Abstract

Background

   Lemur Tail Kinase 3 (LMTK3) promotes cell proliferation, invasiveness
   and therapy resistance, and its expression correlates with poor
   survival in several different malignancies, including breast cancer.
   Crosstalk through extracellular vesicles (EVs) is an increasingly
   appreciated mechanism of cell communication within the tumour immune
   microenvironment, which contributes to different aspects of cancer
   progression and plays a pivotal role in shaping tumour fate.

Methods

   Nanoparticle tracking analysis and transmission electron microscopy
   were used to study the effects of LMTK3 on EV size, while single
   particle interferometry allowed us to examine LMTK3-dependent effects
   on the subpopulation distribution of EVs. Quantitative mass
   spectrometry was used to profile LMTK3-dependent proteomics changes in
   breast cancer-derived EVs. Bioinformatics analysis of clinical data
   along with in vitro and cell-based assays were implemented to explore
   the effects of LMTK3-dependent EV protein cargo on the tumour immune
   microenvironment. To elucidate the mechanism through which LMTK3
   impacts endosomal trafficking and regulates EV biogenesis, we used a
   variety of approaches, including in vitro kinase assays, confocal and
   electron microscopy, as well as in vivo subcutaneous and
   orthotopic breast cancer mouse models.

Results

   Here, we report that LMTK3 increases the average size of EVs, modulates
   immunoregulatory EV proteomic cargo and alters the subpopulation
   distribution of EVs released by breast cancer cells. Mechanistically,
   we provide evidence that LMTK3 phosphorylates Rab7, a key regulator of
   multivesicular body (MVB) trafficking, thereby reducing the fusion of
   MVBs with lysosomes and subsequent degradation of intralumenal
   vesicles, resulting in altered EV release. Moreover, LMTK3 causes
   increased packaging of phosphoserine aminotransferase 1 (PSAT1) in EVs,
   leading to a paracrine upregulation of phosphoglycerate dehydrogenase
   (PHGDH) in monocytes when these EVs are taken up. PSAT1 and PHGDH play
   key roles in the serine biosynthesis pathway, which is closely linked
   to cancer progression and regulation of monocyte behaviour. LMTK3
   EV-induced elevated PHGDH expression in monocytes reduces their
   infiltration into breast cancer 3D spheroids and in vivo breast
   cancer mouse models. Furthermore, these infiltrating monocytes
   preferentially differentiate into pro-tumourigenic M2-like macrophages.
   Additional breast cancer mouse studies highlight the contribution of
   LMTK3-dependent EVs in the observed immunosuppressive macrophage
   phenotype. Finally, in vitro experiments show that
   pharmacological inhibition of LMTK3 reverses the pro-tumourigenic and
   immunomodulatory effects mediated by EVs derived from LMTK3
   overexpressing cells.

Conclusion

   Overall, this study advances our knowledge on the mechanisms of EV
   biogenesis and highlights a novel oncogenic role of LMTK3 in the breast
   TME, further supporting it as a target for cancer therapy.

Supplementary Information

   The online version contains supplementary material available at
   10.1186/s12943-025-02346-2.

   Keywords: LMTK3, Rab7, Extracellular Vesicles, Breast Cancer,
   Monocytes, Macrophages, Tumour Microenvironment

Introduction

   Lemur tail kinase 3 (LMTK3) is a serine/threonine protein kinase with
   critical roles in different malignancies, including breast cancer (BC)
   where its upregulation drives key cellular and oncogenic processes,
   such as estrogen receptor α (ERα) transcription and stabilisation,
   chromatin remodelling, resistance to chemo- and endocrine- therapies,
   invasion and metastasis [[46]1–[47]7]. Through its regulation of these
   oncogenic processes, LMTK3 upregulation in BC strongly correlates with
   poorer survival, higher tumour grade and more advanced disease,
   highlighting its association with more aggressive cancers [[48]8].
   Consequently, it is not surprising that LMTK3 has emerged as a
   promising target for cancer therapy.

   LMTK3 belongs to the LMTK family of kinases. Interestingly, both LMTK1
   and LMTK2 isoforms have been implicated in endosomal trafficking
   [[49]9], while depletion of LMTK2 can lead to impaired recycling of the
   transferrin receptor and result in swollen endosomes [[50]10].
   Endosomal trafficking is controlled by Rab proteins, a large family of
   small GTPases regulating the movement of intracellular cargo between
   organelles [[51]11]. Amongst them, Rab7a is particularly important for
   the maturation of early to late endosomes, which is crucial for the
   sorting and degradation of cellular cargo and the formation of
   multivesicular bodies (MVBs) [[52]12, [53]13]. MVBs are important
   structures in extracellular vesicle (EV) biogenesis, as the membrane of
   the late endosomes can undergo inward budding forming intraluminal
   vesicles (ILVs). The MVBs can then fuse with the plasma membrane,
   releasing the ILVs as exosomes, a subtype of EV specifically arising
   through the endosomal pathway [[54]14]. The regulation of endosomal
   trafficking by LMTKs therefore suggests the kinases may have a role in
   the regulation of EV biogenesis. EVs are lipid bilayer-delimited
   particles naturally released from cells and exosomes are a subtype of
   EV of endosomal origin. In disease, EVs have been proposed as useful
   candidates for biomarkers, as well as promising drug delivery systems
   for therapeutic agents. However, interestingly, EVs also mediate
   intercellular communication by carrying proteins, nucleic acids, and
   other bioactive cargo, contributing to numerous physiological
   processes. In particular, EVs have gained significant recognition for
   their roles in immunity, inflammation and cancer progression through
   facilitating communication between cancer cells and immune cells
   [[55]15].

