Abstract

Objective

   Excessive extra-cellular-matrix production and uncontrolled
   proliferation of the fibroblasts are characteristics of many fibrotic
   diseases, including idiopathic pulmonary fibrosis (IPF). The
   fibroblasts have enhanced glutaminolysis with up-regulated glutaminase,
   GLS1, which converts glutamine to glutamate. Here, we investigated the
   role of glutaminolysis and glutaminolysis-derived metabolite
   α-ketoglutarate (α-KG) on IPF fibroblast phenotype and gene expression.

Methods

   Reduced glutamine conditions were carried out either using
   glutamine-free culture medium or silencing the expression of GLS1 with
   siRNA, with or without α-KG compensation. Cell phenotype has been
   characterized under these different conditions, and gene expression
   profile was examined by RNA-Seq. Specific profibrotic genes (Col3A1 and
   PLK1) expression were examined by real-time PCR and western blots. The
   levels of repressive histone H3K27me3, which demethylase activity is
   affected by glutaminolysis, were examined and H3K27me3 association with
   promoter region of Col3A1 and PLK1 were checked by ChIP assays. Effects
   of reduced glutaminolysis on fibrosis markers were checked in an animal
   model of lung fibrosis.

Results

   The lack of glutamine in the culture medium alters the profibrotic
   phenotype of activated fibroblasts. The addition of exogenous and
   glutaminolysis-derived metabolite α-KG to glutamine-free media barely
   restores the pro-fibrotic phenotype of activated fibroblasts. Many
   genes are down-regulated in glutamine-free medium, α-KG supplementation
   only rescues a limited number of genes. As α-KG is a cofactor for
   histone demethylases of H3K27me3, the reduced glutaminolysis alters
   H3K27me3 levels, and enriches H3K27me3 association with Col3A1 and PLK1
   promoter region. Adding α-KG in glutamine-free medium depleted H3K27me3
   association with Col3A1 promoter region but not that of PLK1. In a
   murine model of lung fibrosis, mice with reduced glutaminolysis showed
   markedly reduced fibrotic markers.

Conclusions

   This study indicates that glutamine is critical for supporting
   pro-fibrotic fibroblast phenotype in lung fibrosis, partially through
   α-KG-dependent and –independent mechanisms, and supports targeting
   fibroblast metabolism as a therapeutic method for fibrotic diseases.

   Keywords: Glutamine, α-KG, H3K27me3, Col3A1, PLK1, Lung fibrosis,
   Running head

Highlights

     * •
       IPF fibroblasts require glutamine to maintain a pro-fibrotic
       phenotype.
     * •
       Glutamine metabolite, α-KG, has limited effects to recover
       pro-fibrotic phenotype under glutamine-free condition.
     * •
       Col3A1 and PLK1 are among the genes reduced by glutamine-free
       condition; adding α-KG rescues Col3A1 but not PLK1.
     * •
       The down-regulated Col3A1 and PLK1 in glutamine-free condition are
       likely through increased H3K27me3 at their promoter.
     * •
       Reduced glutaminolysis decreases lung fibrosis fibrotic markers in
       vitro and in vivo.

1. Introduction

   Glutamine (Gln) is the most abundant amino acid in human circulation
   and intracellular pools; glutamine is also involved in many cellular
   processes [[37]1,[38]2]. Although it is a non-essential amino acid, it
   becomes conditionally essential during catabolic stress, such as during
   states of pregnancy, or critical illness [[39]3,[40]4]. Glutamine is
   converted to glutamate by glutaminase (GLS), and then to
   α-ketoglutarate (α-KG) by glutamate dehydrogenase to enter the
   tricarboxylic acid (TCA) cycle [[41]5]. GLS is the initial enzyme in
   the glutaminolysis pathway. GLS was reported as a key player in cancer
   cell metabolism that promotes tumor growth, and therefore targeting GLS
   has been explored as a potential therapy for cancer treatment
   [[42]6,[43]7]. GLS isoform GLS1 is increased in differentiated
   fibroblasts and fibrotic lung tissues [[44]8].

   Many studies have established the link between metabolism and the cell
   epigenetic status [[45]9,[46]10]. A lot of molecules derived from
   intermediary metabolism are key regulators of chromatin enzymatic
   activities [[47]11]. α-KG is an important metabolite derived from
   glutaminolysis and has a critical role in different metabolic and
   cellular pathways. α-KG serves as a cofactor for multiple dioxygenases
   [[48]12,[49]13]. It regulates the activity of prolyl-4 hydroxylase,
   which controls the biosynthesis of collagen [[50]14]. It is also a
   cofactor for epigenetic enzymes [[51]15], such as ten-eleven
   translocation hydroxylases involved in DNA methylation [[52]16] and
   Jumonji C domain containing (JMJC) lysine demethylases which are major
   histone demethylases [[53]12]. One hypothesis is that altered
   glutaminolysis may affect cellular histone methylation via providing
   intracellular α-KG, which in turn regulates the transcriptional and
   epigenetic state of cells [[54]13]. In our previous studies, we
   demonstrated that decreased histone demethylase activity in
   glutamine-free medium can be rescued by adding exogenous α-KG to the
   medium, which affects the histone demethylase substrate H3K27me3
   [[55]17]. The changes in H3K27me3 levels likely alter the expression of
   the genes that are associated with H3K27me3 thus affecting the cell
   phenotype.

   Profibrotic cell phenotype is characteristics of IPF fibroblast, which
   exhibits increased collagen production, uncontrolled proliferation
   [[56]18], and enhanced glutaminolysis with upregulated GLS1 [[57]19].
   Glutamine controls the synthesis of fibrotic markers such as collagen
   at both the transcriptional and translational levels [[58]20,[59]21]. A
   study in pulmonary fibrosis indicated that glutaminolysis promotes
   collagen translation and stability through α-KG mediated mTOR
   activation and proline hydroxylation [[60]8]. Collagen type I and III
   are the main components of extracellular matrix in pulmonary fibrosis
   [[61]22]. Previously, we have showed that Col3A1 RNA expression is
   associated with histone modification H3K27me3 [[62]23], which histone
   mark could be affected by changes of glutamine levels in IPF
   fibroblasts [[63]17]. In the current study, we explored the role of
   glutaminolysis and its metabolite α-KG on IPF fibroblast phenotype and
   on gene expression profile. We examined whether glutaminolysis
   regulates gene expression through altered histone modification in
   response to metabolic changes. Our study showed that metabolite levels
   provide an additional layer of epigenetic control and transcriptional
   regulation in lung fibroblasts.

