrifuged (1000 rpm, 5 min) and droplets of the supernatant verified for presence of GFP-MxA condensates by fluorescence microscopy. Aliquots of the supernatant fraction were then mixed with anti-MxA pAb or irrelevant pAb (such as anti-protein disulfide isomerase, PDI, pAb), mixed with 50 µL of a GG-buffer washed, 1:5 diluted Protein A-Dynabead suspension (30 mg/mL original suspension, Cat. 2848440, Invitrogen, Baltics, Norway) followed by incubation overnight at 4°C. The beads were then washed 4x using cold PBS and magnetic separation, the pellets air dried and used for tryptic digestion, mass-spectroscopy and unbiased peptide identification by comparison with a human peptide database by Creative Proteomics (Shirley, NY, USA). The signal intensity of respective peptide families were expressed in arbitrary intensity units (Figure 1D).

2.6. Immunofluorescence Assays

The most intense 8-10 such peptide family identifications were tested for association with GFP-MxA condensates in PFA-fixed (4%, 1 hr) transfected cells using two-color immunofluorescence assays as described earlier [24,28,30]. Fluorescence data were collected as mentioned briefly in Section 2.3 above. Between 5-20 cells were imaged per variable. Numerical analyses of colocalization were carried out using Image J FIJI version x86.64 in terms of the Pearson’s correlation coefficient R.

2.7. Western Blotting

Aliquots of prefixed or unfixed cell extracts prepared as outlined above in Section 2.5 were used to immunoisolate complexes reactive with respective antibodies (overnight incubation with Protein A-Dynabeads) followed by washing 3-4x with PBS and resuspension in 2x Laemmli sample loading buffer. Electrophoreses were carried out using 4-20% Mini-PROTEAN TGX precast gels (BioRad, Hercules, CA, USA) or 4-15% SMOBio QPAGE (Stellar Scientific, Baltimore, MD, USA) precast gels followed by Western blotting [24,28,30]. Proteins were displayed using relevant primary rabbit antibodies and CF488A-tagged donkey anti-rabbit IgG from Biotium Inc. (Cat. 20015)(Freemont, CA, USA) with the assistance of an iBright1500 Invitrogen Imaging system (Thermo-Fisher Scientific, Tewkesbury, MA, USA).

2.8. Antibody Reagents

Rabbit pAb to human MxA (H-285) (ab-95926) was purchased from Abcam Inc. (Cambridge, MA, USA). Rabbit pAbs to human BACH1 (Cat. No. 14018-1-AP), to human HO1 (Cat. No. 10701-1-AP), human Nrf2 (Cat. No. 16396-1-AP) and to human p62 (Cat. No. 18420-1-AP) were purchased from Proteintech Group Inc., (Rosemont, IL, USA). Rabbit pAbs to human protein disulfide isomerase (PDI; Cat. No. sc-20132) and to clathrin heavy chain (CHC; Cat. No. sc- 9069) were obtained from Santa Cruz Biotechnology Inc., Dallas, TX, USA). Respective AlexaFluor 488- and AlexaFluor 594-tagged secondary donkey antibodies to rabbit (A-11008 and A-11012) or mouse (A-21202 and A-21203) IgG were from Invitrogen Molecular Probes (Eugene, OR, USA) for immunofluorescence assays, or CF488-tagged donkey anti-rabbit IgG from Biotium Inc. (Cat. 20015)(Freemont, CA, USA) for Western blots.