   Monocytes are short-lived mononuclear myeloid lineage cells that
   circulate in blood and have different subtypes (classical,
   non-classical and intermediate) that are involved in phagocytosis,
   migration, innate immune responses, cytokine secretion and antigen
   presentation [[56]16, [57]17]. In BC, monocyte infiltration and
   expansion in the tumour microenvironment (TME) correlates with disease
   stage and tumour size [[58]18–[59]21]. After their recruitment into
   tumours, monocytes can differentiate into a heterogeneous population of
   macrophages, which, although oversimplified, can be broadly categorised
   as M1-like or M2-like subsets [[60]22]. Pro-inflammatory
   (anti-tumourigenic) M1-like macrophages are key to immune response
   against bacteria and promote T helper 1 (Th1) T cell activity, while
   M2-like anti-inflammatory (pro-tumourigenic) macrophages contribute to
   wound healing and immunosuppression [[61]23]. Macrophages are highly
   plastic cells and can alter their phenotype based on environmental
   cues. Although early BC features many M1-like macrophages that inhibit
   cancer growth through phagocytosis and secretion of tumour necrosis
   factor alpha (TNF-α) and reactive oxygen species (ROS), late-stage
   tumours are characterised by the presence of M2-like cells that
   suppress natural killer (NK) and T cells, and promote BC growth,
   metastasis, and survival [[62]24].

   Among the many processes governing macrophage polarisation is metabolic
   reprogramming, which is an important hallmark of cancer. Changes in
   serine biosynthesis in particular are linked to malignancies and
   elevated levels of 3-phosphoglycerate dehydrogenase (PHGDH), the first
   rate-limiting enzyme of the serine biosynthesis pathway, have been
   implicated in cancer progression. Moreover, in monocytes and
   macrophages, PHGDH has been reported to act as a key metabolic
   checkpoint of macrophage polarisation [[63]25], with increased PHGDH
   activity correlating with an anti-inflammatory M2-like phenotype, while
   inhibition of PHGDH promotes an M1-like pro-inflammatory macrophage
   state [[64]26, [65]27]. Interestingly, a recent study found that Th2
   cytokine, IL-4, and tumour-conditioned media lead to an upregulation of
   PHGDH in macrophages, promoting M2 polarisation. In addition, PHGDH
   ablation in vivo resulted in a shift of macrophages from M2 to M1
   state, a reduction in the infiltration of tumour-associated macrophages
   (TAMs), as well as an enhanced anti-tumour T cell immunity. These
   findings show that targeting PHGDH can reverse the immunosuppressive
   phenotype by promoting an M1-like phenotype in macrophages [[66]28].

   Despite the wealth of knowledge on the roles of LMTK3 in cancer, the
   involvement of LMTK3 in immune modulation remains largely unexplored. A
   study on early onset colorectal cancer (CRC) showed that LMTK3
   mutations positively correlate with tumour mutation burden, CD8^+ T
   cell and macrophage infiltration, as well as with a ‘hot’
   immunophenotype [[67]29]. Similarly, Wei et al. found that elevated
   LMTK3 levels in CRC associate with high immune scores and increased
   abundance of CD4^+ and CD8^+ T cells [[68]30]. However, no study to
   date has explored the role of LMTK3 in immune infiltration in BC.

   In this study, we identify Rab7, a key player in late endosomal
   trafficking, as a novel phosphorylation substrate of LMTK3, through
   which, LMTK3 modulates EV biogenesis, leading to the generation of
   larger EVs with distinct proteomic cargo. These enlarged EVs carry a
   unique set of proteins that can reshape the immune TME. We show in BC
   cells and in vivo models that exposure of monocytes to EVs derived from
   LMTK3-overexpressing cells promotes M2-like polarisation in
   macrophages. Furthermore, EVs from LMTK3 overexpressing cells reduce
   infiltration of monocytes into tumours by promoting PHGDH upregulation.
   Taken together, we present evidence of LMTK3’s implication in EV
   biogenesis and its involvement in the development of an
   immunosuppressive TME.

Materials and methods

Cell lines and cell culture

   T47D, MCF7, MDA-MB-231, THP-1 and HEK293T cells were obtained from
   American Type Culture Collection (ATCC). 4 T1 cells were obtained from
   Wuhan Pricella Biotechnology Co., Ltd. MMTV-Neu mammary carcinoma cell
   lines were generated in Professor Klinakis’ lab. Briefly, endpoint
   mammary tumours in transgenic MMTV-Neu mice were dissociated and sorted
   for CD24 + cells which were kept in culture as luminal tumour cells
   [[69]31]. MDA-MB-231, MMTV-Neu and HEK293T cells were cultured in
   Dulbecco's Modified Eagle Medium (DMEM) (Sigma Aldrich, D6046-500ML)
   supplemented with 10% foetal bovine serum (FBS) (Sigma Aldrich,
   F7524-500ML) and 1% penicillin/streptomycin (P/S) (Sigma Aldrich,
   P4333-100ML), unless otherwise specified. T47D, 4T1 and THP-1 cells
   were cultured in Roswell Park Memorial Institute 1640 medium
   (RPMI-1640) (Sigma Aldrich, R5886-500ML) supplemented with 10% FBS and
   1% P/S. The cells were maintained in a humidified incubator with 5%
   CO[2] at 37 °C and were routinely checked for mycoplasma contamination
   (Sigma Aldrich, MP0035). In experiments involving LMTK3 inhibition, C28
   (MedChemExpress, HY-153896) was used at a final concentration of 10 μM,
   diluted in DMSO.