2. Materials and methods

2.1. Cell culture and treatments

   University of Alabama at Birmingham (UAB) Tissue Procurement Facility
   provided human primary IPF lung culture of fibroblasts, which were
   derived from de-identified tissues ([64]Table 1S online). The protocol
   is approved by the UAB Institutional Review Board. IPF was diagnosed by
   a multidisciplinary approach according to the guidelines from American
   Thoracic Society/European Respiratory Society [[65]24]. The fibroblasts
   were used before passage 5, and were kept in Dulbecco׳s modified
   Eagle׳s medium (DMEM, Life Technologies, Grand Island, NY) with 10%
   fetal bovine serum (FBS, Life Technologies), 100 units/ml penicillin,
   100 μg/ml streptomycin, 1.25 μg/ml amphotericin B, and 2 mm L-glutamine
   (G1251, Sigma) at 5% CO[2] at 37 °C. For glutamine-free condition, the
   culture medium was changed to glutamine-free (0 mM of L-glutamine)
   while all the other components remain the same when the cells reached
   80% confluent. For the condition with added α-KG, the cell permeable
   α-ketoglutarate ([66]AK126628, Ark Pharm Inc., Arlington Heights, IL)
   was added to L-glutamine free medium at 2 mM. The cells were harvested
   for various assays after 48 h or as indicated.

2.2. siRNA transfection

   GLS1 siRNA (NM_0149,054) and Non-targeting (NT) siRNA sequences are
   listed in [67]Table 2S (online supplementary data). siRNA transfections
   were carried out according to manufacturer's instructions with Opti-MEM
   (Thermo Fisher Scientific) using lipofectamine RNAi/MAX (Invitrogen).
   The GLS1 siRNA transfection in IPF lung fibroblasts were done in full
   medium (2 mM L-glutamine) as indicated above, cells were subjected to
   assays after 48 h of transfection.

2.3. Proliferation assay

   The primary IPF fibroblasts were seeded into a 12-well plate at a
   density of 10,000 cells/well in full medium for overnight. Then the
   culture medium was changed to fresh full medium, or glutamine-free
   medium, or glutamine-free medium with 2 mM α-KG. The cells were counted
   using a conventional hemocytometer counting chamber at 24 h, 48 h and
   72 h following published protocol [[68]25].

2.4. Cell migration assay

   When the cells are confluent as a monolayer in full medium, the medium
   was discarded, and a scratch was created with a sterile P20 pipette
   tip. After rinsed the plate with PBS to remove suspended cells, the
   cells were cultured in fresh medium of control (full medium with 2 mM
   glutamine), or glutamine-free (full medium without glutamine), or
   glutamine-free with 2 mM of α-KG. The wound closure was monitored and
   photographed with a Keyence microscope (Itasca, IL) at indicated time
   points. The distance of the edges of the scratch were measured with the
   obtained image.

2.5. Immunofluorescence

   Immunofluorescence staining was performed as descried previously
   [[69]26]. Rabbit anti-Col3A1 was purchased from Abclonal (cat# A3795,
   Woburn, MA), H3K27me3 (Cat#9733) was from Cell Signaling. DAPI
   (Cat#D1306, Thermo Fisher) was used for nuclear staining. Images were
   acquired with a Keyence microscope (Itasca, IL). Fluorescence signals
   were quantitated with ImageJ software.

2.6. RNA extraction and real-time RT-PCR

   RNA was extracted with RNeasy Mini Kit (Qiagen, Valencia, CA), and
   transcribed into cDNA with a cDNA synthesis kit (Takara Bio, Mountain
   View, CA). Real-time RT-PCR was performed in triplicate and normalized
   to β-actin using the ΔΔCt method [[70]27]. Primers used to examine RNA
   level expression are listed in [71]Table 2S.

2.7. RNA-sequencing and data analysis

   Extracted RNA from three primary IPF fibroblast cell lines under the
   three conditions described as above was sent to Genewiz (South
   Plainfield, NJ) for library preparation and RNA-Sequencing (RNA-Seq).
   Details of the data analyzed are described in online supplementary
   data.

2.8. Protein extraction and Immunoblotting

   Whole cell lysates were prepared after washing the cells in cold PBS
   and lysed with 2x SDS reducing sample buffer containing protease
   inhibitors. The lysates protein concentration was quantified using a
   Micro BCA Protein Assay kit (Thermo Scientific). Lysates were then
   subjected to SDS-PAGE; western immunoblotting was performed as
   described previously [[72]28]. Immunoblots were imaged with an Amersham
   Biosciences 600 Imager (GE Healthcare). Signals were quantitated with
   ImageJ software. For antibodies, anti-Col3A1 (A3795) was from Abclonal
   (Woburn, MA), anti-β-actin (#2128) and anti-PLK1 (#4513) were from Cell
   signaling (Beverly, MA), anti-GLS1 (cat #12885-1-AP) was from
   Proteintech (Rosemont, IL).

2.9. Chromatin immunoprecipitation assays

   Chromatin immunoprecipitation (ChIP) assays were performed as per the
   manufacturer's protocol (ab500, Abcam, Cambridge, MA) with minor
   modifications [[73]29]. Antibody against H3K27Me3 (cat#39155) used for
   ChIP assays is from Active Motif (Carlsbad, CA). For ChIP assays,
   primers used are listed in [74]Table 2S. ChIP-DNA was amplified by
   real-time PCR, using SYBR® Green PCR Master Mix (Life Technologies).
   Results are normalized to input DNA.

