Latency-Associated Peptide Rapidly Upregulates Neuraminidase 3 in a Profibrotic Translation-Based Positive Feedback Loop

TL;DR

TGF-β1 rapidly increases NEU3 levels through translation, creating a feedback loop that worsens fibrosis, and this can be blocked by inhibiting DDX3.

Contribution

Discovery of a rapid, translation-based positive feedback loop involving TGF-β1, LAP, and NEU3 in fibrosis.

Findings

TGF-β1 increases NEU3 levels via translation, not transcription.

LAP synergizes with TGF-β1 to upregulate NEU3, forming a feedback loop.

DDX3 inhibition with RK-33 blocks NEU3 upregulation by TGF-β1 and LAP.

Abstract

What are the main findings?
TGF-β1, a key driver of fibrosis, rapidly increases levels of the profibrotic sialidase NEU3 via translational upregulation, independent of new transcription.After activation by NEU3, the TGF-β1 sequestering protein LAP also upregulates NEU3, and works synergistically with TGF-β1, contributing to a TGF-β1 → NEU3 → TGF-β1 positive feedback loop. LAP upregulation of NEU3 is blocked by inhibiting the RNA binding protein DDX3 with RK-33.

TGF-β1, a key driver of fibrosis, rapidly increases levels of the profibrotic sialidase NEU3 via translational upregulation, independent of new transcription.

After activation by NEU3, the TGF-β1 sequestering protein LAP also upregulates NEU3, and works synergistically with TGF-β1, contributing to a TGF-β1 → NEU3 → TGF-β1 positive feedback loop. LAP upregulation of NEU3 is blocked by inhibiting the RNA binding protein DDX3 with RK-33.

What is the implication of the main finding?
LAP is a potent amplifier of profibrotic signaling.DDX3 inhibition is a potential strategy to inhibit the LAP → NEU3 → TGF-β1 pathway.

LAP is a potent amplifier of profibrotic signaling.

DDX3 inhibition is a potential strategy to inhibit the LAP → NEU3 → TGF-β1 pathway.

Fibrosis appears to be an out-of-control wound-healing response that drives a progressive formation of scar tissue in an organ. A key profibrotic cytokine, transforming growth factor beta-1 (TGF-β1), upregulates levels of the extracellular sialidase neuraminidase 3 (NEU3), and NEU3 in turn can activate latent TGF-β1 to release active TGF-β1 from the sequestering latency-associated peptide (LAP). In the mouse bleomycin model of pulmonary fibrosis, NEU3 is both necessary and sufficient for pulmonary fibrosis. In this report, we find that NEU3 protein levels increase both intracellularly and extracellularly in cultures of human lung fibroblasts within 5 min of TGF-β1 exposure. This effect is driven by an increase in translation and is independent of new transcription, supporting a model where TGF-β1 causes a pool of weakly translated NEU3 mRNA to increase translation. By participating in the feedback loop, latent TGF-β1 makes cells more sensitive to TGF-β1. LAP also stimulates NEU3 expression and acts synergistically with TGF-β1 to upregulate NEU3. The positive feedback loop is blocked by NEU3 inhibitors. The RNA helicase DEAD-box helicase 3 (DDX3) mediates NEU3 translation, and the DDX3 inhibitor RK-33 blocks the rapid upregulation of NEU3 by TGF-β1 and LAP. Exposure of cells to TGF-β1 but not LAP induces dephosphorylation of DDX3 within two minutes, suggesting that the mechanisms used by TGF-β1 and LAP to activate DDX3 to increase NEU3 levels may differ. Together, these results suggest that a rapid positive feedback loop involving TGF-β1, LAP, and NEU3 helps drive fibrosis.

Figures

Full text

1. Introduction

Fibrotic disorders, characterized by a progressive buildup of scar tissue in an internal organ, contribute to nearly 45% of deaths in the United States [1,2]. Idiopathic pulmonary fibrosis (IPF), a lung-specific fibrotic disorder, has a median survival of only 4.5 years after diagnosis [3]. While current treatments can slow disease progression, they do not reverse or stop fibrosis, and the prognosis remains poor [4]. A key driver of fibrosis is the extracellular cytokine transforming growth factor beta 1 (TGF-β1) [5,6]. TGF-β1 is synthesized as an inactive pre-pro-protein with an N-terminal prodomain called latency-associated peptide (LAP) [7,8]. After cleavage of the signal peptide, TGF-β1 is trafficked to the endoplasmic reticulum where two TGF-β1 pro-proteins form a disulfide-linked homodimer [9]. After trafficking to the Golgi, proteases cleave the LAP prodomain, LAP and TGF-β1 are secreted, and LAP remains noncovalently associated with TGF-β1, forming the latent TGF-β1 complex [10,11,12]. The latent TGF-β1 complex binds to latent TGF-β-binding protein (LTBP) on the extracellular matrix where LAP sequesters TGF-β1 until activation [9,10]. Multiple mechanisms can release active TGF-β1 from LAP [13,14,15,16,17]. Like many other secreted proteins, LAP is glycosylated [18,19,20], and many glycosylated structures have sialic acids at the distal end of the polysaccharide chain which help regulate various cellular mechanisms [21,22,23,24]. Sialic acid has a negative charge due to its carboxyl group, contributing to the stabilization of the latent TGF-β1 complex [25,26]. Sialidases, also known as neuraminidases, remove the terminal sialic acid from glycoconjugates, and desialylation of LAP loosens the non-covalent association with TGF-β1, releasing active TGF-β1 from sequestration [25,26,27,28].