Transfection and generation of stable cell lines

   Overexpression of LMTK3 was achieved using the pCMV6-FLAG-LMTK3 plasmid
   encoding full-length LMTK3 (Origene, RC223140). Ectopic expression of
   the Rab7 mutants was achieved using site-directed mutagenesis of
   pCMV6-Rab7 vectors generated from mCherry-Rab7a-7, a gift from Michael
   Davidson (Addgene plasmid #55127). The empty pCMV6 vector (Origene,
   PS100001) was used as a control and the pCMV6 vector containing full
   length Rab7 (pCMV6-Rab7) was used for Rab7 overexpression. Rab7
   knockout was achieved by cloning a Rab7-targeting sgRNA into the px459
   vector using the primers: 5′-CACCGTTGCTGAAGGTTATCATCC-3′ forward (Fw),
   5′-AAACGGATGATAACCTTCAGCAAC-3′ reverse (Rv). pSpCas9(BB)−2A-Puro
   (PX459) V2.0 was a gift from Feng Zhang (Addgene plasmid #62988)
   [[70]32]. Following transfection with FuGENE® HD transfection reagent
   (Promega, E2311), the cells were selected with Geneticin (G418) (Sigma
   Aldrich, G8168-10ML) or Puromycin (Gibco, A11138-03) for one week, or
   until no viable cells remained in the non-transfected control group.
   Single cell clone populations were grown and used for the validation of
   successful protein expression through western blotting. PHGDH and PSAT1
   overexpression were achieved using the lentivectors, pLJM5-PHGDH and
   pTRIPZ-PSAT1, a kind gift from Dr Katiuscia Bianchi.

Preparation of conditioned media

   Five million cells were seeded per flask in five T175 flasks in 20 mL
   complete growth media. Once 80% confluence was reached, cells were
   washed three times with 5 mL PBS. 15 mL serum-free media was then added
   to each flask and the cells were placed in an incubator for 24 h. The
   conditioned media was then collected and centrifuged at 2,000 × g for
   10 min at room temperature to remove cells and cell debris. The
   supernatant was transferred to a new tube and either processed
   immediately or stored at −20 °C for future use for up to one month.

Extracellular vesicle separation

   Centrifugation steps were performed using an Optima LE 80-k
   ultracentrifuge and a Type 70 Ti rotor. Conditioned media was
   distributed into four Beckman 25 × 89 mm polycarbonate tubes (Beckman,
   355654) and centrifuged at 10,000 × g for 30 min at 4 °C to pellet cell
   debris and large EV-like particles. The supernatant was then
   transferred into four new tubes by syringing and filtering through a
   0.2 µm filter. It was then centrifuged at 100,000 × g for 90 min at
   4 °C to pellet EVs. To wash the EVs, the four EV pellets were
   resuspended in 1 mL PBS. The EVs were then combined in a new tube. The
   volume was made up to 10 mL with PBS and the samples were centrifuged
   again at 100,000 × g for 90 min at 4 °C. The supernatant was discarded
   by pouring and the pellet resuspended in 100 µL PBS and stored at
   −80 °C for up to three months.

Immunofluorescence

   30,000 cells were plated on sterile coverslips in complete media in a
   24-well plate. Once attached, cells were washed once and fixed in 4%
   paraformaldehyde in PBS for 10 min at room temperature. Cells were then
   permeabilised with 0.1% NP-40 for 15 min at room temperature. The cells
   were then blocked with 3% BSA in PBS for 30 min at room temperature and
   incubated with primary antibodies at 1:100 dilution, in a humidified
   chamber overnight at 4 °C. Following three washes with PBS, the cells
   were incubated with fluorophore-conjugated secondary antibodies at a
   1:1000 dilution for 1 h at room temperature. The coverslips were then
   washed three times with PBS before they were mounted onto slides with
   ProLong Diamond Antifade mountant with DAPI. Cells were then imaged
   with the LSM880 confocal microscope.

Cell lysis and western blotting

   Cells were lysed in radioimmunoprecipitation assay (RIPA) buffer
   (Sigma–Aldrich, R0278) containing protease inhibitors (Roche,
   11697498001) and phosphatase inhibitors (Roche, 4906845001). Protein
   concentration was determined using the BCA assay. 50 μg (cell-based
   experiments) or 10 μg (EV-based experiments) of protein extract was
   mixed with 5 × loading buffer and heated to 95 °C for 10 min. Samples
   were run on 4–12% gels (GenScript, M00653) and transferred onto
   nitrocellulose membranes (ThermoFisher Scientific, IB23001) using the
   iBlot 2 (ThermoFisher Scientific, IB21001). Primary antibodies were
   prepared at 1:1000 dilution, unless otherwise stated in the
   supplementary methods, and incubated overnight at 4 °C. Secondary
   antibodies (1:3000) were incubated with the membrane for 1 h at room
   temperature. Horseradish peroxidase (HRP)-conjugated anti-mouse (Cell
   Signaling Technology, 7076P2) or anti-rabbit (Cell Signaling
   Technology, 7074P2) were used as appropriate. Membranes were visualised
   with SuperSignal West Pico PLUS chemiluminescent substrate
   (ThermoFisher Scientific, 34577) and images were obtained using the UVP
   ChemStudio Imaging System (Analityk jena). Quantification was performed
   using ImageJ densitometry analysis where the intensity of the bands was
   normalised to the loading control and statistical significance was
   determined using Student’s t test or one-way Analysis of Variance
   (ANOVA) as appropriate.

EGFR degradation assay

   300,000 cells were plated per well in a 6-well plate in complete growth
   media. The following day, the media was replaced with serum-free media
   for 24 h. Cells were treated with 50 µg/mL cycloheximide for 1 h prior
   to EGF treatment. 100 ng/mL EGF was added, and samples were collected
   at 0, 30, 60, 120 and 180 min and subjected to western blotting.

