Abstract

   Prolactinoma is the most prevalent pituitary neuroendocrine tumor and
   dopamine agonists (DAs) targeting dopamine D2 receptor (DRD2) are
   recommended as the first-line treatment. However, varying degrees of DA
   resistance limit patient benefit. Our study used transcriptome
   sequencing of surgical tumor samples and found abnormal cholesterol
   metabolism in prolactinoma, especially in DA-resistant tumors. We found
   that cholesterol significantly enhanced the resistance of prolactinoma
   MMQ cell lines to cabergoline in vitro and in vivo xenografts. Further,
   cholesterol did not affect the total protein level of DRD2, but changed
   the distribution of DRD2 with downregulation of its membrane abundance
   and upregulation of cytoplasmic localization. Mechanistically,
   immunoprecipitation combined with mass spectrometry revealed
   cholesterol increased binding affinity between DRD2 and stress granules
   (SGs) core proteins, such as G3BP1. Western blot experiment of G3BP1
   and fluorescent probe were used to confirm the formation of SGs after
   cholesterol treatment in MMQ cells and tumor xenografts, as well as in
   surgical tumor samples. Interfering the formation of SGs by
   overexpressing of USP10 and using the small molecule ISRIB reversed
   cholesterol’s effect on DRD2 cellular distribution and DA resistance in
   MMQ cells. Finally, a non-specificity inhibitor of SGs, anisomycin
   identified by drug repositioning analysis, could attenuate
   cholesterol-induced cabergoline resistance in vitro. Taken together,
   our findings suggest that abnormal cholesterol metabolism reduces DRD2
   membrane localization via stress granules formation, which may be an
   important reason for the DA resistance of prolactinoma patients.

Supplementary Information

   The online version contains supplementary material available at
   10.1186/s40478-025-01986-1.

   Keywords: Prolactinoma, Drug resistance, Cholesterol, Stress granules

Highlights

   Abnormal cholesterol metabolism has been identified in prolactinoma,
   especially in dopamine agonist-resistant tumors.

   Cholesterol promoted prolactinoma dopamine agonist resistance by
   downregulating the membrane localization of DRD2.

   Targeting of stress granules can attenuate cholesterol-regulated
   reduction of DRD2 membrane localization.

Supplementary Information

   The online version contains supplementary material available at
   10.1186/s40478-025-01986-1.

Introduction

   Pituitary neuroendocrine tumors (PitNETs) include nonfunctional and
   functional (hormone-secreting) types, with prolactinomas accounting for
   50% of all pituitary adenomas in both males and females according to
   the latest consensus [[40]1]. Prolactinomas can cause
   hyperprolactinemia and hypogonadism, as well as a greater risk of
   metabolic syndrome and obesity [[41]1]. The dopamine agonist
   cabergoline is the primary medical therapy for patients with
   microprolactinomas and well-encased macroprolactinomas [[42]1].
   Nevertheless, approximately 10% of prolactinoma patients are resistant
   to cabergoline at the recommended doses, and the resistance rate among
   patients with macroprolactinomas is 20% [[43]1–[44]4].

   Primary or secondary resistance to dopamine agonists in prolactinoma
   patients may involve multiple molecular mechanisms that are not yet
   fully understood; furthermore, the expression of dopamine receptor D2
   (DRD2) is a prerequisite for normal dopamine agonist function [[45]5].
   Previous studies have reported that molecules such as filamin A (FLNA)
   and β-arrestin2 are necessary for signal transduction of DRD2 activated
   by dopamine agonists [[46]5]. FLNA is required for DRD2 cell surface
   expression and prevents its lysosomal degradation, and drug-resistant
   prolactinomas exhibit lower FLNA expression [[47]6]. β-arrestin2
   supports DRD2 in reducing the phosphorylation of AKT, and its silencing
   weakens the ability of dopamine agonists to inhibit lactotroph
   proliferation [[48]7]. Currently, the posttranslational transport
   process of DRD2 remains unclear. Rab4 and Rab11 participate in the
   constitutive and ligand-dependent endosomal recycling of membrane DRD2,
   respectively [[49]8]. Furthermore, the regulatory network of DRD2
   cellular localization in prolactinomas requires further research.

   Previous work by our team revealed that patients with PitNETs were more
   susceptible to hypercholesterolemia [[50]9]. A recent study showed that
   the overexpression of patched domain-containing 1 (PTCHD1), which is a
   cholesterol transporter protein, could directly induce the formation of
   stress granules in the absence of other stressors [[51]10]. SGs are
   membraneless aggregates that are formed under stress conditions and
   include a variety of biomolecules, such as G3BP1, USP10, and Caprin-1
   [[52]11]. Caprin-1 promotes the formation of SGs, whereas USP10
   inhibits their formation [[53]11]. A high-cholesterol diet can induce
   the accumulation of SGs in macrophages, the liver and vascular smooth
   muscle cells in mice [[54]12, [55]13]. Moreover, hydroxycholesterol can
   induce the formation of SGs in human neuroblastoma, and saturated fatty
   acids can increase the content of G3BP1 in [56]P18000 pellets
   (containing SGs) during cell lysis [[57]14–[58]16]. SGs are reportedly
   involved in the drug resistance of various tumors, and the inhibition
   of SGs can overcome resistance to chemotherapy agents (such as
   5-fluorouracil) and targeted therapy (including sorafenib) [[59]17,
   [60]18]. However, the roles of cholesterol metabolism and SGs in the
   resistance of prolactinomas to dopamine agonists require further
   research.