2.10. GLS1^+/− heterozygous mice model of bleomycin induced lung fibrosis

   All animal studies were performed in accordance with UAB Institutional
   Animal Care and Use Committee approved protocols. GLS1^+/− heterozygous
   mice (#017956) were purchase from Jackson laboratory. At age of 8-week,
   both wild type (WT) and GLS1^+/− heterozygous (Hets) mice were
   subjected to saline or bleomycin injury. A single dose of normal saline
   or bleomycin sulfate at 1.5 u/kg body weight was instilled
   intratracheally. All mice were sacrificed at day 28 post injury. Whole
   lung tissues were prepared for whole lung lysates for western blots.
   Some mice lung tissues were prepared for histology with H&E stains. A
   semiquantitative Ashcroft scale, ranging from 0 to 8, was used to grade
   severity of lung fibrosis after bleomycin injury [[75]30].

2.11. Statistical analysis

   Data are presented as the mean ± standard deviation (SD). All data were
   statistically analyzed using GraphPad Prism 5.0 (La Jolla, CA). One-way
   ANOVA was used to compare between multiple groups; comparison between
   two groups of the same mice were with Student's t test; comparison
   between two groups of different kinds of mice (wild type vs
   heterozygous) were used Kolmogorov–Smirnov test. A P value of <0.05 was
   considered to be statistically significant.

3. Results

3.1. Glutaminolysis regulates IPF fibroblast proliferation, migration, and
collagen production

   The dysregulated fibroblast proliferation, excessive extracellular
   matrix production, increased migration and apoptosis resistance are
   some of the characteristics of pro-fibrotic fibroblasts in IPF
   [[76]18,[77]31]. Previously, we demonstrated that reduced glutamine
   levels in cell culture increases apoptotic susceptibility of IPF
   fibroblasts [[78]17]. Here, we further examined the effects of
   glutamine withdrawal on IPF fibroblasts proliferation, migration and
   collagen production. At day one, the cells were plated at about 10,000
   cell/well in 12-well plate overnight in control medium. The culture
   medium was then changed to control (full medium) or glutamine-free
   medium. Cell count was determined at day 2 and day 3 after changing the
   media. The cells in control medium reached near 40,000 cell/well at day
   3, whereas the cells in glutamine-free medium hardly doubled
   ([79]Figure 1A). Under similar culture condition, i.e. after cells
   formed a monolayer, scratch assays were carried out to monitor
   fibroblast migration with or without glutamine in the culture medium.
   The cells in control medium migrated much faster than the ones in
   glutamine-free medium, and filled the scratch after 24 h. In contrast,
   the gap of the scratch was still visible after 24 h in glutamine-free
   medium ([80]Figure 1B top two panels). By 48 h, the gaps were filled
   under both conditions, but the glutamine-free condition seemed with
   fewer cells.

Figure 1.

   [81]Figure 1
   [82]Open in a new tab

   Cell proliferation, migration and collagen III deposition in IPF lung
   fibroblasts cultured in medium containing 2 mM glutamine (control),
   0 mM glutamine and 0 mM glutamine plus 2 mM α-KG. The media of cultured
   cells under control condition, was changed to fresh medium with 2 mM
   glutamine, 0 mM glutamine, or 0 mM glutamine with 2 mM α-KG when cells
   reached 80% confluency. A. Cell numbers were counted each day during
   3-day period under these different culture conditions to assess cell
   proliferation (see methods for detailed protocols). Growth curve were
   plotted as average ± standard error of cell numbers from three
   different IPF cell lines of each day. The difference was significant on
   day 3 under each condition (∗p < 0.05). B. Cell migration was examined
   after scratching the cell monolayer with a sterile pipette tip, the
   migration of the cells were recorded by imaging at different time
   points at 4×. The edges of the scratches were marked with dashed lines,
   which were shown by the distance bar with arbitrary unit under 0, 14,
   or 24 h after the scratches were created. The histogram of scratch
   distance under each condition was made with average ± standard error of
   three cell lines. ∗p < 0.05 at 24 h vs 0 h. C. Representative images of
   immunofluorescence stain of collagen III in IPF lung fibroblasts under
   different culture conditions after 48 h. Cell were stained with
   antibodies against Col3A1 (red), and DAPI (blue). Images were also
   merged to visualize both stains in one image. Pictures were taken at
   20× with a Keyence microscope. Histogram is the average ± standard
   error of at least three sildes of immunofluorescence intensity of
   Col3A1 stains measured by ImageJ program. ∗p < 0.05, Gln 0 mM vs Gln
   2 mM, or α-KG 2 mM.

   We then examined if the collagen deposition is affected by these
   conditions. Collagen III is increased in IPF lung fibroblasts [[83]23],
   collagen III presence was shown by immune-fluorescent studies after
   48 h of culturing the cells in control or glutamine-free medium
   ([84]Figure 1C, left and middle panels). Noticeably less staining of
   collagen III in the glutamine-free medium was observed when compared to
   the control medium condition. These data indicate that glutamine
   withdrawal from the culture medium slows IPF fibroblasts proliferation,
   migration and collagen production.

3.2. α-KG restores IPF fibroblasts collagen production in the absence of
glutamine, but not affecting cell proliferation and migration

   Since glutaminolysis-derived α-KG enters the tricarboxylic acid (TCA)
   cycle [[85]32], we reasoned that the addition of exogenous α-KG may
   reverse some of the effects of reduced glutaminolysis on IPF fibroblast
   phenotype. To test this hypothesis, we cultured IPF fibroblasts under
   the glutamine-free condition, in the presence or absence of 2 mM α-KG.
   With exogenous α-KG in the glutamine-free medium, we did not observe
   increased cell proliferation. The proliferation was similar to or even
   slower compared to IPF fibroblasts proliferation under glutamine-free
   condition ([86]Figure 1A). Similar results were observed for migration
   of IPF fibroblasts cultured in glutamine-free medium, in the presence
   or absence of α-KG. With the exogenous α-KG, by 24 h, only some cells
   filled the gap ([87]Figure 1B, bottom panel), in contrast, the gap in
   full medium control was closed. However, immunofluorescence staining of
   collagen III showed that the addition of α-KG to glutamine-free medium
   restored IPF fibroblasts synthesis of collagen III. IPF fibroblasts
   with added α-KG demonstrated heightened fluorescent signal associated
   with collagen III compared to those cells in glutamine-free condition
   without α-KG, and similar to cells cultured in control media (with 2 mM
   glutamine) ([88]Figure 1C, right panel). These data suggests that the
   addition of α-KG in glutamine-free medium has limited effects on cell
   proliferation or migration, but rescues the decreased collagen induced
   by glutamine withdrawal from the culture medium.