Co-immunoprecipitation (co-IP)

   Cells were lysed with a mild lysis buffer (10 mM Tris pH 7.5, 250 mM
   NaCl, 1% Triton X-100) to avoid protein–protein interaction disruption.
   The lysate was sonicated for three cycles, 20 s on and 30 s off in an
   ultrasonic homogeniser (Scientz, JY99-IIDN). 1,500 µg of cell lysates
   were used for each co-IP reaction. The lysates were incubated with
   1:100 of FLAG antibody or mouse IgG isotype antibody for the
   immunoprecipitation of FLAG-LMTK3 and with 1:100 of Rab7 or rabbit IgG
   antibody for the immunoprecipitation of Rab7. The reactions were
   incubated overnight at 4 °C under rotation. The next day, 5 µL of
   Protein G magnetic beads (Thermo Scientific, 88847) for the mouse
   antibodies and Protein A magnetic beads (Thermo Scientific, 88845) for
   the rabbit antibodies were added to each reaction and incubated for 1 h
   at room temperature under rotation. Elution of the co-IP was performed
   by adding 40 µL of 3 × SDS sample buffer (Cell Signaling, 7722) to the
   magnetic bead immunocomplex and the samples were incubated at 95 °C for
   5 min. The eluates were then used for western blotting analysis.

Flow cytometry

   PBMC derived macrophages were analysed by flow cytometry using the BD
   Accuri c6. 500,000 cells were centrifuged at 2,000 × g for 3 min and
   the supernatant was discarded. The cells were resuspended in 500 µL
   resuspension buffer (0.5% BSA in PBS) and blocked with Fc Receptor
   Blocking Solution (Biolegend, 422302) for 30 min. They were then
   stained with antibodies (1:20) for 1 h and propidium iodide (1:1000)
   for 10 min. A minimum of 10,000 cells were analysed per sample.
   Mouse-derived tumour cell suspensions were stained for 20 min, and data
   were acquired by BD FACS Celesta flow cytometer, using BD FACSDiva
   software. Cells were gated as singlets and live cells, according to
   FSC-A/FSC-H and FSC-A/SSC-A scatter plots. Following this, CD45 + cells
   were gated, and total macrophages were selected as the double-positive
   (CD11b + F4/80 +) population. The M1 macrophage population was selected
   as MHCII+CD206- and the M2 population as MHCII-CD206+. FACS analysis
   was performed using FlowJo software, and statistical analysis was
   performed with GraphPad Prism.

In vitro kinase assays

   In vitro kinase assays using radiolabelled adenosine triphosphate (ATP)
   were carried out to assess Rab7 phosphorylation by LMTK3. Recombinant
   human Rab7 was incubated with baculovirus expressed LMTK3 kinase domain
   and [y-^32P]ATP for 60 min at 30 °C, unless otherwise specified. The
   reaction mixture was then separated by SDS-PAGE using a 4–20%
   acrylamide gel. The gel was then stained with Quick Coomassie
   (NeoBiotech, NB-45–00078) overnight and de-stained for 1 h in three
   changes of 1% glycerol. Gels were then dried for 2 h at 64 °C in a
   heated vacuum gel dryer (Cleaver Scientific). X-ray films (Scientific
   Laboratories Supplies, MOL7016) were added to the gels in an
   autoradiography cassette (Amersham Biosciences, [71]RPN11649). After 24
   h of exposure, films were developed using an ECOMAX x-ray film
   processor (Protec).

TEM imaging

   Samples were visualised using the JEOL JEM1400-Plus (Jeol, Tokio,
   Japan) (120 kV, LaB6) microscope with a Gatan OneView 4 K camera.
   Images were taken at 2,000 ×, 10,000 ×, 12,000 × and 30,000
   × magnification. For sample preparation, please refer to the
   supplementary methods.

Quantitative real-time PCR

   Total RNA was extracted using the PureLink RNA mini kit (Invitrogen,
   12183018A). cDNA was synthesised using the high-capacity cDNA reverse
   transcription kit (ThermoFisher Scientific, 4368814). Quantitative
   RT-PCR was performed using the Luna Universal qPCR Master Mix (NEB,
   M3003S) in a StepOne thermal cycler (Applied Biosystems). At least
   three biological and three technical replicates were performed per
   experiment. Glyceraldehyde 3-Phosphate Dehydrogenase (GAPDH) was used
   as an internal control. The following primers were used: GAPDH Fw
   5’-TCGGAGTCAACGGATTTG-3’, Rv 5’-CAACAATATCCACTTTACCAGAG-3’, CD86 Fw
   5’-AGTGAACAGACCAAGAAAAGAGAAA-3’, Rv 5’-AGGTTGCCCAGGAACTTACAA-3’, NOS2
   Fw 5’-CAATGGCAACATCAGGTCGG-3’, Rv 5’-GTGAATTCCACGTTGGCAGG-3’, CD206 Fw
   5’-TGCTACTGAACCCCCACAAC-3’, Rv 5’-AACCAGAGAGGAACCCATTCG-3’, TGFB1 Fw
   5’-GGTGGAAACCCACAACGAAAT-3’, Rv 5’-GAGCAACACGGGTTCAGGTA-3’.

Nanoparticle tracking analysis

   The Nanosight NS300 was used to quantify EVs and measure their size. A
   1 mL syringe and syringe pump were used to feed the EV suspension into
   the flow chamber of the Nanosight. 2–10 µL of EV suspension was diluted
   in 1000 µL PBS and loaded using a 1 mL syringe. A flow rate of 70 AU
   was used, and the camera setting was between 13 and 15. The tubes were
   washed with 2 mL PBS before and after each use to remove debris and the
   flow cell was washed in ddH[2]O prior to use. Five measurements of 60 s
   were taken per sample. The camera type was sCMOS and the laser type was
   Blue488.