   In this study, we performed transcriptome sequencing on peritumoral
   gland tissue and sensitive and resistant prolactinoma samples. The data
   revealed the presence of abnormal cholesterol metabolism in
   prolactinomas and its correlation with dopamine agonist resistance.
   Further cell and animal experiments revealed that cholesterol was able
   to reduce the membrane localization of DRD2 in MMQ cells, thus
   underscoring the importance of continued management of the
   hyperlipidemia that commonly accompanies prolactinomas [[61]19]. The
   interference of the formation of SGs can attenuate the effect of
   cholesterol on the membrane localization of DRD2. We recommend special
   consideration of the roles of abnormal cholesterol metabolism and SGs
   in dopamine agonist-resistant prolactinomas in the future.

Materials and methods

Cell culture and lentiviral transfection

   Rat prolactinoma cell lines MMQ (CRL-10609) and GH3 (CCL-82.1) were
   purchased from the American Type Culture Collection. Cells were
   cultured in F-12 K medium (PYG0036, Boster Biological Technology) with
   15% donor horse serum (04-124-1 A, Biological Industries), 2.5%
   certified fetal bovine serum (04-001-1 A, Biological Industries) and 1%
   penicillin-streptomycin solution (03-031-1B, Biological Industries).
   Cells were cultured in an incubator containing 5% CO2 and 95% air at
   37 °C.

   USP10 and control overexpression lentiviral (Genechem, Shanghai, China)
   were transfected into MMQ cells according to the manufacturer’s
   instructions. Specifically, we selected 10^^5 well-grown MMQ cells in a
   6-well plate, added lentivirus and pro-transfection agent polybrene
   (Genechem, Shanghai, China) with a Multiplicity of Infection (MOI) of
   100, and replaced fresh media containing puromycin after 72 h. Stable
   clones were obtained by using puromycin (2 µg/ml) to for 72 h.
   Transfection efficiency was determined by western blot experiments.

Patients and surgical samples

   Patients presented to the Multidisciplinary Center for Pituitary
   Adenomas in Xinqiao Hospital from January 2020 to December 2023 were
   considered. Responsiveness to dopamine agonists was determined by two
   experienced clinicians according to previously reported criteria
   [[62]20]. Resistance of prolactinomas was defined, according to the
   latest consensus, as the failure to normalize prolactin serum levels or
   achieve a significant reduction in tumor size (a decrease of ≥ 30% in
   the maximum diameter) when treated with standard doses of dopamine
   agonists (7.5–10 mg/day of bromocriptine or 2.0 mg/week of cabergoline)
   for a minimum duration of 6 months [[63]1]. The types of dopamine
   agonists utilized, doses and duration of treatment for 29 patients were
   provided in supplementary Table [64]e1. And the reasons for surgery in
   sensitive patients were provided in supplementary Table [65]2. To
   explore the correlation between serum lipid and drug responsiveness,
   patients formerly consumed lipid-lowering medications were excluded,
   and finally included 29 cases. The patient’s blood cholesterol and
   triglyceride level were reported by preoperative biochemical detection.
   The samples used for sequencing were obtained through transsphenoidal
   surgery, and 5 peritumoral gland tissue, 5sensitive and 4 resistant
   prolactinoma tissue were selected.

RNA sequencing

   The operation of RNA sequencing was carried out as we described before
   [[66]21]. In brief, total RNA of human prolactinoma tissue and
   peritumoral gland tissue was extracted by TRIzol (Invitrogen) according
   to manufacturer’s instructions. The mRNA after quality control was
   enriched by magnetic beads and then fragmented, and cDNA was obtained
   by reverse transcription for end repair. After amplification and
   quantification, the library was sequenced with Illumina Hiseq xten
   (2 × 150 bp read length). SeqPrep
   ([67]https://github.com/jstjohn/SeqPrep) and Sickle
   ([68]https://github.com/najoshi/sickle) were used for quality control
   of raw reads data. Then using
   HiSat2([69]http://ccb.jhu.edu/software/hisat2/index.shtml) for reads
   mapping [[70]22].

WGCNA analysis and identification of differentially expressed genes

   The “WGCNA’” R package was used to found key genes related to
   prolactinoma [[71]23]. After construction of sample clustering and
   co-expression network, the soft-threshold was set to further detect the
   module. Finally, the modules significantly related to traits were
   selected for analysis.

   Differentially expressed genes (DEGs) were analyzed by “edgeR” R
   package and defined as|log(Fold Change)| > 1.5 and p-value < 0.05
   [[72]24]. Then Toppgene website ([73]https://toppgene.cchmc.org/) was
   used for the gene ontology and pathway enrichment analyses of DGEs.
   Z-score of GO term and pathway term was automatically calculated by
   “GOplot” R package [[74]25]. The Z-score referred to here is not the
   conventional statistical measure but rather a metric calculated based
   on the fold change and the number of differentially expressed genes as
   described on ‘[75]https://wencke.github.io/’. This metric serves to
   indicate the likelihood of a biological process, molecular function, or
   cellular component being upregulated (positive value) or downregulated
   (negative value) [[76]25].

   Sequencing data of pituitary gland from female wild mice and pituitary
   hyperplasia (anterior pituitary were already enlarged) from female
   prolactin receptor deficient mice are publicly available under
   accession number [77]GSE100334 [[78]26].