3. Results

3.1. Isolation of GFP-MxA Condensates and Unbiased Peptide Identification

As reported by us earlier [25,30], live-cell fluorescence imaging of transiently expressed wild-type GFP-MxA in OECM1 cells exposed to hypotonic buffer (75-100 mOsm approximately corresponding to saliva-like/beverage-like hypotonicity) confirmed the rapid disassembly of GFP-MxA condensates within 2-4 minutes followed by spontaneous reassembly of condensates by 7-10 min even when cells were continuously kept under hypotonic conditions (Figure 1A). Moreover gently permeabilizing cells with 0.05% saponin led to rapid disassembly of GFP-MxA condensates [24]. Thus, as with the methods of Parker and colleagues for isolation of stress granules [44], it was necessary to prefix cells with 2% PFA for 10 min prior to cell harvesting and sonication. Preliminary observations in transfected A549 cells confirmed that such prefixation allowed GFP-MxA structures to remain intact. The sonicated supernate was clarified by low-speed centrifugation and GFP-MxA condensates immunoisolated using anti-MxA pAb and Protein A-Dynabeads using magnetic pullout. Figure 1B shows low and high magnification images of washed pellets of such isolates confirming the association of GFP-MxA condensates associated with Dynabeads. Figure 1C confirms that such isolation was anti-MxA immune specific and did not occur with the irrelevant control pAb against protein disulfide isomerase (PDI). Subsequently, duplicate samples of anti-GFP-MxA condensate isolates derived from OECM`1 cells were subjected to tryptic digestion, the peptides subjected to mass-spectroscopy and peptide families identified in an unbiased manner by comparison to a human peptide data base using propriety methods at Creative Proteomics, Inc. Data summarized in Figure 1D are in terms of arbitrary intensity units of respective peptide families. As a positive control, both duplicate samples revealed MxA/Mx1 peptides. Of the additional >200 peptide families in the dataset, unexpectedly, the BACH1 transcription factor family (comprising BACH1, BACH2, Nrf1, Nrf2, Nrf3) yielded by far the strongest signal, followed by lesser and variable signals for other peptide families (Figure 1D).

3.2. Confirmation of Association of BACH1 and Nrf2 with GFP-MxA Condensates by Immunofluorescence Assays

Immunofluoresence analyses of cells expressing GFP-MxA condensates were then carried out using respective antibodies to the top 6-8 candidate proteins (listed in Figure 1D) to help validate potential identifications [24,25,28,30]. Figure 2A illustrates a negative control without primary antibody, and, Figure 2B (white arrows) illustrates a positive verifying the presence of immunoreactive MxA (in red) in GFP-MxA condensates. Figure 2C shows a negative result for nucleolin, and Figure 2D (white arrows) a positive result for BACH1.

Figure 3A illustrates a negative control in which a rabbit pAb to BMPR1 (bone morphogenetic protein receptor 1) was used as an irrelevant antibody. Figure 3B (white arrows) reconfirms the detection of BACH1 in GFP-condensates. Since, from among the family members of this transcription factors enumerated in Figure 1D), Nrf2 is already known to be antiviral towards influenza A virus (IAV) as well as additional RNA viruses [37,38,40,41], we wondered whether Nrf2 might also colocalize with the GFP-MxA which is also antiviral towards IAV and several RNA and DNA viruses. The data in Figure 3C (white arrows) reveal that Nrf2 transcription factor strongly colocalized with GFP-MxA condensates.

It is noteworthy that in Figure 2 and Figure 3 the same batch of secondary A594-tagged donkey anti-rabbit IgG was used in all the assays (in the negative controls, the positive MxA control, and in assays for the new candidate proteins). Similar immunoassays were negative for lamin, annexin A, nucleophosmin, plectin and histone species using respective pAb and mAb (data not shown). Thus, this immunoassay-based survey confirmed presence of immunoreactive MxA in GFP-MxA condensates (as positive control), and provided data to support presence of BACH1 and Nrf2 immunoreactive peptides in such condensates (Figure 2 and Figure 3). Since BACH1 and Nrf2 transcription factors are well known for their repressor and activator functions respectively in regulating cellular responses to environmental and oxidative stresses, and in cancer progression and metastases; [31,32,33,34,35,36]), additional techniques were used to validate this association.

Figure 4 illustrates one slice of a high-magnification z-stack collected using confocal microscopy confirming the co-association of BACH1 and Nrf2 with GFP-MxA condensates. Curiously, while significant co-association between GFP-MxA and the respective BACH1 and Nrf2 antigenicity is clear (Pearson’s R=0.845 and =0.796 respectively), some variability in the relative overlap of red and green signals is clearly evident. This suggests that not all GFP-MxA condensates contain the same molar amounts of BACH1 or Nrf2. Distinctive variations in Nrf2 distribution were also evident using wide-filed immunofluorescence microscopy (Figure 5A and 5a; and 5B and 5b). The insets on the right highlight respective condensates 1 through 4 in 5A and 5B showing the differences between red and green pixels in each condensate. These data demonstrate critically that the distribution of BACH1 and Nrf2 is not always necessarily the same in all GFP-MxA condensates even in the same cell; in contrast, in Figure 3B,C it is evenly distributed. These data emphasize the biochemical heterogeneity of GFP-MxA condensates from cell to cell, and even among condensates in the same cell. Additionally, while Figure 5A and 5B focus on spheroidal GFP-MxA condensates, data in Figure 5C confirm the co-association of Nrf2 with fibrillar GFP-MxA.