Single particle interferometric reflectance imaging

   The Unchained Labs Leprechaun was used for EV subpopulation analysis
   according to the manufacturer's instructions using the Leprechaun
   Exosome Human Tetraspanin Kit (Unchained Labs, 251–1044). Conditioned
   media from control and LMTK3 cells was diluted 1:4 with dilution buffer
   and 50 µL was added per chip. Chips were printed with anti-CD9,
   anti-CD63, anti-CD81 and anti-mouse IgG antibodies. The chips were
   incubated overnight and the following day, the chips were washed three
   times then incubated with primary antibodies for 1 h at room
   temperature. The chips were then washed three times with wash buffer
   then in deionised water then dried, imaged, and analysed with the
   Leprechaun.

PBMC and monocyte isolation

   Blood from healthy volunteers was collected by venepuncture in four
   4 mL EDTA tubes and transferred to a 50 mL falcon tube. The isolation
   of peripheral blood mononuclear cells (PBMCs) was performed as
   previously described following the Ficoll-Paque (Merck, Cytiva
   17–1440-02) method [[72]33]. Following this, monocytes were isolated by
   plastic adhesion [[73]33]. Further monocyte population enrichment was
   performed using the PluriSpin Human Monocyte Enrichment solution
   (PluriSelect, 19–01001) according to the manufacturer’s instructions.

Monocyte differentiation

   THP-1 monocytes were differentiated into macrophages by incubation with
   10 ng/mL of 12–0-tetradecanoylphorbol-13-acetate (PMA) (Sigma Aldrich,
   524400) for 24 h. The media was then changed to normal complete media
   and the macrophages were allowed to rest for another 24 h.

3D co-culture of monocytes and breast cancer cells

   Tumour spheroids were cultured by coating each well of a 96-well plate
   with 100 µL of 1% agarose in PBS. 2,500 T47D or MDA-MB-231 cells were
   then added in 100 µL complete media and allowed to grow for 7 days.
   Monocytes were stained with green lipophilic tracer (Invitrogen,
   L7781). The monocytes were then added to spheroids with 2,500 monocytes
   per spheroid. After 24 h of co-culture, spheroids were washed three
   times in PBS and imaged using the JuLi stage live cell imaging
   microscope to assess infiltration. ImageJ was used to quantify the mean
   fluorescence within the spheroids. Briefly, spheroid fluorescence
   intensity was measured by drawing a region of interest around each
   spheroid and using the ‘measure’ function to obtain the mean
   fluorescence intensity. Background intensity was measured in a
   cell-free area and subtracted from the spheroid signal.

Extracellular vesicle uptake

   CD63-RFP was expressed in T47D using pCT-CD63-RFP (System Biosciences,
   CYTO120R-PA-1) through lentiviral transduction and puromycin selection
   to generate a stable cell line. Cells were seeded in T175 flasks and
   EVs were collected as previously described. Monocytes were grown on
   coverslips in a 24-well plate and treated with different concentration
   of CD63-RFP EVs for 24 h. Following this, the plates were centrifuged
   at 300 × g for 3 min and monocytes were fixed with 4% paraformaldehyde
   and mounted to the coverslips in 4',6-diamidino-2-phenylindole
   (DAPI)-containing ProLong Diamond Antifade mountant (ThermoFisher
   Scientific, [74]P36962). They were then imaged with the LSM880 confocal
   microscope and EV uptake was assessed by measuring mean fluorescence
   intensity of cells using ImageJ.

Monocyte proliferation and viability assay

   100,000 THP-1 monocytes were plated per well in a 24-well plate and
   treated with PBS or 10 µg/mL of control or LMTK3 EVs. At 24 h
   intervals, monocytes were counted, using the Countess automatic cell
   counter (ThermoFisher Scientific, [75]C10281), and viability was
   measured using Trypan blue staining.

Immune cell infiltration

   Data analysis of immune cell infiltration into tumours in clinical
   samples was performed using publicly available TCGA (The Cancer Genome
   Atlas) datasets [[76]34–[77]36], which were accessed through cBioPortal
   [[78]37–[79]39]. TIMER2.0 (Tumour Immune Estimation Resource version
   2.0) [[80]7] and xCell were used to obtain infiltration scores for
   immune cells in different tumours [[81]40]. LMTK3 expression levels
   from RNA sequencing data were taken from cBioPortal and separated into
   four quartiles based on levels of LMTK3 and the first (Q1) and last
   (Q4) quartiles were compared to assess immune cell infiltration in
   tumours. For gene correlation analysis, mRNA levels for different M1
   and M2 macrophage marker genes was then compared between these patient
   groups.

Animal models and sample processing

   For the MMTV-Neu model, twelve male, Rag1-/- mice between 10 and 14
   weeks old, obtained from Jackson labs
   ([82]https://www.jax.org/strain/002216) were used in this experiment.
   MMTV-Neu control and LMTK3 overexpressing cells were used to grow
   subcutaneous or orthotopic tumours in the mice. Mice were grouped into
   two groups of six mice each, one for control and one for LMTK3 cells.
   2 × 10^6 cells in 100 μL DMEM/Matrigel mix (in 3:1 ratio) were injected
   in each murine flank for subcutaneous implantation or into the mammary
   fat pad for orthotopic implantation. From the day of injection (Day 0),
   mice were monitored for tumour growth twice a week, and tumour size was
   measured using a calliper.

   For the 4T1 mouse model, SPF grade female BALB/c mice aged 6–8 weeks
   were purchased from Shanghai SLAC Experimental Animal Co., Ltd.
   (Shanghai, China). 4T1 cells were used to grow subcutaneous tumours in
   the right flanks of animals by injection of 5 × 10^5 cells in 100 μL
   DMEM/Matrigel mix (in 1:1 ratio). Tumours were measured with a calliper
   and mice were injected with 50 µg EV protein/mouse in 50 μL PBS
   intratumourally three times per week.