Cell viability assay

   MMQ cells were seeded in 96-well plates at a density of 5000 cells per
   well and cultured with palmitic acid (HY-N0830, MedChemExpress),
   cabergoline (HY-15296, MedChemExpress) and cholesterol/Mβ-CD (C4951,
   Sigma Aldrich) for different times. The solution of palmitic acid was
   developed as described previously [[79]27]. DMSO (HY-Y0320,
   MedChemExpress) and Mβ-CD (C4555, Sigma Aldrich) were used as vehicle
   for cabergoline and cholesterol, respectively. After adding CCK-8
   solution (HY-K0301, MedChemExpress), the optical density at 450 nm
   (OD450) was measured after 1 h of culture in an incubator containing 5%
   CO2. Lipid depleted fetal bovine serum (C3840, Vivacell) was utilized
   instead of standard fetal bovine serum and donor horse serum to
   formulate the medium, thereby minimizing the impact of endogenous
   lipids. The kill effect of cabergoline was assessed by inhibition rate
   defined as (vehicle[OD450]-cabergoline[OD450])/vehicle[OD450] × 100%.

Apoptosis analysis

   An annexin V-FITC apoptosis detection kit (C1062S, Beyotime) was used
   to measure the apoptosis level of MMQ cells. In brief, MMQ cells were
   collected and mixed with 195ul binding buffer, 5ul Annexin-V-FITC and
   10 µl propidium iodide. After incubated at room temperature for 20 min,
   the MMQ cells were immediately analyzed using flow cytometry (Guava
   easyCyte, Cytek).

Xenograft experiments

   Three weeks old female BALB/cA-nu mice were purchased from Beijing
   Vital River Laboratory Animal Technology Co., Ltd. and fed of
   high-cholesterol diet (XT109C, Xietong Pharmaceutical Bio-engineering)
   or control diet (XT102C, Xietong Pharmaceutical Bio-engineering) for
   two weeks in SPF conditions. Then nude mice were subcutaneously
   inoculated with 2 × 10^6 MMQ cells and divided into four groups:
   high-fat diet with cabergoline administration, high-fat diet with
   vehicle administration, control diet with cabergoline administration,
   and control diet with vehicle administration. Cabergoline was
   administered at a dose of 0.5 mg/kg, once every two days, starting 8
   days after cell inoculation. Tumor volumes were calculated using the
   following formula: V (mm3) = [ab^2]/2 (a was the tumor maximum length
   and b was the minimum length). The tumors were resected and the mice
   heart blood was preserved for the measurement of triglycerides and
   cholesterol three weeks later. All animal experimental procedures were
   conducted under review and approval by the Internal Animal Care and Use
   Committee.

Reverse transcription and qPCR experiment

   Total RNA was extracted using RNA extraction kit (AG21022, Accurate
   Biotechnology). PrimeScript RT Reagent Kit (PR047A, TaKaRa) was used to
   obtain cDNA from 1 µg RNA. Then qPCR experiment was performed by using
   TB Green Premix Ex Taq II (RR820A, TaKaRa) and CFX96 Real-time System
   (Bio-Rad). The primers were described as follows, forward (Drd2):
   AGTTTCCCAGTGAACAGGCG, reverse (Drd2): CGTGCCCATTCTTCTCTGGT, forward
   (Actb): GAGGGAAATCGTGCGTGAC, reverse (Actb): GCATCGGAACCGCTCATT.

Western blot

   Total protein was obtained by using RIPA lysis buffer (WB001, Invent
   Biotechnologies) with protease inhibitor cocktail (GRF101, Epizyme
   Biotech). Plasma membrane and cytoplasmic proteins were isolated by
   using Minute™ plasma membrane protein isolation and cell fractionation
   kit (SM-005, Invent Biotechnologies). Protein was separated by TGX
   Stain-Free™ FastCast™ acrylamide kits (1610183, Bio-Rad) and
   transferred to PVDF membranes (Millipore, ISEQ00010). After blocking
   with 5% skim milk mixed in TBST solution, the membranes were incubated
   with primary antibodies listed as follows, DRD2 (55084-1-AP,
   Proteintech, 1:1000), G3BP1 (66486-1-Ig, Proteintech, 1:1000), USP10
   (19374-1-AP, Proteintech, 1:1000), Beta-Actin (81115-1-RR, Proteintech,
   1:4000). Then membranes were incubated with HRP-conjugated secondary
   antibodies (SA00001-2/SA00001-1, Proteintech, 1:2000) and detected with
   ECL substrate (BG0001, Bioground Biotech).

Immunofluorescence

   Before immunostaining, GH3 cells were cultured in 24-well plate coated
   with poly-L-lysine (ST509, Beyotime) and MMQ cells were cultured in
   24-well plate with Shi-fix™ coverslips (SB-Shifix-25, SHIKHAR BIOTECH).
   The treated cells were fixed with 4% paraformaldehyde and permeabilized
   with Saponin (P0095, Beyotime). Then the cells underwent sequential
   steps of blocking, primary antibody incubation, Alexa Fluor™
   594-conjugated secondary antibody (A-11012, Invitrogen) incubation, and
   finally mounting with an antifade mounting medium containing DAPI
   (P0131, Beyotime).

Immunohistochemistry

   Tumors implanted subcutaneously in nude mice were fixed in 4%
   paraformaldehyde and embedded with paraffin, and then sliced into 4 μm
   sections. The subsequent immunohistochemical staining was performed
   following the previously reported procedures [[80]9]. The intensity of
   immunostaining was measured by the IHC toolbox plug-in provided by Fiji
   [[81]28].