There are four known mammalian sialidases, neuraminidase 1 through 4 (NEU1-NEU4), each with distinct substrate specificities and subcellular localizations [27,29]. NEU3 is found intracellularly in endosomes, on the extracellular side of the plasma membrane, and may also be released into the extracellular environment [29,30,31,32,33]. NEU3 has been implicated in a range of pathological processes across organ systems including intestinal inflammation and colitis [34], liver fibrosis [35], and atherosclerosis [36]. NEU3 knockout mice do not develop bleomycin-induced pulmonary fibrosis [37], and NEU3 aspiration is sufficient to induce pulmonary fibrosis in mice [31], indicating that NEU3 is both necessary and sufficient for pulmonary fibrosis in mice. Inhibition of NEU3 with general sialidase inhibitors such as 2,3-didehydro-2-deoxy-N-acetyl-neuraminic acid (DANA) and oseltamivir, and more specific NEU3 inhibitors such as 2-acetyl pyridine (2-AP) and 4-amino-1-methyl-2-piperidinecarboxylic acid (AMPCA), inhibits pulmonary inflammation, decreases active TGF-β1 levels in the lungs, and decreases pulmonary fibrosis in a mouse bleomycin model [28,31,32,37]. NEU3 contributes to fibrosis through three key mechanisms: (1) NEU3 desialylates the endogenous serum antifibrotic protein serum amyloid P (SAP), which inhibits SAP activity [38,39], (2) NEU3 upregulates the pro-inflammatory cytokine IL-6 in human peripheral blood mononuclear cells, which in turn upregulates NEU3, functioning in a positive feedback loop [28,37,39], and (3) NEU3 desialylates LAP, releasing TGF-β1 from latent TGF-β1, and the released active TGF-β1 upregulates NEU3 translation and decreases NEU3 degradation, functioning in a second positive feedback loop [32,40].

For about two-thirds of the proteins responsive to TGF-β1, regulation occurs at the transcriptional level through the classic multi-step process of (1) influencing a transcription factor, (2) nuclear translocation, (3) gene transcription, (4) mRNA processing, (5) mRNA export from the nucleus, and (6) translation [41,42]. This multi-step process imposes a temporal delay such that increases in protein levels typically occur after several hours [43,44,45,46]. In contrast, the remaining one-third of proteins are controlled at the level of translation where the protein levels may be changed without having to complete steps 1–5, potentially allowing a relatively quicker response [47,48].

In human lung fibroblasts, there are at least 182 proteins, including NEU3, whose levels are increased by TGF-β1 without a corresponding increase in their mRNA levels, but with a shift in these mRNAs from free and monosome fractions to polysomes, indicating translational regulation [49]. Of the 182 proteins, 180 (including NEU3) share a 20-nucleotide motif. This motif is necessary and sufficient for TGF-β1-induced translation of NEU3 [49]. DEAD-box helicase 3 (DDX3) is an ATP-dependent RNA helicase that is upregulated in the lungs of mice with bleomycin-induced pulmonary fibrosis and in fibrotic lesions of IPF patients [49]. In response to TGF-β1, DDX3 increases its binding to the 20-nucleotide common motif [49]. DDX3 inhibition with the DDX3 inhibitor RK-33 [50] reduces TGF-β1 upregulation of NEU3 in human lung fibroblasts, and in mice, RK-33 injections reduce bleomycin-induced lung inflammation and fibrosis and lung tissue levels of DDX-3, TGF-β1, and NEU3 [49]. Notably, although DDX3 is involved in other cellular processes, RK-33 has no reported toxicity in various animal models at the therapeutic dose, and RK-33 is currently being developed as a cancer therapeutic [50,51,52]. This suggests that DDX3 may play a key role in the TGF-β1-mediated regulation of NEU3.

In this report, we show that NEU3 is rapidly upregulated by TGF-β1 via a translation-dependent mechanism, with intracellular and extracellular levels rising within 5 min. We further show that the NEU3 → TGF-β1 → NEU3 positive feedback loop is activated and amplified within 5 min, and the latent TGF-β1 complex enhances NEU3 upregulation. Notably, LAP itself rapidly upregulates NEU3 and acts synergistically with TGF-β1, revealing an unrecognized signaling role for LAP. This loop is blocked by the NEU3 inhibitors 2-AP and AMPCA, and the DDX3 inhibitor RK-33 blocks the rapid upregulation of NEU3 by TGF-β1 and LAP. Exposure of cells to TGF-β1 induces rapid dephosphorylation of DDX3 within two minutes, whereas exposure of cells to LAP has no significant effect on DDX3 phosphorylation levels. Together, these findings suggest that along with TGF-β1, LAP plays an important biological role in NEU3 upregulation, which is mediated by DDX3 generating a very fast wound-healing response that, when dysregulated in an active state, could contribute to fibrosis.

2. Materials and Methods

2.1. Primary Human Lung Fibroblast Cell Culture

Eight human lung fibroblast (HLF) cell lines from 3 healthy males (M2, M3, NL-83), 1 healthy female (NL-44), 3 female IPF patients (IPF-36, IPF-8, F5), and 1 male IPF patient (IPF-32), were gifts from Dr. Carol Feghali-Bostwick, Medical University of South Carolina, Charleston, SC, USA. Frozen cell lines were thawed and cultured in 6-well tissue culture dishes (10062-892, VWR, Radnor, PA, USA) in DMEM (15-017-CV, Corning, Glendale, AZ, USA) supplemented with 10% bovine calf serum (BCS) (SH30072.03, Cytiva, Wilmington, DE, USA), 100 U/mL penicillin and 100 μg/mL streptomycin (SV30010 Cytiva, USA), and 2 mM glutamine (SH30034.01, Cytiva, USA) (referred herein as “culture medium”), for 2–4 days at 37 °C in a humidified 5% CO_2_ incubator, replacing the culture medium every 2 days. After reaching ~60–80% confluence, the medium was removed, cells were rinsed twice with 1 mL of 1× phosphate-buffered saline (PBS) (21-040-CV, Corning) before detaching with 0.5 mL of Accutase (AT-104, Innovative Cell Technologies, San Diego, CA, USA). After detaching, Accutase was neutralized with 1 mL of culture medium, 14 mL of PBS was added, and cells were collected by centrifugation at 200× g for 5 min. The cells were resuspended in culture medium to 1.5 × 10^4^ cells/mL, and added to 24-well plates (353047, Corning) at 500 µL/well or 96-well black μ-plates (89626, Ibidi Fitchburg, WI, USA) at 200 µL/well. For all experiments, the healthiest growing cell lines were used at the lowest available passage number (all cells were used at passages 4–10, and no abnormal morphology differences were observed in this passage range) while making a concerted effort to include both male and female cells from healthy and IPF donors. “Protein free medium” (DMEM/100 U/mL penicillin/100 μg/mL streptomycin/2 mM glutamine) was used for most experimental dilutions and as a negative control. Culture medium and protein free medium were warmed to 37 °C before use.