Tumour isolation

   All mice were sacrificed on day 32 (MMTV-Neu subcutaneous), day 23
   (MMTV-Neu orthotopic) or day 29 (4T1) and tumours were excised for
   enzymatic digestion. Additionally, tumour biopsies were stored at
   −80 °C from each tumour, as well as tumour slices for paraffin
   embedding. Part of the tumours was cut and put in dissociation buffer,
   composed of 1 mg/mL Collagenase D (Roche, 11088858001) and 1 mg/mL
   Dispase (Sigma-Aldrich, D4693-1G) in DMEM F12 medium for 45 min. Once
   enzymatically dissociated, a single cell suspension was obtained after
   mechanical dissociation and passing through 70 μm cell strainer. The
   resulting mixture was centrifuged at 400 × g for 5 min. Red blood cells
   were lysed using ACK lysis buffer for 10 min, and RBC-free cell
   suspension was obtained to be used for flow cytometry. IHC was
   performed as described in the supplementary methods.

Statistical analyses

   Unless otherwise specified, data analysis was carried out using
   GraphPad Prism 9. Statistical tests were used depending on the
   experimental design as appropriate. Experiments were performed in
   triplicate unless otherwise stated. Column and line graphs show mean
   ± standard error of the mean (SEM) unless otherwise stated. Statistical
   significance was assessed using Student's t-test for comparison between
   two groups and one-way ANOVA followed by Tukey’s multiple comparison
   test for multiple groups. A p-value < 0.05 was considered statistically
   significant.

   For microscopy experiments where the size of structures is given,
   ImageJ was used. A line was drawn across the longest axis of each
   structure and the ‘measure’ function was used to determine the
   diameter.

Data availability

   The proteomics datasets generated in this study are both provided as
   supplementary tables S1 and S2.

Results

LMTK3 increases the size and regulates the proteomic cargo of EVs released by
breast cancer cells

   Considering the involvement of LMTK1 and LMTK2 in the endocytic pathway
   and the structural similarity between LMTK isoforms, we hypothesised
   that LMTK3 may also play a role in this fundamental process, as well as
   in EV biogenesis. To explore this, we initially generated a panel of BC
   cell lines (T47D, MCF7, and MDA-MB-231) stably overexpressing LMTK3 to
   explore how LMTK3 abundance impacts EVs. Western blotting confirmed the
   successful overexpression of LMTK3 (Fig. [83]1a). As previously
   reported, both bands represent full-length LMTK3, which may be subject
   to post-translational modifications [[84]1]. Following collection of
   conditioned media, we used differential ultracentrifugation (DUC) to
   separate EVs, then validated the enrichment of EVs according to the
   minimal information for studies of extracellular vesicles (MISEV)
   guidelines [[85]41] (Fig. [86]1b). Nanoparticle tracking analysis (NTA)
   was employed to determine particle size distribution and concentration
   (Fig. [87]1c-d). The analysis showed a significant increase in average
   EV size in all LMTK3-overexpressing BC cell lines (Fig. [88]1c),
   although, the EV concentration was not significantly altered
   (Fig. [89]1d). Furthermore, LMTK3 inhibition following treatment with
   C28 caused a decrease in average EV size without altering EV
   concentration. This suggests that LMTK3 affects an essential aspect of
   EV biogenesis without altering the quantity of EVs produced. Particle
   size distributions across the BC cell lines displayed the expected
   pattern, with a mode of ~ 100 nm and a right-skewed distribution
   (Figure S[90]1a-c). LMTK3 overexpression increased the average diameter
   of EVs by approximately 20–30%, translating into a significant increase
   in EVs volume. The increase in EVs size was confirmed by an orthogonal
   measurement using transmission electron microscopy (TEM)
   (Fig. [91]1e-h).

Fig. 1.

   [92]Fig. 1
   [93]Open in a new tab

   Overexpression of LMTK3 increases average EV size, alters EV protein
   cargo, and changes the EV subpopulation distribution. a. Western blot
   for LMTK3 in LMTK3-overexpressing cells compared to control (pCMV6
   empty vector) cells. GAPDH is shown as a loading control (n = 3). b.
   Western blot of cell lysate and EV fractions showing enrichment of the
   EV markers, CD9 and CD81, presence of HSP70 and depletion of GM130 in
   EVs compared to cell lysates (n = 3). c. Summary data showing mean EV
   size. d. NTA showing concentration of EVs. e–g. Representative
   transmission electron micrographs showing EVs from control and
   LMTK3-overexpressing breast cancer cells. Scale bars = 500 nm (12,000
   x) and 100 nm (30,000 x). h. Quantification of (e–g). 500 EVs were
   measured for each condition. A representative distribution of EV size
   is shown. Median is shown with a solid line, and quartiles are shown
   with dotted lines. i. Proteomic analysis of EVs from control and T47D
   LMTK3-overexpressing cells. Volcano plot showing EV protein cargo
   changes as log[2](fold change) of LMTK3 EVs/Control EVs. Significantly
   upregulated hits are shown in red and downregulated hits in blue
   (Welch’s t test, p < 0.05). j,k. Western blotting of T47D (j) and
   MDA-MB-231 (k) cell lysate and EV fractions for control (Ct) and
   LMTK3-overexpressing cells. PSAT1, LDHB, CD63 and TSG101 were validated
   as hits from the proteomics analysis. CD9 and HSP70 were used as
   positive EV markers. GM130 was used as a negative EV marker (n = 3). l.
   Quantification of PSAT1 and LDHB levels in LMTK3 EVs relative to
   control EVs compared to HSP70. m. Quantification of CD63 and TSG101
   levels in EVs relative to control EVs compared to HSP70. n. EV diameter
   for control and LMTK3 cells as measured by SP-IRIS. An increase in EV
   size is seen when LMTK3 is overexpressed for all three capture
   antibodies. o. The percentage abundance of particles positive for CD63,
   CD81 and CD9. CD63 positive particles were downregulated when LMTK3 was
   overexpressed. Graphs show the mean ± SEM of three independent
   experiments (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001;
   two-tailed unpaired Student’s t-test) (c,d,l,m,n,o)