3.3. Alternative Confirmation of the Association of GFP-BACH1 and GFP-Nrf2 with HA-MxA Condensates

The association of BACH1 and Nrf2 with MxA condensates was evaluated using an approach and reagents different from the immunoassays summarized in Section 3.2 above. OECM1 cells were transiently transfected with an expression vector for HA-MxA together with expression vectors for either GFP-BACH1 or GFP-Nrf2. Two days later the cultures were fixed, and the transfected cells imaged for respective GFP (in green) and for MxA using an anti-HA tag pAb. Thus in these experiments the imaging strategy and reagents are completely different from those in Figure 2, Figure 3, Figure 4 and Figure 5. Figure 6 summarizes representative data obtained using this alternative approach. Figure 6A,B confirm the colocalization of GFP-BACH1 with HA-MxA condensates, while Figure 6C confirms the colocalization of GFP-Nrf2 with HA-MxA condensates.

3.4. Biochemical Interaction Studies Using Western Blots

Western blot assays were carried out to verify whether the Protein A Dynabead-immunoisolated protein pellets used in the analyses in Figure 1D did indeed contain full-length endogenous BACH1 and full-length endogenous Nrf2 proteins. Thus, respective aliquots of the protein pellets from GFP-MxA expressing cells isolated using anti-MxA pAb (PrtPellet1 and PrtPellet2), similar to those used to derive the data in Figure 1D, were probed for Nrf2 (Figure 7A, left-most lane) or for BACH1 (Figure 7B, both lanes) by Western blotting. The data in Figure7A and 7B confirm the presence of full-length Nrf2 and BACH1 proteins in these MxA-condensate isolates, validating the discovery of corresponding peptides in such isolates as in Figure 1D.

The strong colocalizations observed between HA-MxA on the one hand and either GFP-BACH1 or GFP-Nrf2 on the other hand in the experiments summarized in Figure 6 (note that Pearson’s R were approximately 0.95 in all “merge” panels), suggested that this co-transfection approach (preparing extracts from cells transiently co-transfected with vectors of respective full-length tagged constructs) might be useful for protein interaction studies. Figure 7A (right two lanes, arrowheads) show evidence that anti-MxA pAb cross-isolated Nrf2 as well as BACH1 from a prefixed extract of GFP-BACH1 and HA-MxA cotransfected cells. Note that the middle lane in Figure 7A shows that endogenous Nrf2 was indeed pulled out using anti-MxA pAb (as was BACH1 also) indicating that MxA, Nrf2 and BACH1 wre present in the same complex. These data consolidate the evidence derived from the immunofluorescence assays shown in Figure 2 through 6.

Side-by-side comparisons of immunopullout assays and Western blots using extracts of co-transfected OECM1 cells suggest that MxA interacts with a subset of cellular BACH1 In Figure 8, Panels A and C the amounts of BACH1 in the respective anti-MxA pullout lanes are lower than in the anti-BACH1 lanes. Curiously, this interaction between MxA and BACH1 did not require prefixation of cells with PFA. In Figure 8, Panel B the extent of BACH1 pullout using anti-MxA pAb was comparable in cells prefixed with PFA or kept in PBS alone. Figure 8, Panel C compares the anti-MxA and anti-BACH1 pullout to controls without added primary pAb or with the irrelevant anti-CHC pAb. Taken together the data in Figure 7 and Figure 8, validate the presence of full-length BACH1 and Nrf2 in association with MxA.

Human MxA is a cytoplasmic protein which generates a cytoplasmic GFP-MxA condensate [48,49,50,51,52,53,54,55]. In contrast its murine ortholog Mx1 is a nuclear protein and we have shown previously that murine GFP-Mx1 generates nuclear structures with properties of condensates [27]. Since endogenous Nrf2 provided strong data showing immunofluorescence colocalization with human GFP-MxA condensates in the cytoplasm (Figure 3C and 9A), we investigated whether transcription factor Nrf2 might also associate with murine Mx1 nuclear condensates. For this evaluation a vector for murine GFP-Mx1 was transfected into human OECM1 cells, and the cells allowed to form nuclear GFP-Mx1 condensates over 2 days. The presence of Nrf2 in such condensates was then evaluated in immunoassays. The data in Figure 9B show the absence of colocalization between murine GFP-Mx1 nuclear condensates and endogenous Nrf2. Thus, the association of Nrf2 with Mx family members is selective for the human GFP-MxA condensate.