3.3. α-KG partially rescues IPF fibroblast gene expression profile under
glutamine-free condition

   We next examined the regulation of IPF fibroblast gene expression by
   glutaminolysis. As previously described, IPF fibroblasts were cultured
   in control or glutamine-free media supplemented with or without α-KG.
   We performed RNA-Seq with primary IPF fibroblasts derived from three
   different individuals under these culture conditions. When cells
   reached 80% confluency, the control medium was changed to fresh medium
   containing either 2 mM glutamine, glutamine-free (0 mM), or
   glutamine-free plus α-KG (0 mM glutamine with 2 mM α-KG). The cells
   were cultured for 48 h before collecting RNA for sequencing.
   Principal-components analysis (PCA) of our RNA-seq data projected a
   strikingly large variance between the cells cultured in glutamine-free
   medium compared to cells cultured in control medium. The addition of
   α-KG only rescued limited expression profile towards those observed
   under control condition (with 2 mM glutamine). Each cell line seems to
   cluster together under glutamine-free and α-KG conditions
   ([89]Figure 2A). However, the greatest variance is observed between the
   donors, which reflected the heterogeneity of the IPF fibroblasts
   [[90]33]. PCA also identified one sample under glutamine-free condition
   to be an outlier; therefore, for further analysis, we only used two
   samples that were obtained from cells cultured under glutamine-free
   condition (used IPF-B and C, under Gln 0 mM), while we used all three
   samples for the other two conditions (Gln 2 mM and Gln 0 mM plus α-KG
   2 mM). There are many genes that are down-regulated by glutamine
   withdrawal ([91]Figure 1SA, online supplementary data). However, adding
   α-KG to the glutamine-free medium only has limited effects on a small
   portion of genes that were susceptible to glutamine withdrawal from the
   culture media ([92]Figure 1SB–D).

Figure 2.

   [93]Figure 2
   [94]Open in a new tab

   RNA-Seq data from three primary IPF lung fibroblasts culture in the
   presence of 2 mM glutamine, 0 mM glutamine, or 0 mM glutamine plus α-KG
   2 mM. RNAs were collected 48 h after treatment and subjected to
   RNA-Seq. A. PCA plot showing sample clustering. Three different samples
   are shown by different colors. Lung fibroblasts from IPF patient A
   (Red), patient B (Green), and patient C (Blue), were cultured with 2 mM
   glutamine (square), 0 mM glutamine (triangle), or 0 mM glutamine with
   2 mM α-KG (circle). Significant variability can be observed between the
   different samples. Following analysis for the 0 mM glutamine condition
   only used samples from patients B and C (green and blue triangle), see
   text for details. B. Pathway enrichment analysis of altered genes under
   glutamine 0 mM vs 2 mM glutamine. Top 10 most significantly enriched
   terms are presented in the figure. C. GO term enrichment analysis of
   altered genes in glutamine 0 mM vs 2 mM. GO term enrichment (Biological
   Process database) showed that most significant terms are related to
   cell cycle. GO term analysis returned 168 highly down-regulated genes
   (fold change >1.5) associated with the term “cell cycle”. Top ranked
   most significantly enriched terms are presented in the table. D.
   Heatmap showing the top 20 most differentially expressed genes when
   compared to 2 mM glutamine (Gln2, q < 0.05). More RNA-Seq analysis are
   in online supplementary data. The gene expression in response to the
   addition of α-KG in the glutamine-free medium is shown with purple box.

   The RNA-Seq data from IPF cells indicated that glutamine withdrawal
   altered many pathways ([95]Figure 2B,C). GO term enrichment shows the
   most affected pathway is cell cycle ([96]Figure 2C). From the top 20
   most differentially expressed genes, some genes, such as BIRC5, have
   been previously reported to modulate IPF fibroblast phenotype [[97]17]
   ([98]Figure 2D, and online [99]Table 3S). Overall, the RNA-Seq data
   described above show significantly altered gene expression profile of
   IPF fibroblasts under condition of glutamine deficiency, however, with
   limited rescue following the addition of α-KG in the glutamine-free
   medium.

3.4. Glutaminolysis regulates Col3A1 and PLK1 expression via α-KG-dependent
and -independent mechanisms

   In fibrotic diseases, fibroblasts have been shown to express higher
   levels of Col3A1 [[100]23,[101]34]. Recent studies have demonstrated
   that glutaminolysis is increased in fibrosis [[102]8,[103]33,[104]35];
   and we observed glutamine-free condition would reduce Col3A1, which is
   rescued by adding α-KG ([105]Figure 1C). Thus, we further examined
   mechanisms by which glutaminolysis regulates Col3A1 expression in IPF
   fibroblasts. Under glutamine free condition, Col3A1 expression is
   decreased at protein and mRNA levels ([106]Figure 3A–C) suggesting a
   regulation of Col3A1 expression at both transcriptional and
   post-transcriptional levels. Cell-permeable α-KG added in the
   gluntamine-free medium, partially recovered Col3A1 expression both at
   protein and RNA levels ([107]Figure 3A–C). This observation is in
   consistence of the data that we showed in [108]Figure 1C, and suggests
   that glutaminolysis regulates Col3A1 expression via an α-KG-dependent
   mechanism.

Figure 3.