2.2. Coomassie and Silver Staining of Gels and Western Blot

HLF lines in 24-well plates were grown in culture medium for 2–4 days at 37 °C in a humidified 5% CO_2_ incubator. After reaching ~60–80% confluence, the liquid was removed, cells were washed twice with PBS, and 200 μL of protein free medium with or without 10 ng/mL of recombinant human active TGF-β1 (781802, BioLegend, San Diego, CA, USA, diluted from the vendor 0.2 mg/mL stock in protein free medium) was added to the wells. At the indicated times, the liquid was removed, cells were washed once with PBS, and cells were lysed in situ with 150 μL of Laemmli sample buffer with 10% 2-mercaptoethanol (6050 EMD Millipore Corporation, Norwood, OH, USA) and left for 5 min. The liquid was pipetted up and down twice to collect all the cells and transferred to Eppendorf tubes. Samples were heated at 95 °C for 5 min and 5 μL were loaded on two 4–20% Tris-glycine Mini-Protean TGX gels (Bio-Rad, Hercules, CA, USA), one for Coomassie staining and one for a Western blot. Where indicated, 6× native gel running buffer was added to samples. The gels ran at 90 volts for 1.5 h in 1× Tris/glycine/SDS (25 mM, 192 mM, 0.1% respectively) running buffer. For total protein evaluation of samples, one gel was stained with Coomassie staining solution (0.1% w/v Coomassie brilliant blue (0615-10G, Bio-Rad)/50% methanol/10% glacial acetic acid/40% water) for 1 h with gentle shaking, and destained with destaining solution (10% glacial acetic acid, 40% ethanol, 50% water) overnight with gentle shaking (after the first hour, the gel was rinsed with water and the destaining solution was replaced). For evaluation of total phosphorylated proteins, gels were silver-stained following [53]. Coomassie and silver-stained gels were imaged using a ChemiDoc XRS+ System (Bio-Rad) and the integrated density of lanes was measured using Image Lab software (Version 6.1, Bio-Rad).

For extracellular NEU3, following TGF-β1 exposure for 5 min, culture supernatants in 24-well plates were gently pipetted up and down, collected in Eppendorf tubes, and placed on ice. 50 μL of PBS was added to each well to gently rinse cells and was collected and added to the supernatants. The conditioned medium was clarified by centrifugation at 200× g for 10 min, the top 80% of the supernatant was collected and clarified at 10,000× g for 10 min, and 20 μL was taken from the top of each tube, mixed with 4 μL of 6× Laemmli buffer, and processed as above.

For Western blots, proteins were transferred to polyvinylidene difluoride membranes (88518, Thermo Fisher Scientific) in Tris/glycine/SDS buffer containing 20% methanol [32]. Membranes were blocked for 1 h at room temperature with blocking buffer (2% IgG-free BSA (BSA-50, Rockland Immunochemicals, Gilbertsville, PA, USA) in Tris-buffered saline with Tween 20 (TBST; 150 mM NaCl, 10 mM Tris-HCl pH 7.4 with 0.1% Tween 20 (0777-1L, VWR)). Membranes were incubated with 1 μg/mL rabbit anti-NEU3 antibody (27879-1-AP, ProteinTech, Rosemont, IL, USA), or 1 μg/mL rabbit anti-caldesmon-1 antibody (D5C80, Cell Signaling Technology, Danvers, MA, USA), or 1 μg/mL rabbit anti-DDX3 antibody (NBP2-67-121, Bio-Techne, Minneapolis, MN, USA) diluted in blocking buffer overnight at 4 °C on a platform rocker. The next day, blots were washed 3 times in TBST for 10 min each at room temperature and incubated with peroxidase-conjugated donkey F(ab′)2 anti-rabbit secondary antibody (711-036-152, Jackson ImmunoResearch, West Grove, PA, USA) at 400 ng/mL in blocking buffer for 1 h at room temperature. Blots were then washed 3 times in TBST for 10 min each at room temperature and SuperSignal West Pico Chemiluminescence Substrate (34577, Thermo Fisher Scientific, Waltham, MA, USA) was used following the manufacturer’s protocol to visualize the peroxidase using a ChemiDoc XRS+ System. The integrated density of bands was measured using FIJI (Version 1.54p) [54].

2.3. Immunofluorescence and Image Analysis

HLF cell lines were added to 96-well black μ-plates at 2000 cells per well and grown in culture medium for 24–48 h at 37 °C in a humidified 5% CO_2_ incubator. After reaching ~50–60% confluence (this lower confluence was used for these experiments to avoid cell crowding to allow an accurate measure of fluorescence for each cell), the medium was removed, and cells were washed twice with PBS and treated with 200 μL of protein free medium with or without 10 ng/mL recombinant human active TGF-β1 at 37 °C in a humidified 5% CO_2_ incubator. Other TGF-β1 concentrations or treatments were also used as indicated. Where indicated, recombinant human latent TGF-β1 (299-LT/CF, R&D systems), recombinant human latency associated peptide (LAP-H5213, Acro Biosystems, Newark, DE, USA), and lipopolysaccharide (437627, Caliobiochem Corporation, La Jolla, CA, USA) were diluted in protein free medium and added to cells. Where indicated, a 50 kDa cutoff spin filter (UFC505024, MilliporeSigma, Burlington, MA, USA) was pre-wet with 500 μL of culture medium, spun at 12,000× g for 5 min, and the flow through and retentate were discarded. 500 μL of 100 ng/mL latent TGF-β1 was added to the spin filter and spun at 12,000× g for 30 min. The retentate was reconstituted to 500 μL with protein free medium and diluted to the indicated concentrations. The NEU3 inhibitors 2-acetylpyridine (2-AP) (W325104-1KG-K, Sigma-Aldrich, St. Louis, MI, USA) [28,32] and AMPCA (manufactured by Sundia MediTech Company Ltd., Shanghai, China) [28,32] were made as 10 mM stocks in water, diluted in protein free medium, and added to cells simultaneously with TGF-β1. For staining, wells were washed with 200 μL of PBS and fixed with 100 μL of 4% paraformaldehyde (19210, Electron Microscopy Sciences, Hatfield, PA, USA) in PBS for 10 min at room temperature. After fixing, the liquid was removed, cells were washed with 200 μL of PBS, and permeabilized with 100 μL of 0.5% Triton X-100 (J66624, Thermo Fisher) in PBS for 5 min. The liquid was removed, cells were washed with 200 μL of PBS, and then blocked with 200 μL of 1% IgG-free BSA in PBS (PBSB) for 1 h at room temperature. After blocking, liquid was removed, and 100 μL of either rabbit anti-NEU3 antibody (OACA04338, Aviva Systems, San Diego, CA, USA) or rabbit anti-caldesmon antibody (D5C80, Cell Signaling Technology) was added at 1:1000 in PBSB and left overnight at 4 °C. The next day, cells were washed 3 times for 5 min per wash with 0.05% Tween 20 in PBS (PBST) at room temperature. After washing, 100 μL of 1 μg/mL donkey F(ab′)2 anti-rabbit Alexa Fluor 488 secondary antibody (711-546-152, Jackson ImmunoResearch) in PBSB was clarified by centrifugation at 12,000× g for 5 min and added for 30 min at room temperature. Cells were then washed 3 times for 5 min per wash with PBST. After washing, 50 μL of 2 μg/mL DAPI (422801, Biolegend) in PBS was added to each well. Images were taken with an Eclipse Ti2 microscope (Nikon, Melville, NY, USA) and analyzed with FIJI software. For each experiment, a control well stained with only secondary antibody was imaged to evaluate background fluorescence.