3.5. Association of the Redox-Protective Enzyme Heme Oxygenase 1 (HO1) but Not p62/Sequestosome 1 with GFP-MxA Condensates

Both BACH1 and Nrf2 regulate the expression of genes involved in the oxidative stress response - BACH1 is typically a transcriptional repressor while Nrf2 is a transcriptional activator [31,32,33,34]. Heme Oxygenase 1 (HO1), an oxidative-stress protective protein with antiviral activity [2,34,39], is the downstream product of a target gene regulated up (by Nrf2) or down (by BACH1) through a balance between these two transcription factors. Moreover, HO1 itself displays antiviral activity towards RNA viruses and mediates functions of innate immunity [34,38,39]. Thus, we investigated whether HO1 might be associated with GFP-MxA condensates in cells which also display strong Nrf2 (Figure 3C and 9A). The immunoassay data in Figure 10A,B revealed the association of the HO1 with GFP-MxA cytoplasmic condensates with R >0.8.

Nrf2 is known to be sequestered in the cytoplasm in condensates of p62/sequestosome-1 referred to as “p62 bodies” [38,45,46,47]. The protein p62 serves as a selective autophagy receptor which binds ubiquitinated proteins for subsequent degradation, including as part of the antioxidant stress response. Thus, we investigated whether GFP-MxA condensates might represent such p62 bodies. The data in Figure 10C show that GFP-MxA condensates (which we show above in Figure 2, Figure 3, Figure 4, Figure 5, Figure 6, Figure 7, Figure 8, Figure 9 and Figure 10 to include Nrf2, BACH1 and XO1) did not include p62 (R only 0.2). Thus, MxA condensates do not represent p62 bodies.

3.6. Dynamic Tonicity-Driven Regulation of Nrf2 Transcription Factor Condensates in Oral Cancer Cells

In the “exposome” hypothesis [3,4,25], we posit that in the absence of overt causes of oral cancer (such as tobacco, betel-nut chewing or alcohol consumption), the exposure of the floor of the mouth (the “high-risk region”) to environmental stresses of tonicity, temperature and pH (for example hypotonicity due to saliva and to beverages such as water, tea or coffee) sets into motion repeated cycles of condensate disassembly and reassembly. We envision that pro-carcinogenic transcription factors in cells subjected to such cycling over long periods of time (months or years) might on occasion lead to aberrant signaling triggering a cancer initiation event. With the present discovery of the association of the positive transcription factor Nrf2 with cytoplasmic condensates in oral cells, we investigated whether Nrf2-positive structures in such cells might display dynamic disassembly and reassembly. In these experiments (as in Figure 11), disassembly and then reassembly of GFP-MxA into new condensates helped track the process. The data in Figure 11B show the rapid cytoplasmic dispersal of both GFP-MxA and transcription factor Nrf2 in cells exposed to beverage-like hypotonicity (ELB at 50 mOsm) for 5 min. Remarkably, both Nrf2 and GFP-MxA reassembled together into new condensates when cells were subsequently exposed to isotonic medium. Taken together, the data in Figure 11 provide a foundation for the possibility that assorted pro-carcinogenic transcription factors can indeed cycle in and out of cellular condensates in oral cells subjected to environmental stresses. The question whether repeated recycling of pro-carcinogenic transcription factors out of and then back into condensates in this manner might lead to aberrant signaling and trigger carcinogenesis is intriguing.

4. Discussion

An unbiased proteomics-based evaluation of peptides associated with cytoplasmic GFP MxA condensates isolated from human oral cancer cells revealed the unexpected presence of peptides of the BTB and b-Zip transcription factor family with the identification of BACH1 protein family in the condensate immunoisolates. Two members of this protein family – the transcriptional repressor BACH1 and the transcriptional activator Nrf2 have been investigated extensively in the past 2-3 decades for their roles in cellular responses to oxidative and environmental stresses, antiviral defense and innate immunity, and cancer progression and metastasis [31,32,33,34,35,36]. Thus, with this initial peptide-family identification, we focused on validating the possible association of full length BACH1 and Nrf2 proteins with GFP-MxA condensates. Respective immunofluorescence assays of oral cancer cells expressing GFP-MxA condensates, Western blots including cross-immunopullout experiments, and studies with full-length GFP-BACH1 and GFP-Nrf2 coexpressed in cells together with HA-MxA validated the co-association of these proteins with the same condensates. Moreover, unexpectedly, HO1, the downstream product of the activation of Nrf2 was also included in GFP-MxA condensates. Thus, the inclusion of human MxA, BACH1, Nrf2 and HO1 defines a novel class of “antioxidant” cytoplasmic condensates. These structures were distinct from “p62 bodies” which comprise of the p62/sequestome-1 protein which binds Ub-Nrf2 targeting Nrf2 towards autophagic degradation [38,45,46,47]. Moreover, in a further distinction, murine Mx1, which forms nuclear condensates, did not associate with Nrf2.