   [109]Figure 3
   [110]Open in a new tab

   Expression of Col3A1 and PLK1 in IPF fibroblasts grown in media of
   different glutamine concentrations. Col3A1 and PLK1 expression in IPF
   lung fibroblasts in culture medium with 2 mM glutamine (control), 0 mM
   glutamine, or 0 mM glutamine with 2 mM α-KG for 48 h. A/B: (A)
   Representative western blotting of Col3A1 expression, with β-actin as
   loading control. (B) Col3A1/β-actin determined by densitometry under
   culture conditions described above, n = 3 experimental repeats. (C)
   Col3A1 mRNA levels expression was assessed by real-time PCR. Triangles,
   squares, or circles indicate three different patient-derived IPF
   fibroblast lines. Expressed values represent mean ± SD; n = 3
   experimental replicates for each cell line. D/E: (D) Representative
   western blotting of PLK1 expression with β-actin as loading control.
   (E) PLK1 ratio to β-actin determined by densitometry under the same
   culture conditions as in D with n = 3 experimental repeats. F: PLK1
   mRNA levels expression were determined by real-time PCR. Triangles,
   squares, or circles indicate three different patient-derived IPF cell
   lines. Expressed values represent mean ± SD; n = 3 experimental
   replicates of each cell line. ∗p < 0.05, Glutamine 0 mM (Gln 0) vs Gln
   2 mM or Gln0 plus α-KG as indicated. Additional cell lines response
   under the same condition by western blots are shown in online data
   [111]Figure 2SA.

   From RNA-Seq data comparing control to glutamine-free condition, among
   the top 20 differentiated genes, PLK1 is one of the cell cycle related
   genes that was previously reported to be involved in lung fibroblasts
   proliferation [[112]36]. We further examined PLK1 expression under
   control and glutamine deficiency conditions. Similarly to Col3A1,
   glutamine-free condition induced a decrease in PLK1 expression at both
   the protein and mRNA levels ([113]Figure 3D–F). However, unlike Col3A1,
   addition of α-KG did not rescue PLK1 expression either at the protein
   or mRNA levels ([114]Figure 3D–F). This indicates that glutaminolysis
   regulates PLK1 expression via an α-KG-independent mechanism. Taken
   together, these data suggest that glutaminolysis differentially
   regulates Col3A1 and PLK1 expression via α-KG-dependent and
   -independent mechanisms.

3.5. GLS1 silencing mimics the effects of glutamine-free condition on Col3A1
and PLK1 expression

   GLS1 is the key enzyme that controls glutaminolysis in IPF, converts
   glutamine to glutamate [[115]35]. To determine whether
   glutamine-mediated regulation of Col3A1 and PLK1 involves GLS1, we
   silenced GLS1 using siRNA in IPF fibroblasts cultured under 2 mM
   glutamine containing control medium. 48 h post siRNA transfection,
   cells were collected to examine the genes expression. In GLS1-silenced
   cells, the Col3A1 and PLK1 expression was significantly down-regulated
   at both the protein and RNA levels ([116]Figure 4). The addition of
   α-KG to the control medium (with 2 mM glutamine), which bypass the
   blocked conversion of glutamine to glutamate, upregulated Col3A1 at
   both protein and mRNA levels ([117]Figure 4 A–C); however, the addition
   of α-KG failed to rescue PLK1 at protein or mRNA levels ([118]Figure 4
   A,B,D). Individual differences also play a role in response to altered
   metabolic conditions. Here, we showed representative data of two
   different IPF cell lines. In response to GLS1 silencing, one cell line
   has modest (represented by triangles in [119]Figure 4C) while the other
   one has robust reduction of Col3A1 expression ([120]Figure 4C). Despite
   the heterogeneity of IPF fibroblasts, this data confirms that reduced
   glutaminolysis down-regulates Col3A1 and PLK1 expression, while the
   metabolite α-KG only rescues a subset of genes that are α-KG dependent,
   such as Col3A1.

Figure 4.

   [121]Figure 4
   [122]Open in a new tab

   Expression of Col3A1 and PLK1 in IPF fibroblasts transfected with
   non-targeting (NT) or GLS1 siRNA. IPF cells are transfected with NT or
   GLS1 siRNA in the presence of 2 mM glutamine. GLS1-silenced cells were
   cultured in presence of glutamine 2 mM with or without adding α-KG
   2 mM. The cells were collected 48 h after transfection for western
   blots or real-time PCR. A/B: (A) Representative western-blotting to
   examine the protein levels for GLS1, Col3A1 and PLK1. (B) Densitometry
   analysis of Col3A1 and PLK1 associated signals: ration of Col3A1 and
   PLK1 to β-actin, as shown in A, n = 3 experimental replicates.
   ∗p < 0.05, GLS1 siRNA compared to NT siRNA, or GLS1 siRNA with 2 mM Gln
   plus α-KG. C/D: RNA expression of Col3A1 (C) or PLK1 (D) in two IPF
   fibroblast cell lines transfected with NT or GLS1 siRNA cultured in the
   presence of 2 mM glutamine with or without added α-KG as in A.
   Triangles or squares represent two different IPF patient-derived
   primary cells. Expressed values represent mean ± SD, 3 experimental
   replicates of each cell line. ∗p < 0.05, siRNA GLS1 vs NT, or vs siRNA
   GLS1 with 2 mM glutamine plus α-KG.

3.6. Col3A1 and PLK1 association with histone H3K27me3

   In previous studies, we demonstrated that α-KG availability directly
   affects H3K27me3 demethylases activities in lung fibroblasts [[123]17].
   Here, we checked overall baseline levels of H3K27me3 in IPF primary
   cell lines under different glutaminolysis conditions. We confirmed by
   western blotting that H3K27me3 levels are low in control medium (with
   2 mM glutamine), increased in glutamine-free medium, and reduced with
   added α-KG in glutamine-free medium ([124]Figure 5A,B and online
   [125]Figure 2SB). Immunofluorescence staining of H3K27me3 showed most
   intense staining under glutamine-free condition compared to control
   with 2 mM glutamine, or glutamine-free with added α-KG ([126]Figure 5C,
   and online data [127]Figure 3S).

Figure 5.