For each well, 2–5 images were taken (based on cell density), and all the cells were selected with the freehand line tool on FIJI. For images with high cell density, all cells in 2–3 randomly selected regions were outlined. The area and integrated density were measured. Integrated densities were then divided by the corresponding area and the intensity/area of the cells stained with secondary antibody alone was subtracted. These readings were then averaged. The average was then normalized to the experimental negative control (cells with no treatment, stained with anti-NEU3 or anti-caldesmon antibodies) as percent change, with the negative control as 100%. For several experiments, image acquisition and quantification was performed by different individuals, with at least one of them blinded to the treatment conditions. Independent analyses produced consistent results.

2.4. Transcription and Translation Inhibition and DDX3 Inhibition with RK-33

Cycloheximide (94271, VWR), actinomycin D (BVT-0089-M005, Adipogen, San Diego, CA, USA), or RK-33 (S8246, Selleck Chemicals, Houston, TX, USA) were dissolved in DMSO to 10 mg/mL, 0.5 mg/mL, and 10 mM stocks respectively, and further diluted in protein free medium. HLF cell lines were seeded on 96-well black μ-Plates as above. After reaching ~50–60% confluence, liquid was removed, cells were washed twice with PBS and pre-treated with 100 μL of either 50 μg/mL cycloheximide, 1 μg/mL actinomycin D, 10 μM RK-33, or an equal volume dilution of DMSO in protein free medium (control) at 37 °C in a humidified 5% CO_2_ incubator. For actinomycin D and cycloheximide, after 10 min, liquid was removed, cells were washed with 200 μL of 1× PBS and treated with 200 μL of protein free medium with or without active TGF-β1 for the indicated times. For RK-33, protein free medium with or without 10 ng/mL of active TGF-β1 or 0.3 ng/mL recombinant human LAP was added directly to the wells and left for 5 min. After the indicated times, wells were then washed with 1× PBS and stained by immunofluorescence as above.

2.5. Phosphorylation of DDX3

HLF lines in 6-well plates were grown in culture medium for 4–5 days at 37 °C in a humidified 5% CO_2_ incubator. After reaching ~70–80% confluence, the liquid was removed, cells were washed twice with PBS, and 1 mL of protein free medium with or without 10 ng/mL of recombinant human active TGF-β1 or 0.3 ng/mL of recombinant human LAP was added to cells. At the indicated times, cells were washed once with PBS, detached with 0.5 mL of Accutase for 5 min at room temperature and collected in Eppendorf tubes. 1 mL of PBS was added to dilute the Accutase, and cells were collected by centrifugation at 200× g for 5 min at room temperature. Supernatants were removed, and cells were resuspended and washed in 1.5 mL of PBS and collected by centrifugation at 200× g for 5 min at room temperature. The washes were repeated for a total of three washes with the last centrifugation at 500× g. Supernatants were removed and phosphorylated proteins were isolated using TALON PMAC Phosphoprotein Enrichment Kit (635641, Takara Bio, San Jose, CA, USA) following the manufacturer’s protocol. Phosphorylated proteins and supernatants of total cell lysates were stained for DDX3 using Western blots as described above.

2.6. Evaluation of TGF-β1 Upregulation of GAPDH

To determine whether TGF-β1 has an effect on GAPDH, monosome and polysome fractionation Supplemental data set 3 from Chen et al. [49] was examined. GAPDH was identified in the raw ribosomal data using its Ensemble ID ENSG00000111640. The abundance of GAPDH mRNA in the monosome and polysome fractions was evaluated for the four primary human lung fibroblast cell lines treated with and without 10 ng/mL of TGF-β1.

2.7. Identification of Potential G-Quadruplex Regions in NEU3 mRNA

To identify possible G-quadruplex regions in the NEU3 mRNA, the NM_006656.6 NEU3 transcript from the National Center for Biotechnology Information (NCBI) database was analyzed using QGRS Mapper [55].

2.8. Statistical Analysis

Statistical analyses were performed using Prism 10 (GraphPad Software, La Jolla, CA, USA). Statistical significance between two groups was determined by unpaired t-test, or between multiple groups using ANOVA with Dunnett’s post hoc test.