Nevertheless, the foundational antiviral activity which led to the recognition starting in 1963 of nuclear murine Mx1 and in the 1970s of its ortholog cytoplasmic human MxA was the inhibition of replication of influenza A virus (IAV)[48,49,50,51,52,53,54]. This antiviral repertoire of Mx family members has since been expanded to include an extensive collection of RNA- and DNA-containing viruses with some distinctions (for example, the cytoplasmic MxA shows antiviral activity against the rhabdovirus VSV, but the nuclear Mx1 does not}[27,48,49,50,51,52,53,54]. In cell-free assays, MxA inhibits early RNA viral transcription through a mechanism not yet completely elucidated [52,53,54]. Amazingly, Nrf2 also inhibits IAV replication at an early stage of infection [37,38,40,41]. Moreover, Nrf2 also displays antiviral activity towards additional RNA and DNA viruses such as respiratory syncytial virus (RSV), SARS-Cov-2, herpes simplex 1 and vaccinia virus [37,38,40,41]. Even HO1 displays antiviral activity towards RNA viruses [38,39]. Thus, taken together, the literature points to the possibility of multiple antiviral mechanisms mediated jointly by the components of the MxA, Nrf2, and HO1-containing “antioxidant condensate”. In the case of MxA, it is the dispersed phase that appears to possess antiviral activity, with the condensate representing a storage depot [25,55].

The balance between the cytoplasmic and nuclear pools of Nrf2 and BACH1 regulates their respective transcriptional activities [31,32,33,34,35,36,56]. Increased cellular BACH1 is associated with increased metastasis of triple-negative breast cancer, and increased Nrf2 with hepatocellular carcinoma [57,58]. p62 condensates incorporate Ub-Nrf2 and target it towards degradation – mutations in p62 and associated proteins represent gain in function changes which enhance Nrf2 function in carcinogenesis [45,46,47].

While human MxA is known to be expressed in healthy gingival/buccal epithelium, likely through induction by IFN-λ species or α-defensins [29], in the present study we used GFP-MxA as a reporter for studies of condensate dynamics in oral cancer cells. In previous experiments, we had observed that cytoplasmic GFP-MxA condensates in oral epithelial cells were rapidly (within 1-3 min) dispersed by saliva- and beverage-like hypotonicity (range 30–100 mOsm). This then triggered an active cellular process which mediated a “spontaneous” reassembly of MxA into a new set of condensates in the next 5–10 min [25]. This process reflected water influx and “uncrowding” of the cytoplasm in the disassembly phase, and water efflux and “recrowding” of the cytoplasm in the reassembly process [25,28,30]. This process was temperature sensitive (faster at 50°C and slower at 5°C; [25]) and involved the WNK-SPAK/OSR1 kinase pathway which regulates chloride and water influx and efflux [25]. The discovery that the transcription factors BACH1 and Nrf2 associate with MxA condensates suggested that these transcription might also be regulated by water and chloride influx and efflux in oral epithelial cells. Indeed, the present data (Figure 11) show that Nrf2 disassembles in oral cancer cells in response to hypotonicity and reassembles into new condensates when cells are exposed to isotonic conditions. These data provide a basis for suggesting that hypotonicity, temperature and environmental stresses preferentially inflicted on cells in the floor of the mouth along the path of liquid transit might particularly disrupt the normal function of Nrf2 and BACH1 and HO1 in this location in the mouth. We posit this mechanism as a possible basis of this region in the mouth for representing a “high-risk zone” for cancer development.

5. Conclusions

All of us repeatedly challenge our oral mucosa with the stresses of tonicity and temperature every single day. We posit that these stresses trigger repetitive cycles of disassembly/reassembly of biomolecular condensates in oral epithelial cells in the U-shaped high cancer risk region in the mouth along the path of liquid transit (the “exposome”)[3,4]. The association of transcription factors BACH1, Nrf2 and the protective protein HO1 with MxA condensates in such cells defines a novel “antioxidant condensate” in such cells. The data suggest a novel subcellular mechanism for the initiation of oral cancer in the “high-risk” region – the repetitive condensate disassembly and reassembly cycles of these prooncogenic transcription factors resulting in aberrant episodic cancer initiation events.