   [128]Figure 5
   [129]Open in a new tab

   Histone H3K27me3 in IPF fibroblasts and its association with Col3A1 or
   PLK1 promoter region. IPF lung fibroblasts were kept in control medium
   (with 2 mM glutamine), when the cells were at 70% confluency, medium
   was changed to fresh medium with glutamine 2 mM, 0 mM, or 0 mM with
   α-KG (2 mM) for 48 h. A. Representative western blots of H3K27me3
   overall baseline levels in two different IPF primary cell lines under
   the culture conditions described above, H3 is the loading control.
   Additional IPF cell lines responsed by WB are in online data
   [130]Figure 2SB. B. Densitometry analysis of H3K27me3 signals with
   ratio to H3 as shown in A, n = 3 experimental repeats. The results are
   expressed as mean ± SD from three independent experiments from these
   two IPF cell lines. ∗p < 0.05, compared with glutamine 2 mM or with
   added 2 mM α-KG. C. Representative merged images of immunofluorescence
   of IPF fibroblasts stained with antibodies against H3K27me3 (green) and
   DAPI (blue). Note, the number of cells seeded in the control medium was
   half compared to the number of cells seeded for the glutamine-free and
   glutamine 0 mM plus 2 mM α-KG conditions to avoid overcrowding of
   cells. The images were taken with a Keynence microscope, at 20×
   magnification. Additional images with DAPI or H3K27me3 alone are in
   online supplementary data ([131]Figure 3S). Right panel: densitometry
   of immunoflurescencent stain of green (H3K27me3) ratio to blue (DAPI).
   ∗p < 0.05, Gln 0 vs Gln 2 or α-KG. D and E. The IPF fibroblasts were
   collected after 48 h of each experimental condition. The association of
   H3K27me3 with Col3A1 (D) or PLK1 (E) was analyzed by ChIP assays. DNA
   was immunoprecipitated with specific antibody against H3K27me3. Primers
   for PCR are listed in [132]Table 2S. Negative control represents IgG
   pull-down. qPCR data were analyzed by using the 2-ΔΔCt method, and
   results were normalized to input DNA, expressed as fold changes
   relative to control samples (cells culture under glutamine 2 mM). The
   values are expressed as mean ± SD from average of three independent
   experiments of one representative cell line. ∗p < 0.05, compared to
   glutamine 2 mM (Gln2), #p < 0.05, compared to glutamine-free (Gln0).

   We then examined the changes of H3K27me3 at specific gene promoter
   regions. The association of H3K27me3 with Col3A1 and PLK1 promoter
   regions was examined by ChIP with primers listed in [133]Table 2S. We
   observed a significant increase in H3K27me3 association with both
   Col3A1 and PLK1 promoters under glutamine-free condition compared to
   control ([134]Figure 5D,E) likely underpinning reduced mRNA expression
   of Col3A1 and PLK1 under glutamine-free condition. With added α-KG, the
   association between the repressive mark H3K27me3 and Col3A1 promoter
   region is depleted, which likely drives the up-regulation of Col3A1
   expression at mRNA levels ([135]Figure 3C). However, the α-KG
   compensation did not reduce the association of H3K27me3 with Col3A1 to
   the level observed under 2 mM of glutamine, which indicates that other
   factors may be also involved in the regulation of Col3A1 expression by
   glutamine. For PLK1, the addition of α-KG did not significantly reverse
   the increased association between H3K27me3 and PLK1 promoter region
   under glutamine-free condition ([136]Figure 5E). This data indicates
   that the overall baseline levels of H3K27me3 are sensitive to changes
   in the concentration of glutamine and its metabolite α-KG. However, the
   association of H3K27me3 with specific gene promoter region is regulated
   differentially. In summary, H3K27me3 is enriched at the promoter region
   of Col3A1 and PLK1 in glutamine-free condition, likely repressing their
   expression; however, the addition of α-KG under glutamine-free
   condition only depleted H3K27me3 association with Col3A1 not that with
   PLK1.

3.7. GLS1^+/− heterozygous mice showed reduced fibrotic markers in response
to lung injury with bleomycin

   Previously, we have shown that inhibition of GLS1 attenuates lung
   fibrosis [[137]19]. Since GLS1^−/− null mice die shortly after birth
   [[138]37], we used a heterozygous (GLS1^+/−, Hets) strain to examine
   effects of reduced GLS1 expression on lung fibrosis and expression of
   collagen III and PLK1 in lung tissues. Studies showed that these
   GLS1^+/− mice demonstrated significant downregulation (up to 50%) of
   GLS1 enzymatic activity in the brain and kidney compared to their WT
   (GLS1^+/+) counterparts [[139]37,[140]38]. We subjected 8-week old
   wild-type (WT) and GLS1^+/− heterozygous mice to bleomycin-induced lung
   fibrosis. In this murine model of lung fibrosis, extracellular matrix
   accumulation has been shown to start between day 14 to day 20 after
   injury [[141]39], and almost reach the max of total increase of
   collagen at day 28 [[142]40]. We collected whole lung tissues at day 28
   to examine collagen III and PLK1 expression levels ([143]Figure 6A).
   GLS1^+/− heterozygous mice showed less lung fibrosis compared to WT
   mice in response to bleomycin injury by H&E stain ([144]Figure 6B), and
   by the semi-quantitative Ashcroft score ([145]Figure 6C). Whole lung
   lysates from GLS1^+/− heterozygous mice have reduced GLS1
   ([146]Figure 6D,E), and less collagen III compared to lung lysates from
   WT mice with bleomycin injury ([147]Figure 6D,F) as shown by
   western-blot analysis. Similarly, PLK1 is also reduced in GLS1 ^+/−
   heterozygous mice with bleomycin injury compared to WT mice
   ([148]Figure 6D,G). Taken together, our data showed that GLS1^+/−
   heterozygous mice showed reduced collagen III and PLK1 expression in
   the lung tissues of a murine model of lung fibrosis.

Figure 6.