Stress granules detection

   Protein isolation of stress granules was consistent with previously
   reported procedures [[82]15]. The lysis buffer was composed of
   100mmol/L potassium acetate, 50mmol/L Tris-HCl with pH 7.4, 0.5mmol/L
   dithiothreitol, 2mmol/L magnesium acetate, 0.5% NP-40 solution,
   50 µg/ml heparin and protease inhibitor cocktail. Collected cells were
   lysed in lysis buffer, followed by centrifugation at 1000 g for 5 min,
   and the supernatant was collected as whole cell lysate (Wcl). After
   determining the concentration by using BCA protein assay kit (P0010,
   Beyotime), equal amounts of protein were further centrifuged at
   18,000 g for 20 min and the pellet was rich in stress granules
   ([83]P18000).

   The previously reported small molecule fluorescent probe (TASG) was
   used for fluorescence detection of stress granules [[84]29]. The TASG
   was quantified by Multi-point tool provided by Fiji to count the
   granules in each cell [[85]28].

Immunoprecipitation (IP)

   According to the manufacturer ‘s instructions, Magnetic Protein A/G
   IP/Co-IP Kit (YJ201, Epizyme Biotech) was used to explore the
   substrates binding with DRD2. In brief, MMQ cells were collected and
   washed with PBS, then incubated with lysis buffer provided in IP kit.
   The protein lysis after centrifugation was incubated with primary
   antibody against DRD2 overnight, then mixed with magnetic beads for 4 h
   at 4℃. Then the proteins were eluted from the magnetic beads for mass
   spectrometry sequencing.

Protein–protein interaction (PPI) network construction

   STRING database ([86]https://cn.string-db.org/) was used for the
   protein-protein interaction network construction and using Cytoscape
   (v3.9) for visualization. The importance of proteins in the network was
   ranked by the algorithm provided by cytoHubba plug-in of Cytoscape
   [[87]30].

GTEx database analysis

   The GEPIA2 web server ([88]http://gepia2.cancer-pku.cn/) was used to
   analysis the molecules expression correlation in normal pituitary
   tissue in GTEx database [[89]31].

Drug repositioning analysis

   The repositioning analysis of potential drugs for DA resistance
   prolactinoma was performed by R package “cogena” basing on the CMap
   database [[90]32].

Statistical analysis

   Data in this study was presented as the mean ± standard deviation (SD).
   The chi-squared test (one side) was used to determine the correlation
   between hypercholesterolemia and DA resistance. Significant differences
   between the two groups were ascertained by employing an unpaired t-test
   (two tailed) or, in case of unequal variances, a Mann-Whitney
   nonparametric test. Statistically significant was defined as a
   P-value < 0.05 (displayed as * P < 0.05, ** P < 0.01, *** P < 0.001,
   **** P < 0.0001).

Results

Transcriptome analysis revealed abnormal lipid metabolism in prolactinoma

   We conducted an analysis of the transcriptomic sequencing data of 14
   samples, including 5 peritumoral gland tissues and 9 prolactinoma
   tissues. A total of 1277 upregulated and 643 downregulated DEGs were
   identified and visualized via a volcano plot (Fig. [91]1A). A weighted
   gene coexpression network was constructed (Supplementary Fig. 1A), and
   a heatmap of module-trait relationships (Fig. [92]1B) in peritumoral
   gland tissue and prolactinoma was constructed based on the Pearson
   correlation coefficient. The significantly related gene modules,
   including the blue module (r = 0.7, p = 0.005), magenta module
   (r = 0.64, p = 0.01) and turquoise module (r = 0.76, p = 0.002), were
   selected together with the DEGs to construct a Venn diagram. Finally,
   862 overlapping genes were defined as core DEGs between prolactinoma
   and peritumoral gland tissue (Fig. [93]1C). We further performed
   functional annotation analysis on the core DEGs (Fig. [94]1D). Multiple
   extracellular matrix-related terms, such as extracellular matrix
   structural constituent (term ID: GO0005201, Z-score =-2.710687) were
   repressed. Pathway enrichment analysis of the core DEGs revealed
   several lipid metabolism pathways (Fig. [95]1E). We subsequently
   analyzed public pituitary hyperplasia (the anterior pituitary already
   being enlarged) sequencing data from wild-type (WT) and prolactin
   receptor (Prlr) knockout (KO) female mice at different ages (with this
   knockout potentially leading to prolactinoma), and found that the DEGs
   observed between Prlr-KO and WT mice changed over time in relation to
   lipid metabolism (Fig. [96]1F) [[97]26]. These results indicate that
   abnormal lipid metabolism is involved in the tumorigenesis of
   prolactinoma.

Fig. 1.

   [98]Fig. 1
   [99]Open in a new tab

   Transcriptome analysis revealed abnormal lipid metabolism in
   prolactinoma. (A) Volcano plot of DEGs between peritumor gland tissue
   and prolactinoma samples. (B) Heatmap of correlation between gene
   modules and clinical traits in peritumor gland tissue and prolactinoma
   samples. (C) Venn analysis of gene modules and DEGs to identify core
   DEGs between prolactinoma and peritumor gland tissue. (D) GO function
   analysis of core DEGs between prolactinoma and peritumor gland tissue.
   (E) Pathway enrichment of core DEGs between prolactinoma and peritumor
   gland tissue. (F) DEGs of public pituitary hyperplasia (anterior
   pituitary were already enlarged) sequencing data between prolactin
   receptor knockout and wild-type mice with different months, and their
   pathway enrichment. (G) Heatmap of DEGs between DA resistant and
   sensitive prolactinoma samples. (H) Pathway enrichment analysis of
   resistance related DEGs. (I) The intersected genes between resistance
   related DEGs and genes in the term metabolism of lipids in Figure F.
   (J) Clinical features of 14 DA sensitive and 15 DA resistant
   prolactinoma patients. (K-L) Blood triglyceride (K) and total
   cholesterol (L) levels of sensitive and resistant prolactinoma
   patients. * (P) < 0.05