   Given the impact of LMTK3 on EV size, we next investigated whether
   LMTK3 can influence the proteomic cargo of BC cell-derived EVs. Tandem
   mass tag (TMT)-based mass spectrometry revealed substantial alterations
   in the protein content of EVs secreted by LMTK3-overexpressing BC cells
   (LMTK3 EVs) compared to control cells (Fig. [94]1i, Table S[95]1).
   Interestingly, LMTK3 protein was also detected within the EVs following
   its overexpression. Gene ontology (GO) pathway enrichment analysis
   highlighted significant upregulation of proteins associated with amino
   acids, endocytosis, Wnt and ERBB signalling, and other biological and
   cellular processes. Pathway analysis of the data indicate that LMTK3
   EVs may play a functional role in modulating the TME via different
   mechanisms and signalling pathways (Figure [96]S1d) [[97]42]. Western
   blotting was used to validate different EV cargo proteins, including
   phosphoserine aminotransferase 1 (PSAT1) and lactate dehydrogenase B
   (LDHB) that were identified through our proteomics approach. These key
   metabolic enzymes, previously reported to be implicated in cancer
   progression [[98]43–[99]45], were upregulated in the EVs secreted by
   LMTK3 overexpressing cells, while their total cellular levels remained
   unchanged, suggesting that LMTK3 selectively enhances their
   incorporation into EVs rather than regulating their cellular expression
   (Fig. [100]1j-m). Interestingly, several key components of exosome
   biogenesis were downregulated in EVs when LMTK3 was overexpressed,
   including syndecan-2 (SDC2), tumour suppressor gene 101 (TSG101) and
   tetraspanins (CD63 and TSPAN14), which further points to a role of
   LMTK3 in regulating EV biogenesis. Of these, CD63 and TSG101 were
   further validated as downregulated in LMTK3 overexpressing cell-derived
   EVs through western blotting (Fig. [101]1j-m).

   EVs are heterogeneous in size, biogenesis, and function [[102]46,
   [103]47]. To determine whether LMTK3 influences specific EV
   subpopulations, single-particle analysis was performed using the
   Unchained Labs Leprechaun® instrument, which employs single particle
   interferometric reflectance imaging sensor (SP-IRIS) technology to
   quantify surface EV markers [[104]48]. EVs were immunocaptured with the
   tetraspanins CD9, CD63 and CD81, then immunolabelled with fluorescent
   antibodies against these proteins. The results showed a significant
   reduction in CD63-positive particles in LMTK3-overexpressing cells,
   suggesting a disruption in the tetraspanin-dependent exosome biogenesis
   pathway. This reduction aligns with the proteomics findings showing
   lower CD63 levels in LMTK3-derived EVs (Fig. [105]1n-o).

   Taken together, using various experimental approaches, we demonstrated
   that LMTK3 upregulation in BC results in the formation of larger EVs
   with distinct protein cargo and changes to EV subpopulations, strongly
   suggesting a role for LMTK3 in EV biogenesis.

LMTK3 regulates EV biogenesis and EV cargo loading by phosphorylating Rab7 at
Ser72

   Next, we investigated the possibility that LMTK3 regulates EV
   biogenesis through protein phosphorylation. To explore this hypothesis,
   we analysed previously published phosphoproteomics data [[106]7] to
   identify prospective LMTK3 substrates involved in EV biogenesis. Our
   focus laid on Rab7 (Ser72), a small GTPase with key roles in MVB
   maturation and exosome release.

   Previous studies have shown LMTK3 to co-localise with endosomal and
   Golgi markers in neurons, where it is involved in the endocytic
   trafficking of N-methyl D-aspartate (NMDA) receptors, while in BC
   cells, it is predicted to localise to the Golgi apparatus and early
   endosomes [[107]49]. Rab7 is mainly detected in late endosomes (MVBs),
   which are the subcellular structures where exosomes are generated
   through the formation of ILVs and then released through fusion with the
   plasma membrane. To assess the prospective co-localisation between
   LMTK3 and Rab7, confocal microscopy was used. Following
   immunofluorescence staining, our analysis showed overlapping
   localisation between LMTK3 and Rab7, confirming that LMTK3 is present
   in late endosomes (Fig. [108]2a-b). Furthermore, co-immunoprecipitation
   assays using antibodies for FLAG-LMTK3 or Rab7 also confirmed the
   interaction between LMTK3 and Rab7 (Fig. [109]2c).

Fig. 2.