   [149]Figure 6
   [150]Open in a new tab

   Reduced fibrotic markers in the lungs of bleomycin-injured GLS1+/−
   heterozygous mice. A. Schematic timeline of animal studies. 8-week old
   WT or GLS1+/− heterozygous mice were subjected to saline or bleomycin
   injury. All mice were sacrificed 28 days after injury, samples were
   collected for assays. B. H&E stain of WT or GLS1+/− hets mice lungs
   subjected to saline or bleomycin. C. Semiquantitative Ashcroft score of
   lung sections with WT or GLS1+/− heterozygous mice subjected to saline
   or bleomycin on day 28 after injury. ∗p < 0.05, WT saline vs Bleo, or
   WT vs Hets of bleomycin injured mice (n = 3). D. Mice lung tissues were
   collected and the expression of GLS1, Col3A1, and PLK1 were examined by
   western blots for each of the following group: WT saline (n = 4), WT
   with bleomycin injury (n = 5), and Hets with bleomycin injury (n = 6).
   β-actin is the loading control. E-G. Densitometry of GLS1 (E), Col3A1
   (F), and PLK1 (G) relative to β-actin (mean with standard error) as
   shown on the western blots in D. ∗p < 0.05, by t-test (WT saline vs
   bleo) or Kolmogorov–Smirnov test (WT bleo vs Hets bleo). H. Schematic
   diagram of possible mechanisms of the findings from this study.
   Glutamine control cell phenotype through an α-KG dependent and
   independent mechanisms to regulate gene expression.

4. Discussion

   Previously, our group demonstrated that glutaminolysis is required for
   fibroblast differentiation [[151]35,[152]41], and regulates fibroblast
   apoptosis [[153]17]. In our present study, we further characterized the
   cell phenotype changes under different metabolic conditions. We
   compared the differential expression of genes with and without
   glutamine, and examined if α-KG could compensate for glutamine-free
   condition. In this study ([154]Figure 6H), we found many genes are
   glutamine dependent, among those, some are α-KG dependent, like Col3A1,
   and some are α-KG independent, such as PLK1. We also explored if
   altered glutaminolysis modulating gene expression through association
   with histone modification H3K27me3. Our study confirmed that reduced
   glutaminolysis would lessen lung fibrosis. We provide the novel
   connection of metabolic alterations resulted gene expression change, at
   least partially, through histone modification in lung fibroblasts.

   Although glutamine is a non-essential amino acid, most mammalian cells
   still require at least millimolar levels of glutamine to survive and
   proliferate [[155]42]. Studies have shown those fast growing cells,
   like cancer cells, require more glutamine, and is currently considered
   as essential in cancer cells [[156]43]. Therefore, glutamine
   deprivation has been considered as a therapeutic approach in cancer.
   IPF is a disease that has many similarity to cancer [[157]44,[158]45],
   and we have shown that glutaminolysis is required for lung fibroblast
   differentiation and activation [[159]35], and reduced glutamine induces
   apoptosis in IPF fibroblasts [[160]17]. In the current study, we
   explored the effects of glutamine on IPF fibroblast proliferation,
   migration and matrix deposition. Glutamine is converted to glutamate
   and then to α-KG to enter tricarboxylic acid cycle [[161]43]. We
   thought adding α-KG in glutamine-free medium would compensate and
   recover many functions of the cells with glutamine withdrawal.
   Unexpectedly, adding α-KG to the glutamine-free medium did not improve
   proliferation nor migration in these IPF fibroblasts. This may due to
   the inability to recover certain important genes that are involved in
   such cellular functions, such as PLK1 that we examined closely. As the
   RNA-Seq data indicated, the top pathways affected by glutamine withdrew
   are cell cycle related. Adding α-KG only rescued few genes compared to
   the glutamine-free condition. Although the added α-KG did not improve
   the proliferation or migration in IPF lung fibroblasts, α-KG has many
   functions in cell growth [[162]15]. In a study with endothelial
   metabolism, the addition of α-KG rescued cellular growth in
   glutamine-depleted media [[163]46]. In another study with NIH3T3 and
   chondrocytes, the α-KG addition in standard culture media was reported
   to increase proliferation [[164]47], likely due to the increased
   extracellular matrix. Some of these observations are consistent with
   our observation of the increased collagen III in the cells compensated
   with α-KG. Nonetheless, our data indicate that only few changes caused
   by glutamine-free could be rescued by adding α-KG in IPF lung
   fibroblasts. The RNA-Seq data showed some genes are rescued, but fewer
   with significant cut off criteria. Of the three primary IPF fibroblast
   cell lines we examined, the gene expression has huge differences at
   baseline among these individuals, which are larger than the culture
   conditions. The heterogeneity and small size of the samples are the
   main limitation of this study. However, despite the heterogeneity of
   the different IPF cell lines, each cell line presented similar
   expression pattern under the glutamine-free and with added α-KG
   conditions.

   Ours and previous studies have demonstrated that glutamine is critical
   for profibrotic genes in IPF lung fibroblasts, withdrew glutamine would
   reduce such gene expression, such as Col3A1 in this study, Col1A1, and
   α-SMA reported before [[165]8,[166]35]. In a note, besides IPF cell
   lines, we also examined fibroblasts from IPF and non-IPF donors for
   these individual gene expression, which non-IPF showed similar pattern
   of changes as with IPF under these conditions. Among the altered gene
   expression due to glutamine withdrawal, there are α-KG-dependent, and
   -independent genes. We examined two genes in detail. Col3A1 is involved
   in IPF pathogenesis [[167]23], belongs to the collagen family, which
   were reported to be affected by different metabolic conditions
   [[168]48]. PLK1 is reported to be involved in IPF cell proliferation
   [[169]49], and is on top of the differential gene expression list from
   the RNA-Seq data. PLK1 remained down-regulated after adding α-KG
   supports the observation of no improved proliferation with added α-KG
   in glutamine-free medium. A prior IPF-related study showed reduced
   glutaminolysis down-regulates collagen in lung fibroblasts [[170]8]
   through α-KG mediated mTOR activation and proline hydroxylation at
   translational level. These previous studies demonstrated glutamine
   regulate collagen expression at the post-transcriptional level
   [[171]8,[172]21]. In addition, our study showed that glutamine and its
   metabolite α-KG also control Col3A1 at transcriptional level through
   histone H3K27me3 at its promoter region.