   To further investigate the molecular mechanism of DA resistance, we
   identified 341 upregulated and 383 downregulated DEGs between 5
   DA-sensitive prolactinomas and 4 DA-resistant prolactinomas
   (Fig. [100]1G). Functional annotation analysis of these
   resistance-related DEGs revealed lipid metabolism pathways
   (Fig. [101]1H). We extracted the genes related to lipid metabolism, as
   shown in Figure F, and intersected these genes with the drug
   resistance-related DEGs. Five overlapping genes (UGT8, ACACB, SLC44A3,
   PLIN2 and PON1), which are predominantly involved in cholesterol and
   fatty acid metabolism, were obtained (Fig. [102]1I) [[103]33–[104]35].
   Therefore, we further reviewed the clinical features of 14 DA-sensitive
   prolactinoma patients and 15 DA-resistant prolactinoma patients. We
   found that DA-resistant prolactinoma patients were more likely to
   exhibit elevated total cholesterol levels (p = 0.0384) (Fig. [105]1L),
   whereas triglyceride levels were not significantly different between
   the two groups (Fig. [106]1K). There was no significant difference in
   BMI between the two groups (Fig. [107]1J).

Cholesterol promoted MMQ cell resistance to Cabergoline treatment

   We subsequently investigated whether cholesterol and fatty acids
   influence the effect of cabergoline on MMQ cells in vitro. We selected
   palmitic acid for analysis, which is the most abundant saturated fatty
   acid in human serum, and the CCK-8 results revealed that exogenous
   addition of palmitic acid did not alleviate the degree of
   cabergoline-induced inhibition of MMQ cells (Supplementary Fig. 1B).
   However, the presence of 20 µg/ml Mβ-CD balanced cholesterol
   (CHO/Mβ-CD) significantly mitigated the inhibitory effect of
   cabergoline on MMQ cells without impacting cell viability
   (Fig. [108]2A). The results of flow cytometry revealed that 20 µg/ml
   CHO/Mβ-CD reduced the apoptosis of MMQ cells induced by cabergoline
   treatment for 48 h (Fig. [109]2B-C).

Fig. 2.

   [110]Fig. 2
   [111]Open in a new tab

   Cholesterol promoted the resistance of MMQ cells to DAs. (A) Cell
   viability assays of MMQ cells after treatment with CHO/Mβ-CD and
   cabergoline alone or in combination. (B)-(C) Flow cytometry analysis of
   MMQ cell apoptosis after treatment with CHO/Mβ-CD and cabergoline alone
   or in combination. (D)-(E) Detection of cholesterol and triglyceride
   levels in the peripheral blood of nude mice after they were fed a
   high-fat diet or control diet for six weeks. (F) Weights of the nude
   mice prior to tumor resection. (G) Images of tumor xenografts in nude
   mice after being exposed to different diets and drugs (H-(I). Weights
   and volumes of the tumor xenografts. * P < 0.05, *** P < 0.001, ****
   P < 0.0001

   In vivo experiments revealed that a high-cholesterol diet for six weeks
   resulted in elevated levels of cholesterol and triglycerides in nude
   mice (Fig. [112]2D-E), and the weights of the nude mice did not
   significantly change during the course of the experiment
   (Fig. [113]2F). In the absence of cabergoline, a high-fat diet did not
   affect the weights or volumes of tumor xenografts compared to a control
   diet. However, when cabergoline was administered, the high-fat diet
   attenuated the ability of cabergoline to reduce the weights and volumes
   of tumor xenografts (Fig. [114]2G-I). These results suggest that
   abnormal cholesterol metabolism may be involved in the process of DA
   resistance in prolactinoma.

DRD2 membrane localization was decreased by cholesterol

   To explore the potential mechanism of cholesterol-induced drug
   resistance in prolactinoma, we performed transcriptomic sequencing on
   CHO/Mβ-CD- or Mβ-CD-treated MMQ cells. The results revealed that
   CHO/Mβ-CD treatment decreased the expression of genes involved in
   cholesterol synthesis, such as Hmgcr, Sqle and Hmgcs1 (Fig. [115]3A).
   Moreover, the expression of cholesterol efflux-related genes and
   sterol-responsive transcription factors, such as Abca1 and Srebf1, was
   increased (Fig. [116]3A). These results indicated that the MMQ cells
   did response to cholesterol stimulation. Functional analysis revealed
   that cholesterol synthesis and the GPCR signaling pathway were
   downregulated (Fig. [117]3B), whereas the nuclear receptor signaling
   pathway was activated (Fig. [118]3C). Given that the target of
   cabergoline (the DRD2 receptor) belongs to the GPCR family, we further
   focused on potential alterations in DRD2 caused by cholesterol. The
   sequencing results revealed that the transcription of long subtype of
   Drd2 was significantly increased after CHO/Mβ-CD treatment, which was
   also confirmed via qPCR experiments (Supplementary Fig. 1C). However,
   the total cellular protein expression level of DRD2 was not
   significantly altered (Fig. [119]3D). DRD2 receptor availability at the
   cell surface has been proposed as being a molecular marker of
   prolactinoma pharmacological responsiveness in the absence of genetic
   defects [[120]6]. Therefore, we further explored whether the spatial
   distribution of DRD2 was changed due to CHO/Mβ-CD treatment. The
   results demonstrated that CHO/Mβ-CD treatment reduced the protein
   abundance of DRD2 on the cell membrane and increased its content in the
   cytoplasm (Fig. [121]3E-F). The immunofluorescence results also
   revealed decreased membrane localization and increased cytoplasmic
   localization of DRD2 in CHO/Mβ-CD-treated MMQ and GH3 cells
   (Fig. [122]3G-H). Finally, we isolated the cell membrane proteins from
   the xenografts and found that the membrane localization of DRD2 was
   lower in the high-fat diet group compared to the control diet group
   (Fig. [123]3I).