3. Results

3.1. TGF-β1 Upregulates NEU3 Levels Within 2 Min

We previously observed that a 48 h exposure to 10 ng/mL of TGF-β1 in human lung fibroblasts causes a group of mRNAs (including NEU3) to shift from monosomes to polysomes and increase corresponding protein levels without an increase in mRNA levels, indicating translational regulation [49]. In contrast, in response to TGF-β1 another group of proteins (including caldesmon) have stable monosome-to-polysome ratios, and increased mRNA levels with a corresponding increase in protein level, indicating transcriptional regulation [49]. Translation regulation can result in a very rapid protein increase in response to a signal [56,57,58]. To determine how rapidly NEU3 is upregulated, we examined NEU3 expression as a function of time after exposure to TGF-β1 and compared this to caldesmon expression. Male and female human lung fibroblasts were treated with 10 ng/mL of TGF-β1, and NEU3 and caldesmon levels were evaluated using Western blots normalized to Coomassie-stained gels of samples (Figure 1A–D and Figure S1A) and immunofluorescence (Figure 1E–G and Figure S2). Western blots were normalized to Coomassie-stained gels as TGF-β1 upregulates commonly used loading controls such as β-actin [59,60] and GAPDH [49,61]. Additionally, re-analysis of Supplemental data set 3 from Chen et al. [49] revealed that human lung fibroblasts exposed to TGF-β1 for 48 h had increased GAPDH translation (shift from monosomes to polysomes) (Supplementary Figure S1B). As previously observed, due to two isoforms of NEU3, the anti-NEU3 antibodies detected two bands on Western blots [28,30,33,40] (Figure 1A). NEU3 protein levels increased 2 min after TGF-β1 exposure on both Western blots and immunofluorescence, whereas the earliest rise in caldesmon was seen after 4 h (Figure 1B,D,G), similar to other TGF-β1 transcriptionally regulated proteins [43,44,45,46]. On the Western blots, TGF-β1 caused NEU3 levels to increase to 183 ± 23% of baseline (100%) at 2 min (Figure 1C). Analysis of the data in Supplemental data set 5 from Chen et al. [49] showed that after 48 h of TGF-β1 exposure, the median increase of all proteins whose levels were increased by TGF-β1 was to 150% of the baseline level. The TGF-β1-induced increase in NEU3 at 2 min was thus higher than the median TGF-β1-induced protein increase at 48 h. We observed that baseline levels of NEU3 decrease over time in protein free medium but not in serum-containing culture medium (Supplementary Figure S3A,C,E,F), whereas caldesmon levels showed less of a decrease (Supplementary Figure S3B,D,G). This indicates that some serum factor maintains baseline NEU3 levels. At 24 h in protein free medium, baseline levels of NEU3 decreased, but relative levels of NEU3 with TGF-β1 were increased compared to protein free medium control as reflected in the quantification of immunofluorescence and Western blots (Figure 1C,D,F,G). These data indicate that, compared to TGF-β1 upregulation of caldesmon, TGF-β1 can rapidly increase NEU3 levels.

3.2. A 5-Min TGF-β1 Exposure Upregulates Extracellular NEU3

NEU3 is found intracellularly in endosomes, on the extracellular side of the plasma membrane, and under some conditions, it is also released to the extracellular environment [31,32,33]. To determine if a brief TGF-β1 exposure increases extracellular NEU3, human lung fibroblasts were exposed to TGF-β1, and Western blots of the conditioned media were stained for NEU3. Only one band of NEU3 was observed in the cell culture supernatants (Figure 1H), which is consistent with previous observations [31,33,40]. Extracellular NEU3 protein levels increased at 5 min with 30 and 100 ng/mL of TGF-β1 (Figure 1H,I). These data suggest that, in addition to rapidly increasing levels of intracellular NEU3, TGF-β1 also rapidly increases levels of extracellular NEU3 at higher concentrations.

3.3. The Rapid TGF-β1-Mediated Upregulation of NEU3 Occurs in the Presence of a Transcription Inhibitor but Not in the Presence of a Translation Inhibitor

TGF-β1 can upregulate NEU3 protein without a corresponding increase in NEU3 mRNA levels [49], but it was unknown whether this upregulation required transcriptional regulation of another upstream factor or signaling molecule. To test the hypothesis that rapid NEU3 upregulation is independent of transcriptional control, we examined the NEU3 response to TGF-β1 in the presence of the transcription inhibitor actinomycin D (Act D) and the translation inhibitor cycloheximide (CHX) [54,62]. Fibroblasts were pre-treated with either Act D or CHX, or an equal volume of DMSO buffer control, and were then stimulated with 10 ng/mL of TGF-β1. When pre-treated with Act D, levels of the apparently transcriptionally regulated protein caldesmon did not significantly increase in response to TGF-β1 (Figure 2A,B, Supplementary Figures S4A and S5A,B). In contrast, TGF-β1 exposure rapidly upregulated NEU3 both in the Act D pre-treated and control conditions (Figure 2A,C, Supplementary Figures S4B and S5C,D). In cells treated with CHX [63], TGF-β1 did not significantly upregulate NEU3 (Figure 2D,E, Supplementary Figures S4C and S5E,F). In the DMSO buffer control conditions, caldesmon levels increased sooner (20 min) than the previously observed increase at 4 h (Figure 1G). This is likely due to the presence of DMSO, which increases transcription through multiple mechanisms including weakening histone–histone interactions and relieving chromatin-mediated repression [64] as well as altering RNA polymerase conformation, enhancing its initiation activity [65,66]. Notably, NEU3 control levels did not change with DMSO, supporting the idea that TGF-β1 upregulation of NEU3 is not mediated by a transcription mechanism. Overall, these data support the idea that TGF-β1 upregulation of NEU3 is mediated by translation and does not require TGF-β1 regulation of transcription.