   With the established link that metabolites reprogram epigenetic status
   of the cells to regulate gene expression [[173]50,[174]51]; many
   metabolites are reported to be involved in histone modifications
   [[175]52,[176]53]. α-KG is involved in affecting H3K27me3 levels
   through regulating JMJC family histone demethylase [[177]13]. It is
   well established that histone modifications are critical for gene
   expression [[178]54]. Histone H3K27me3 is a repressive mark [[179]55].
   Its demethylases UTX and JMJD3 belongs to JMJC family which uses
   glutaminolysis metabolite α-KG as a cofactor [[180]13]. We have
   confirmed that the overall H3K27me3 levels are upregulated
   significantly with reduced glutaminolysis [[181]17]. For specific
   genes, the increased association of the repressive mark H3K27me3 with
   gene promoter region represses gene expression [[182]56]. We observed
   robust increased H3K27me3 association at the promoter region in both
   Col3A1 and PLK1 in response to glutamine deficiency, likely due to the
   decreased demethylase activity of the reduced glutamine metabolite
   [[183]17]. With the compensation of α-KG in glutamine-free media,
   reduced association of H3K27me3 at Col3A1 promoter region was noticed,
   probably due to the increased demethylase activity with added α-KG at
   the specific gene Col3A1. However, we did not observe such changes of
   H3K27me3 at PLK1 promoter region with added α-KG, which may indicate
   other α-KG-independent mechanisms are dominant in regulating PLK1 mRNA
   level expression under this condition.

   As we mentioned early, besides as a co-factor for histone related
   enzymes [[184]17], α-KG is also involved in regulation through mTOR
   activation at translational levels [[185]8]. Thus, changes in metabolic
   condition resulted differences are a combined effects to each specific
   gene and/or protein expression changes. We also have to point out the
   well-known fact of IPF fibroblast heterogeneity [[186]57], which would
   contribute to the different individual cell line reaction to the
   altered micro-environment. For example, some IPF cell lines have more
   robust changes than others in response to the changed culture
   conditions.

   In our previous study we showed reduced GLS1 in lung fibroblasts
   attenuates bleomycin-induced lung fibrosis [[187]19]. In this study, we
   used a GLS1^+/− heterozygous [[188]37,[189]38] animal model of lung
   fibrosis further confirmed that reduced GLS1 is protective for lung
   fibrosis. The whole lung tissue showed reduced Col3A1 and PLK1. Since
   the GLS1^+/− heterozygous mice have an overall reduced GLS1 expression,
   we would regard the reduced fibrosis as an overall effects. We realize
   that reduced GLS1 would have other effects than regulating H3K27me3, as
   we have demonstrated that some genes are not regulated through the
   GLS1-αKG-H3K27me3 axis. Therefore, we did not pursue to relate those
   gene expression changes with H3K27me3 levels in the whole lung tissues.
   Also, our previous study showed that ablation of GLS1 attenuates
   bleomycin induced lung fibrosis [[190]19], thus we did not further
   evaluate the hydroxyproline content in bleomycin injured GLS1^+/− hets
   mice, which would be the limitations of this study.

   Nonetheless, our current study confirmed that reduced glutaminolysis
   would alter cell phenotype, and α-KG only partially compensates for
   glutamine-free condition regarding cellular functions and gene
   expression. Among the genes that are glutamine dependent, some are α-KG
   dependent, like Col3A1, and some are α-KG independent, such as PLK1.
   The altered expression by glutamine deficient is likely, at least
   partially, through repressive histone mark H3K27me3. Our study supports
   that metabolic control of cellular function is likely through
   epigenetic regulation of histone modification to modulate gene
   expression.

5. Conclusions

   We found that glutaminolysis is important in maintaining IPF
   fibroblasts cell phenotype. Compensating the main metabolite, α-KG, to
   the glutamine-free medium has limited effects on the cell phenotype,
   such as cell proliferation and migration, but would compensate for
   collagen accumulation. RNA-Seq data indicate that only some of the
   differential expressed genes under glutamine-free condition are
   α-KG-dependent, while most are α-KG-independent. Col3A1 is an
   α-KG-dependent gene that is down-regulated at both RNA and protein
   levels, either by glutamine-free or siRNA GLS1 knockdown, which can be
   rescued by adding α-KG. While PLK1, one of the top hits of the cell
   cycles genes from RNA-seq data that are down-regulated in
   glutamine-free condition, its expression did not change with added
   α-KG. As α-KG is a cofactor for histone demethylase of H3K27me3, the
   associations of the repressive histone mark H3K27me3 with Col3A1 and
   PLK1 were depleted under glutamine-free condition. In an animal model
   of bleomycin induced lung fibrosis with reduced glutaminolysis of
   heterozygous GLS1 ^+/− mice, decreased fibrosis were observed with
   reduced fibrotic markers. This study indicates that glutamine is
   critical for supporting pro-fibrotic cell phenotype, partially through
   α-KG-depended and –independent mechanisms, and supports the new
   treatment approach of targeting metabolism for fibrotic diseases.

Funding sources

   This work was supported by the National Institutes of Health
   (R01AG050567 and R01HL151702 to YYS). The funders had no involvement in
   the study design, collector, analysis, interpretation of data or report
   writing.

Authorship contributions

   ZX, LB, JZ, KB: conducted research; ZX, LB, JZ, KB, GL, YYS: analyzed
   data; RC and JRS: analyzed RNA-Seq data; KB, YYS: prepared the
   manuscript; YYS: Conceptualized and designed the study. All authors
   have read and approved the final manuscript.

Footnotes

   ^Appendix A

   Supplementary data to this article can be found online at
   [191]https://doi.org/10.1016/j.molmet.2022.101655.

Conflict of interest

   None declared.

Appendix A. Supplementary data

   The following are the Supplementary data to this article:
   Multimedia component 1
   [192]mmc1.docx^ (2.1MB, docx)
   Multimedia component 2
   [193]mmc2.xlsx^ (17.9KB, xlsx)

Data availability

   Data will be made available on request.

References