Fig. 3.

   [124]Fig. 3
   [125]Open in a new tab

   Cholesterol treatment reduced the membrane localization of DRD2. (A)
   Volcano plot of DEGs sequenced from MMQ cells after CHO/Mβ-CD or Mβ-CD
   treatment. (B-C) Pathway enrichment results of downregulated and
   upregulated DEGs. (D) Total DRD2 protein levels in MMQ cells after
   CHO/Mβ-CD or Mβ-CD treatment. (E-F) Membrane and cytoplasmic protein
   levels of DRD2 in MMQ cells after CHO/Mβ-CD or Mβ-CD treatment. (G-H)
   Immunofluorescence results of DRD2 in MMQ and GH3 cells after CHO/Mβ-CD
   or Mβ-CD treatment. (I) The expression level of DRD2 in the cell
   membranes of xenografts. Scale bar: 5 μm. Western Blot experiments had
   three biological replicates and two technical replicates

Cholesterol promoted the formation of stress granules

   We performed immunoprecipitation and mass spectrometry sequencing of
   endogenous DRD2 in MMQ cells to investigate the potential mechanism of
   decreased DRD2 abundance on the cell membrane. The binding affinity of
   several proteins (such as UBAP2L, G3BP1 and RAB7A) for DRD2 was altered
   following CHO/Mβ-CD treatment (Fig. [126]4A). Functional annotation
   analysis revealed that proteins with decreased binding affinity were
   mainly involved in protein localization to the membrane (Fig. [127]4B),
   whereas proteins with increased binding affinity were mainly related to
   stress granule assembly (Fig. [128]4C). Due to the fact that stress
   granules are dynamic assemblies of RNA and proteins that form in
   response to various cellular stresses, we focused on proteins with
   increased binding affinity after cholesterol treatment. The key marker
   of stress granule formation (G3BP1) was identified as the predominant
   protein among the increased-binding proteins via the construction of a
   protein interaction network (Fig. [129]4D). We observed a significant
   negative correlation between G3BP1 and DRD2 expression in the pituitary
   gland according to the GTEx database (Fig. [130]4E).

Fig. 4.

   [131]Fig. 4
   [132]Open in a new tab

   Cholesterol promoted the formation of stress granules. (A) Plot of
   proteins identified via immunoprecipitation combined with mass
   spectrometry sequencing. (B-C) Functional annotation analysis of
   proteins with decreased or increased binding affinity to DRD2 after
   CHO/Mβ-CD treatment in MMQ cells. (D) Protein interaction network of
   proteins with increased binding affinity to DRD2 and the predominant
   ranks. (E) Correlation plot of G3BP1 and DRD2 expression in the
   pituitary gland according to the GTEx database. (F) Protein level of
   G3BP1 in whole-cell lysates (wcl) and [133]P18000 pellets from MMQ
   cells. (G) Fluorescence staining of stress granules using TASG in MMQ
   cells. (H) Immunohistochemical staining of G3BP1 in tumor xenografts.
   (I) Western blot analysis of G3BP1 in tumor xenografts. (J)
   Immunohistochemical staining of G3BP1 in sensitive and resistant human
   prolactinomas. Scale bar: 5 μm

   We further examined the formation of stress granules after CHO/Mβ-CD
   treatment. The protein abundance of G3BP1 significantly increased in
   the [134]P18000 protein, which represents the abundant proteins of
   stress granules (Fig. [135]4F) [[136]15]. In addition, the staining of
   stress granules with the fluorescent probe TASG revealed significant
   accumulation of stress granules after CHO/Mβ-CD treatment (p = 0.0193)
   (Fig. [137]4G). The immunohistochemistry results revealed that the
   expression of G3BP1 was greater in the xenografts of the high-fat diet
   compared to the control diet group (Fig. [138]4H). Western blot
   analysis revealed similar results (Fig. [139]4I). Finally, the
   immunohistochemistry results revealed increased expression of G3BP1 in
   resistant human prolactinomas compared with sensitive prolactinomas
   (Fig. [140]4J).

Targeting of stress granules could attenuate the cholesterol-mediated
reduction in DRD2 membrane localization

   Subsequently, we explored whether stress granules mediate the
   regulatory effect of cholesterol on DRD2 membrane localization. We
   overexpressed USP10 in MMQ cells, and the Western blot results revealed
   that the total protein level of DRD2 was similar between USP10^Mock and
   USP10^OE MMQ cells (Fig. [141]5B). Compared to the USP10^Mock group,
   the assembly of stress granules in the USP10^OE group treated with
   CHO/Mβ-CD was decreased, as shown via fluorescence staining of TASG
   (Fig. [142]5C). The membrane abundance of DRD2 was decreased after
   CHO/Mβ-CD treatment in USP10^Mock MMQ cells but was not significantly
   altered in USP10^OE MMQ cells (Fig. [143]5D-E). Immunofluorescence
   staining revealed decreased membrane localization of DRD2 in
   CHO/Mβ-CD-treated USP10^Mock MMQ cells and constant membrane
   localization in USP10^OE MMQ cells (Fig. [144]5F-G). The results of the
   cell viability assay revealed that the inhibition rate of cabergoline
   was attenuated upon the addition of CHO/Mβ-CD to USP10^Mock cells,
   whereas cholesterol had a diminished effect on the inhibition rate of
   cabergoline in USP10^OE MMQ cells (Fig. [145]5H-I). Furthermore, we
   used ISRIB (Integrated Stress Response Inhibitor) to suppress the
   formation of stress granules, as previously reported [[146]36,
   [147]37]. We found that 1 µM ISRIB effectively inhibited the induction
   of stress granules by cholesterol (Fig. [148]5J). Western blot
   experiments revealed that the membrane abundance of DRD2 was decreased
   upon the addition of ISRIB after CHO/Mβ-CD treatment (Fig. [149]5K).
   Cell proliferation assays revealed that ISRIB treatment suppressed the
   ability of cholesterol to reduce cabergoline-induced MMQ-induced cell
   death (Fig. [150]5L).