3.4. NEU3 Functions in a Positive Feedback Loop with TGF-β1

In the latent TGF-β1 complex, LAP is sialylated [18,19,20], and NEU3 desialylates LAP [28]. This desialylation causes the release of active TGF-β1 from LAP sequestration [28]. Based on these observations, we previously indirectly showed that NEU3 functions in a positive feedback loop with TGF-β1 where NEU3 releases active TGF-β1 from the latent TGF-β1 complex, and the released TGF-β1 increases levels of NEU3 [28,39,40]. A prediction of this positive feedback amplification model is that for cells exposed to a low concentration of active TGF-β1, which causes a slight upregulation of NEU3, addition of latent TGF-β1 would allow the slightly increased levels of extracellular NEU3 to activate the latent TGF-β1, releasing active TGF-β1, causing further upregulation of NEU3. To directly study this feedback loop, we first examined the effect of active TGF-β1 and observed that at 1 ng/mL and below, although there was a slight upward trend, active TGF-β1 did not significantly increase NEU3 at 5 min as assessed by immunofluorescence (Figure 3A). Similarly, latent TGF-β1 alone (gray line in Figure 3B) upregulated NEU3 at 3, 10, and 30 ng/mL, but 1 ng/mL and lower concentrations of latent TGF-β1 did not significantly increase NEU3 (Figure 3B). Next, 0.1, 0.3, or 1 ng/mL of active TGF-β1 (which did not significantly upregulate NEU3) were mixed with a range of concentrations of latent TGF-β1 and added to fibroblasts for 5 min (Figure 3B). In the presence of latent TGF-β1, all three of these active TGF-β1 concentrations upregulated NEU3, with higher concentrations of active TGF-β1 needing less latent TGF-β1 for NEU3 upregulation, supporting the positive feedback model (Figure 3B). The commercial latent TGF-β1 preparation has a small amount of active TGF-β1 already present (Supplementary Figure S6A and [67]). To test if the ability of latent TGF-β1 to upregulate NEU3 is due to this active TGF-β1 contaminant (~22 kDa), the commercial latent TGF-β1 (~75 kDa) was filtered using a 50 kDa cutoff spin filter and the retentate was added to human lung fibroblasts. Cells were treated with either filtered latent TGF-β1 alone or filtered latent TGF-β1 with 0.3 ng/mL of active TGF-β1. The filtered latent TGF-β1 upregulated NEU3 at 100 ng/mL (Figure 3C) but not at 3, 10, or 30 ng/mL as seen with the unfiltered commercial preparation (Figure 3B), suggesting that the NEU3 upregulation at lower concentrations of unfiltered latent TGF-β1 was due to the active TGF-β1 contaminant. Addition of 0.3 ng/mL of active TGF-β1 to 0.1 or 0.3 ng/mL of filtered latent TGF-β1 upregulated NEU3 (Figure 3C). As observed with other TGF-β1 mediated effects, high concentrations of active TGF-β1, unfiltered latent TGF-β1, or combinations did not significantly upregulate NEU3 [68,69]. Together, these results indicate that the presence of latent TGF-β1 potentiates the sensitivity of cells to TGF-β1 with respect to NEU3 upregulation.

3.5. The NEU3 → TGF-β1 → NEU3 Feedback Loop Is Inhibited by NEU3 Inhibitors

The human and mouse NEU3 inhibitors 2-AP and AMPCA inhibit pulmonary inflammation, decrease levels of active TGF-β1 in the lungs, and decrease pulmonary fibrosis in a mouse bleomycin model [28]. To test if the NEU3 inhibitors influence the latent TGF-β1 amplification effect we observed, the latent TGF-β1 amplification experiment was repeated in the presence and absence of 2-AP and AMPCA. 0.1 ng/mL of latent TGF-β1 was mixed with the three suboptimal concentrations of active TGF-β1 (0.1, 0.3, 1 ng/mL) in the presence and absence of 10 nM 2-AP or AMPCA for 5 min and NEU3 levels were measured by immunofluorescence. In protein free media alone and 0.1 ng/mL of latent TGF-β1 alone, 2-AP and AMPCA did not significantly change the levels of NEU3 (Figure 3D). In contrast, both 2-AP and AMPCA blocked the amplification effect of NEU3 for all three active TGF-β1 concentrations (0.1 ng/mL latent TGF-β1 + 0.1, 0.3, or 1 ng/mL active TGF-β1). For unknown reasons, 0.1 ng/mL active TGF-β1 + 0.1 ng/mL latent TGF-β1 + 2-AP decreased NEU3 levels below baseline (Figure 3D). These data indicate that the NEU3 inhibitors 2-AP and AMPCA can block the positive feedback loop, inhibiting the upregulation of NEU3.

3.6. LAP Itself Upregulates NEU3 at 5 Min

Although 3 ng/mL of active TGF-β1 alone is needed to upregulate NEU3 expression (Figure 3A), a combination of 1 ng/mL of active TGF-β1 with only 0.01 ng/mL of latent TGF-β1 was sufficient to significantly increase NEU3. This suggests that even if all of the active TGF-β1 were released from LAP in the latent TGF-β1 complex, the total concentration of active TGF-β1 would still remain below the 3 ng/mL threshold. Increased levels of NEU3 with a total possible TGF-β1 concentration below 3 ng/mL were also observed for the 0.1 and 0.3 ng/mL active TGF-β1 combinations (Figure 3B,C). This suggested the LAP component of the activated latent TGF-β1 complex may exert its own bioactivity to contribute to NEU3 upregulation that is independent of TGF-β1, as independent functions of LAP have been observed in other instances [13,70,71,72]. To determine whether LAP alone can upregulate NEU3, recombinant human LAP was added to human lung fibroblasts for 5 min. NEU3 was significantly upregulated with 0.01, 0.03, 0.1, 0.3, and 1 ng/mL of LAP (Figure 4A). We verified that the recombinant LAP protein did not have any detectable active TGF-β1 or any other protein contaminant (Supplementary Figure S6A). Individual concentrations of lipopolysaccharide (LPS), a potent immunologic stimulant from Gram negative bacteria that frequently contaminates commercial reagents [73], did not significantly affect NEU3 levels (1-way ANOVA, Dunnett’s test), suggesting that the LAP (and TGF-β1) effect on NEU3 levels was not due to a possible LPS effect (Supplementary Figure S6B). However, for unknown reasons, there seemed to be a sex-specific effect for LPS; combining the data from the 10 LPS concentrations, LPS caused NEU3 in male cell lines to be 109 ± 2 and NEU3 from female cell lines to be 83 ± 3% of the no-LPS control (p < 0.0001, t-test). To determine whether LAP and TGF-β1 have a synergistic effect to upregulate NEU3, suboptimal levels of LAP (0.001 and 0.003 ng/mL) and suboptimal levels of TGF-β1 (0.1 and 0.3 ng/mL) were combined and added to human lung fibroblasts for 5 min. As seen previously (Figure 3 and Figure 4), suboptimal concentrations of LAP and TGF-β1 alone did not upregulation NEU3, but the combination of these suboptimal concentrations did significantly upregulate NEU3 (Figure 4B). Overall, these data indicate that LAP rapidly upregulates NEU3 and works synergistically with TGF-β1 to potentiate TGF-β1 signaling.