Fig. 5.

   [151]Fig. 5
   [152]Open in a new tab

   Stress granules mediated the cholesterol-regulated reduction in DRD2
   localization on the cell membrane. (A) USP10 can inhibit the assembly
   of stress granules, according to a previous study. (B) Overexpression
   of USP10 in MMQ cells and its effect on DRD2 protein levels. (C)
   Fluorescence staining of stress granules with TASG in USP10^Mock and
   USP10^OE MMQ cells. (D-E) Western blot analysis of the cell membrane
   proteins of USP10^Mock and USP10^OE MMQ cells. (F-G) Fluorescence
   staining of DRD2 in USP10^Mock and USP10^OE MMQ cells. (H-I) CCK-8
   results of USP10^Mock and USP10^OE MMQ cells treated with CHO/Mβ-CD and
   cabergoline alone or in combination. (J) Fluorescence staining of
   stress granules with TASG in MMQ cells. (K) Western blot analysis of
   the levels of cell membrane proteins in MMQ cells. (L) CCK-8 results
   for MMQ-treated cells. *** P < 0.001, **** P < 0.0001. Scale bar: 5 μm

   Finally, we conducted a drug repositioning analysis to identify
   pharmaceuticals that are capable of targeting stress granules to
   overcome cabergoline resistance. The DEGs observed between sensitive
   and resistant prolactinoma samples were divided into 3 coexpressed gene
   sets by using the “NbClust” R package (Supplementary Fig. 1D-E). We
   mapped the gene cluster onto the Cmap database and identified the
   natural compound anisomycin, which has been reported to possess the
   ability to inhibit stress granule assembly (Supplementary Fig. 1F)
   [[153]38, [154]39]. Fluorescence staining with TASG revealed that 2 µM
   anisomycin was able to inhibit the stress granule assembly after
   treatment with CHO/Mβ-CD (Supplementary Fig. 1G). A cell viability
   assay revealed that CHO/Mβ-CD had a weaker effect on the inhibition
   rate of cabergoline in MMQ cells after the addition of 2 µM anisomycin
   (Supplementary Fig. 1H-I).

Discussion

   Cabergoline, which is a preferred dopamine agonist, has high efficacy
   in decreasing serum prolactin levels, ameliorating clinical
   manifestations and reducing prolactinoma size [[155]1]. However,
   approximately 10% of prolactinoma patients develop resistance, and the
   underlying molecular mechanism of this resistance is not currently
   known [[156]1–[157]4]. This study identified dysregulated cholesterol
   metabolism in prolactinoma, especially in dopamine agonist-resistant
   tumors, by transcriptomic sequencing analysis of peritumoral gland
   tissue and sensitive and resistant prolactinoma samples. Further in
   vitro and in vivo experiments revealed that cholesterol promoted MMQ
   cell resistance to cabergoline treatment by reducing the membrane
   localization of DRD2. The interference of the formation of stress
   granules was able to attenuate the effect of cholesterol on the
   membrane localization of DRD2.

   Abnormal lipid metabolism has been implicated in oncogenesis, as well
   as in conferring resistance to therapeutic drugs. Clinical data
   revealed that patients with resistant prolactinomas were more likely to
   exhibit elevated blood cholesterol. In particular, 5 lipid
   metabolism-related genes (UGT8, ACACB, SLC44A3, PLIN2 and PON1) were
   identified as DEGs between resistant and sensitive prolactinomas.
   UGT8-mediated sulfatide synthesis can regulate BAX localization and
   influence the sensitivity of colorectal cancer cells to apoptosis
   [[158]40]. ACACB-encoded acetyl-CoA carboxylase elicits a shift from
   glycolysis-dependent to lipogenesis-dependent activities, thereby
   endowing head and neck squamous cell carcinoma with resistance to
   cetuximab [[159]41]. The lipid droplet-coating protein PLIN2 mediates
   the accumulation of lipid droplets, which act as functional acentriolar
   microtubule organizing centers, thereby interfering with mitotic
   spindle polarity and endowing tumors with resistance to paclitaxel
   [[160]42]. The stability of PON1 has been shown to support the
   expansion of resistant clones in a tumor regression model [[161]43]. In
   vivo and in vitro experiments revealed that cholesterol promoted MMQ
   resistance in cells without affecting their proliferation ability.
   However, palmitic acid, which is the most abundant saturated fatty acid
   in the human serum, did not influence the resistance of MMQ cells, thus
   indicating specific underlying resistance effects among lipid
   metabolites in MMQ cells.