3.7. DDX3 Mediates NEU3 Upregulation by LAP and TGF-β1, and Exposure to TGF-β1, but Not LAP, Causes DDX3 Dephosphorylation

TGF-β1 increases the binding of the RNA helicase DDX3 to a common 20 nucleotide motif which is found in 180 human lung fibroblast mRNAs (including NEU3) whose levels are not significantly affected by TGF-β1, but whose translation is upregulated by TGF-β1 [49]. RK-33 is a DDX3 inhibitor that attenuates bleomycin-induced pulmonary fibrosis in young male mice [49]. To determine whether DDX3 inhibition attenuates the rapid upregulation of NEU3, HLFs were pre-treated with RK-33 then stimulated with either TGF-β1 or LAP for 5 min, and NEU3 was measured. As observed above, a 5 min treatment with TGF-β1 and LAP increased NEU3 levels, and RK-33 blocked this upregulation (Figure 5A,B). For unknown reasons, LAP but not TGF-β1 in the presence of RK-33 decreased NEU3 levels to levels below those seen in control cells. Changes in phosphorylation are often involved in signaling cascades [74]. To determine if TGF-β1 and/or LAP affects DDX3 phosphorylation, HLFs were treated with either TGF-β1 or LAP for 5 min and the phosphorylation level of DDX3 was measured and compared to total DDX3 levels. Exposure of cells to TGF-β1 caused dephosphorylation of DDX3 within 2 min without significantly changing DDX3 levels (Figure 5C–E and Supplementary Figure S7) whereas exposure of cells to LAP had no significant effect on phosphorylation or total DDX3 levels (Figure 5F–G and Supplementary Figure S7). Together, these data indicate that DDX3 is required for the rapid NEU3 upregulation by TGF-β1 and LAP, and suggest that LAP and TGF-β1 may converge on DDX3 through distinct upstream mechanisms.

4. Discussion

In this report we show that, unlike transcriptionally regulated proteins that increase after several hours of TGF-β1 exposure, NEU3 protein levels rise rapidly within 5 min both intracellularly and extracellularly. The rapid upregulation of NEU3 is independent of new transcription, supporting the idea that TGF-β1-induced upregulation of NEU3 is regulated on the translational level. NEU3 activates latent TGF-β1, and the released active TGF-β1 upregulates NEU3 in a rapid positive feedback loop. This upregulation is potentiated by the latent TGF-β1 complex. Notably, the LAP itself can upregulate NEU3 and works synergistically with TGF-β1. The NEU3 → TGF-β1 → NEU3 positive loop is inhibited by the NEU3 inhibitors 2-AP and AMPCA, supporting the direct role of NEU3 in this loop (Figure 6). In line with previous studies demonstrating a role of DDX3 in NEU3 translational regulation [49], we find that the rapid NEU3 upregulation by TGF-β1 and LAP is blocked by inhibiting DDX3 with RK-33. Exposure of cells to TGF-β1 induces rapid DDX3 dephosphorylation within 2 min, but exposure to LAP has no significant effect on DDX3 phosphorylation, indicating that these two factors may converge at DDX3 through distinct mechanisms (Figure 6). In our studies we tested male and female cell lines from both healthy donors and IPF patients. For our experiments we found no significant differences between cell lines derived from healthy donors vs. IPF patients. For unknown reasons, we did observe that in some cases, some female cell lines had a slightly stronger NEU3 response than male cell lines following TGF-β1 stimulation.

Fibrosis appears to be due to constitutively activated wound-healing mechanisms [48]. From an evolutionary standpoint, a rapid wound-healing response is crucial to minimize infection and further damage [6]. A small amount of positive feedback can significantly increase the sensitivity and response time of a system [76]. TGF-β1 is pre-positioned in the latent form in the extracellular environment, ready for rapid activation [9,10]. In incisional wounds, active TGF-β1 is detectable within minutes [77], and subsequent TGF-β1-mediated changes in phosphorylation of downstream effectors are also detected within minutes [78,79]. Together, this supports the idea that a fast positive feedback loop mechanism evolved to contribute to a rapid wound-healing response.

For translation to be rapidly upregulated without de novo NEU3 mRNA synthesis, the NEU3 transcript is likely pre-made and stored in the cytoplasm halted at the level of translation, as seen with other RNAs [80,81]. For instance, in yeast, in preparation for stressful conditions, basal levels of mRNAs are transcribed and stored in the cytoplasm. Under stressful conditions, these mRNAs are translated within minutes. This is seen for the yeast proteins GCN4 [82,83,84], HAC1 [85], and ICY2 [86,87]. Similar mechanisms can be seen in human cells with mRNAs that are stored in stress granules [88]. Together with our current findings, these results highlight a potential mechanism where NEU3 mRNA is pre-made and stored in the cytoplasm, halted at the level of translation. Upon activation of latent TGF-β1, the released active TGF-β1 and LAP rapidly induce a translational regulator (such as DDX3) which activates the translation of NEU3 mRNA within 5 min to form the TGF-β1 → NEU3 part of the TGF-β1 → NEU3 → TGF-β1 positive feedback loop. NEU3 then exerts its enzymatic activity to activate TGF-β1 that is pre-positioned at the extracellular matrix, generating the NEU3 → TGF-β1 part of the positive feedback loop for a rapid wound-healing response. The LAP component of the latent TGF-β1 complex exerts independent bioactivity to rapidly upregulate NEU3 and works synergistically with TGF-β1. Since LAP desialylation can release active TGF-β1 [28], it is possible that after NEU3 cleavage, the desialylated form acquires a distinct signaling property that acts at an unknown receptor to reinforce the TGF-β1 → NEU3 → TGF-β1 feedback loop. Although LAP is classically presumed to function only as a sequestering agent for TGF-β1, LAP possesses independent biological activity in other contexts including regulating cell adhesion and migration through its engagement with integrins through its RGD motif [13,72], and modulating immune responses as a chemotactic and anti-inflammatory ligand [70]. There may be multiple receptors that LAP associates with, but it is most well-known to bind to integrins alphavbeta3, alphavbeta5, alphavbeta6, and alphavbeta8 [71,72]. The observation that 30 ng/mL and lower concentrations of latent TGF-β1 do not significantly upregulate NEU3 (Figure 3C), whereas 0.01–1 ng/mL LAP increases NEU3 (Figure 4A) suggest that the LAP in the latent TGF-β1 complex has a relatively poor ability to upregulate NEU3 compared to free LAP. These observations support a model where after activation and release of TGF-β1, LAP acquires a distinct signaling capability that reinforces the positive feedback loop.