   The first and most obvious explanation for dopamine agonist resistance
   in prolactinomas is the loss of receptor expression on the tumor cell
   surface, and many studies have attempted to validate this hypothesis
   [[162]6]. The results of sequencing and qPCR experiments revealed
   increased transcription of total Drd2 after cholesterol treatment,
   whereas the total protein level of DRD2 remained unchanged, as
   confirmed via Western blotting. Due to the fact that the GPCR signaling
   pathway was downregulated according to the functional analysis results,
   we further focused on the protein function of DRD2. Interestingly, we
   found that cholesterol was able to reduce the membrane localization of
   DRD2, which may explain the acquired resistance of MMQ cells. This
   finding expands the comprehension of the role of cholesterol in
   prolactinoma resistance and indicates that cholesterol metabolism may
   influence the surface abundance of other drug receptors, such as EGFR
   and PD-L1 [[163]44, [164]45].

   Currently, the regulatory mechanism of DRD2 cellular localization in
   prolactinomas is poorly understood. Previous study reported that
   leucine-rich repeat kinase-2 (LRRK2) and tetraspanin-7 (TSPAN7)
   regulated the transport and membrane localization of DRD2 [[165]46,
   [166]47]. G protein-coupled receptor kinase-2 (GRK2), β-arrestins and
   FLNA reportedly control the internalization and recycling of membrane
   DRD2 [[167]6, [168]48]. We performed immunoprecipitation to explore the
   potential mechanism of cholesterol-induced changes in DRD2 membrane
   localization, and functional annotation analysis showed that these
   proteins with increased binding affinity to DRD2 were mainly related to
   stress granules. We subsequently overexpressed USP10 in MMQ cells
   (which can inhibit SG formation) and found that the membrane abundance
   of DRD2 remained unchanged in USP10^OE MMQ cells after cholesterol
   treatment [[169]11]. Similar results were obtained via the use of
   ISRIB, which inhibits SG formation. In addition, anisomycin, a
   non-specificity inhibitor of SGs, was identified through drug
   reposition analysis, and showed attenuate effect on MMQ resistance
   induced by cholesterol in vitro. These findings indicate a potential
   association between SGs and DRD2 membrane localization, and USP10 may
   serve as a tool for modulating DRD2 membrane localization and enhancing
   prolactinoma sensitivity to cabergoline. This is not only significant
   for treating prolactinoma drug resistance, but also potentially
   provides insights for diseases such as Parkinson’s disease dopamine
   resistance.

   Some limitations remain to be addressed from this study. First, the
   manner in which SGs regulate DRD2 membrane localization is still
   unclear, and it is likely that this relationship is not
   straightforward. Another concern is whether the formation of SGs
   induced by other stressors can lead to resistance to cabergoline in
   prolactinomas.

   In summary, our results indicate that abnormal cholesterol metabolism
   plays an important role in the resistance of prolactinoma to DAs.
   Cholesterol was able to reduce the membrane localization of DRD2 in MMQ
   cells, and interference with the formation of SGs can attenuate the
   effect of cholesterol on the membrane localization of DRD2. Future
   research and clinical practice may consider the role of stress granules
   in resistant prolactinomas.

Conclusions

   Our findings revealed dysregulated cholesterol metabolism in
   prolactinoma, especially in dopamine agonist-resistant tumors.
   Cholesterol can reduce the membrane localization of DRD2, which may be
   an important reason for the resistance of prolactinomas to DAs.

Electronic supplementary material

   Below is the link to the electronic supplementary material.
   [170]Supplementary Material 1^ (4.4MB, tif)
   [171]Supplementary Material 2^ (2.9MB, tif)
   [172]Supplementary Material 3^ (370.9KB, docx)
   [173]Supplementary Material 4^ (84MB, zip)

Abbreviations

   DRD2
          Dopamine receptor D2

   DA
          Dopamine agonist

   G3BP1
          G3BP stress granule assembly factor 1

   PitNETs
          Pituitary neuroendocrine tumors

   GRK2
          G protein-coupled receptor kinase 2

   DEGs
          Differentially expressed genes

   Wcl
          Whole cell lysate

   IP
          Immunoprecipitation

   PPI
          Protein–protein interaction

   Mβ-CD
          Mβ-CD balanced cholesterol

   USP10
          Ubiquitin specific peptidase 10

   PD-L1
          Programmed-death ligand 1

Author contributions

   Hui Yang and Song Li designed this work. Yuyang Peng and Chengcheng
   Wang carried out experiments. Yuyang Peng and Huachun Yin carried out
   data analysis. Xiangdong Pei and Yuan Zhang conducted clinical data
   collection. Yuyang Peng written the manuscript. Hong Liang provided
   suggestions. Xin Zheng, Song Li and Hui Yang revised the manuscript.

Funding

   This work was supported by Natural Science Foundation of Chongqing
   (CSTB2024NSCQ-MSX0021, CSTB2022NSCQ-MSX1534) and Young Doctor
   Incubation Program of Xinqiao Hospital (2022YQB046, 2023YQB0384,
   2023YQB055).

Data availability

   The data covered in this article are available on reasonable request
   from the corresponding author.

Declarations

Ethical approval

   This study was conducted in according to the principles of the Helsinki
   declaration and approval of the Ethics Committee of Xinqiao Hospital,
   Army Medical University (authorization ID: 2022-356-01). All animal
   experimental procedures were conducted under review and approval by the
   Internal Animal Care and Use Committee of the Army Medical University
   (authorization ID: AMUWEC20224532).

Competing interests

   The authors declare no competing interests.

Footnotes

   Publisher’s note

   Springer Nature remains neutral with regard to jurisdictional claims in
   published maps and institutional affiliations.

   Yuyang Peng, Huachun Yin, Xiangdong Pei and Yuan Zhang contributed
   equally to this work.

Contributor Information

   Hui Yang, Email: huiyangtg2018@aliyun.com.

   Song Li, Email: tmmuls@tmmu.edu.cn.

References