Changes in phosphorylation are one of the most rapid and effective regulatory mechanisms in signaling cascades [89,90]. Our data suggest that exposure of cells to TGF-β1 rapidly induces dephosphorylation of DDX3, which in turn upregulates NEU3. Since dephosphorylation is a first-order reaction, depending only on the concentration of the phosphorylated substrate, it can proceed more rapidly than phosphorylation, which is a second-order reaction requiring both the substrate and ATP [91]. This kinetic difference often results in faster signal termination or activation through phosphatase activity. Similar rapid dephosphorylation events have been observed in other pathways such as MAPK signaling cascades, where phosphatase-mediated responses can occur within seconds [89]. Interestingly, although LAP rapidly upregulates NEU3 and is blocked by RK-33, exposing cells to LAP does not significantly influence the phosphorylation (at least as assessed with the TALON PMAC kit) or total protein level of DDX3, suggesting that LAP engages DDX3 through a phosphorylation-independent mechanism. This also suggests the possibility that TGF-β1 mediated dephosphorylation of DDX3 is not required for the NEU3 response. Our previous work demonstrated that TGF-β1 enhances DDX3 binding to the common 20-nucleotide motif (shared among several mRNAs regulated by TGF-β1 on the translational level) after 1 h of TGF-β1 stimulation but not at 30 min [49]. Although DDX3 binding to the motif showed an upward trend at 30 min, this association was not statistically significant. There are multiple possible models that could explain how TGF-β1 and LAP converge on DDX3 to produce the same output. One possibility is that LAP and/or TGF-β1 increase DDX3 binding to the common 20-nucleotide motif at early time points, which was not previously detected. Alternatively, either LAP and/or TGF-β1 may cause DDX3 to interact with a different region of the NEU3 transcript. DDX3 has multiple binding sites for RNA [92,93] and preferentially binds structured RNAs with G-quadruplex elements through its intrinsically disordered N-terminal region [94]. The NEU3 transcript contains multiple predicted G-quadruplex structures (Figure 6), raising the possibility that DDX3 binds these regions to increase translation at early time points. Another possibility is that TGF-β1-mediated DDX3 dephosphorylation alters its interaction network or helicase activity in a way that indirectly enhances NEU3 translation without direct RNA binding, such as modulating components of the translation machinery, whereas LAP modulates a different aspect such as DDX3 binding, which both result in rapid NEU3 upregulation.

There are at least two reasons that the rapid positive feedback loop does not invariably turn on in response to a single extracellular molecule of NEU3, TGF-β1 or LAP. First, NEU3, TGF-β1, and LAP have finite lifetimes allowing, as with other extracellular factors, a basal equilibrium to be reached between release rate and degradation/uptake rate. Second, the TGF-β1 signal transduction pathway has negative feedback components such as Smad7 [95], developmental endothelial locus-1 (Del-1) [96], and the RelA subunit of nuclear factor κB (NF-κB) [97], causing a threshold effect where low levels of TGF-β1 (and LAP) essentially do not activate the pathway. We hypothesize that in a wound or in fibrosis, a sufficiently high level of the various factors that increase NEU3 levels activate the loop. Under normal conditions, a small amount of TGF-β1 and/or LAP causes a small upregulation of NEU3 which in turn increases the amount of TGF-β1 and LAP to a new steady state level where the efficiency of the system is controlled (loop gain less than 1). In this normal environment, a decreased level of the injury signals and/or upregulation of inhibitory signals indicate a completion of healing and cause the loop to return to quiescent levels of NEU3, TGF-β1, and LAP. In fibrosis, three possibilities that could cause the loop to stay in an activated state include an excess of signals that activate the loop (due to increased amount of signal or decreased degradation), insufficient levels of signals that inhibit the loop, or an increase in the efficiency of the loop where NEU3 more efficiently upregulates TGF-β1 and LAP and/or TGF-β1 and LAP more efficiently upregulate NEU3 (loop gain equal to or more than 1).

High concentrations of TGF-β1 and LAP do not upregulate NEU3, which likely reflects a tightly controlled protective mechanism, as seen in multiple biological systems. For example, hormones often operate within complex dose-dependent negative feedback loops where high levels of a downstream effector downregulates its own production or secretion [98]. Additionally, continuous receptor stimulation can trigger receptor downregulation through decreased affinity and internalization [99]. However, these mechanisms typically take longer than a few minutes. A rapid downregulation mechanism could include NEU3 degradation in response to high activator concentrations [100,101], rapid phosphorylation of downstream effectors that could inhibit NEU3 translation [78,79], analogous to short-term desensitization of G-protein coupled receptors where phosphorylation of the receptor by GPCR kinase recruits β-arrestin to rapidly shut off the response [102,103], or rapid receptor phosphorylation and uncoupling which is seen with serine threonine receptors [104]. Although the mechanism remains unclear, this non-linear dose effect indicates that NEU3 upregulation is likely a controlled process with mechanisms in place to avoid overstimulation. Together, this report supports a model where NEU3 mRNA is stored in the cytoplasm, ready for translational activation for a very rapid wound-healing response which is amplified by a positive feedback loop that may get stuck in an ‘on’ state in fibrosis.

5. Conclusions

Together, these findings identify a rapid profibrotic translation-based positive feedback loop involving TGF-β1, LAP, and NEU3. TGF-β1 induces a rapid increase in both intracellular and extracellular NEU3 protein levels within 5 min through translational upregulation, independent of transcription. Surprisingly, in addition to TGF-β1, LAP itself can independently upregulate NEU3 within 5 min, and works synergistically with TGF-β1, reframing LAP from a passive protein to an active signaling component. Further, DDX3 inhibition effectively blocks the rapid NEU3 induction by both TGF-β1 and LAP. We propose that this rapid feedback mechanism likely evolved to support efficient wound healing, but when persistently engaged, becomes locked in an “on” state that drives fibrosis.