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

   LMTK3 directly phosphorylates Rab7 at Ser72. a. Immunofluorescence
   images of LMTK3 co-localisation with Rab7 in T47D and MDA-MB-231 cells.
   Scale bars = 10 µm and 1 µm (zoom). Images were taken with the LSM880
   with Airyscan and a 63 × objective lens. b. Graph showing mean
   Pearson’s correlation coefficient (r) for co-localisation between LMTK3
   and Rab7 (n = 50 cells). c. Co-immunoprecipitation western blot showing
   interaction between LMTK3 and Rab7 in T47D LMTK3 cells. IgG is used as
   a negative control. Anti-FLAG (top) or anti-Rab7 (bottom) antibody was
   used to capture FLAG-LMTK3 and Rab7 respectively from cell lysates (n =
   3). d. Western blot showing overexpression of LMTK3 leads to an
   increase in phospho Rab7 (serine 72). Tubulin was used as a loading
   control (n = 3). e. Quantification of intensity of pRab7/Rab7 in T47D
   and MDA-MB-231. Graphs show the mean ± SEM of three independent
   experiments (*p < 0.05, **p < 0.01; two-tailed unpaired Student’s
   t-test). f. Western blot showing a decrease in Rab7 Ser72
   phosphorylation following LMTK3 inhibition by C28. Tubulin is shown as
   a loading control (n = 3). g. Quantification of intensity of pRab7/Rab7
   in T47D and MDA-MB-231. h. Western blot showing overexpression of LMTK3
   in C28-treated cells restores Rab7 phosphorylation, showing specificity
   of C28 (n = 3). i. Quantification of (h). j. In vitro kinase assay of
   Rab7 and HSP27 (positive control) as substrates. An autoradiogram was
   produced to show phosphorylated proteins. 1 µM C28 was used to inhibit
   LMTK3 kinase activity. k. Western blot showing Rab7 phosphorylation by
   LMTK3 at Ser72. Following an in vitro kinase assay, proteins were
   separated by SDS PAGE. Western blot for phosphorylated Rab7 (Ser72) is
   presented (n = 3). l. Western blot showing overexpression of Rab7
   mutant constructs in T47D and MDA-MB-231. Tubulin was used as a loading
   control (n = 3). m. Nanoparticle tracking analysis showing Rab7 S72E
   overexpression results in an increase in average EV size in T47D and
   MDA-MB-231 cells. n. Nanoparticle tracking analysis showing
   concentration of EVs from Rab7 mutant T47D and MDA-MB-231 cells. o,p.
   EVs were collected from the Rab7 mutant cell lines. Cell lysates and
   EVs were subjected to SDS-PAGE followed by western blotting. Figure
   showing a representative western blot showing PSAT1 levels in Rab7
   mutant T47D (o) and MDA-MB-231 (p) cells. GM130 was used as a cellular
   marker and a negative EV marker. HSP70 and CD9 were used as positive EV
   markers (n = 3). q. Quantification of (o,p). Graphs show the mean ± SEM
   of at least three independent experiments (*p < 0.05, **p < 0.01, ***p
   < 0.001, ****p < 0.0001; one-way ANOVA, followed by Tukey’s multiple
   comparisons test)

   To investigate whether Rab7 mediates the observed LMTK3-dependent
   changes in EV biogenesis, we first assessed Rab7 phosphorylation at
   Ser72 in LMTK3-overexpressing cells. Western blotting confirmed that
   LMTK3 upregulation increases Rab7 phosphorylation without significantly
   altering Rab7 protein levels (Fig. [112]2d-e). Furthermore, inhibition
   of LMTK3 using the selective LMTK3 inhibitor (C28) [[113]7, [114]50,
   [115]51] reduced Rab7 Ser72 phosphorylation in wild-type cells
   (Fig. [116]2f-g), an effect which was restored following overexpression
   of LMTK3 (Fig. [117]2h-i).

   To elucidate if Rab7 is directly or indirectly phosphorylated by LMTK3,
   we performed in vitro kinase assays, using recombinant LMTK3 kinase
   domain as source of enzyme activity and Rab7 as substrate. Heat shock
   protein 27 (HSP27) was used as a positive control of LMTK3
   phosphorylation [[118]7]. Our results demonstrated that LMTK3 directly
   phosphorylates Rab7 in vitro and that this effect is abolished
   following LMTK3 pharmacological inhibition (Fig. [119]2j and Figure
   S2). Moreover, in vitro kinase assays using non-radiolabelled ATP
   followed by western blotting for Rab7 (phospho-Ser72) further validated
   the exact phosphorylation residue (Ser72) (Fig. [120]2k).

   To further explore the functional significance of Rab7 phosphorylation
   at Ser72, single cell clones (T47D and MDA-MB-231) overexpressing
   wild-type Rab7, Rab7 phospho-mimetic (S72E; Ser72Glu) or phospho-null
   (S72A; Ser72Ala) mutants were generated (Fig. [121]2l). Having
   previously shown that LMTK3 upregulation increases EV size, we explored
   whether the phospho-mimetic Rab7 mutant (S72E) could recapitulate this
   effect. NTA of EVs confirmed that S72E mutation significantly increased
   EV size, mirroring the effects seen on EV biogenesis following LMTK3
   overexpression (Fig. [122]2m). The overexpression of Rab7 mutants did
   not, however, have any impact on EV number (Fig. [123]2n). Furthermore,
   as expected, overexpression of the Rab7 S72A mutant did not alter the
   EVs, since it did not change the levels of phospho-Rab7 in the cells.
   Notably, as shown in Fig. [124]2o-q, the S72E phospho-mimetic Rab7
   mutant also leads to enhanced packaging of PSAT1 into EVs, consistent
   with the effect observed upon overexpression of LMTK3. Together, these
   data support the hypothesis that LMTK3 alters EV cargo via Rab7
   phosphorylation.

LMTK3 upregulation disrupts multivesicular body trafficking

   To examine the impact of LMTK3 overexpression on a cellular scale, we
   evaluated whether Rab7 phosphorylation by LMTK3 impacts
   endosome-lysosome trafficking using an EGFR degradation assay. Rab7
   typically interacts with RILP to facilitate the transport of endosomes
   to lysosomes, however, previous studies have shown that phospho-Rab7 at
   Ser72 disrupts this interaction [[125]52–[126]54], and as a result,
   MVBs are degraded at a reduced rate. EGF treatment of serum-starved
   cells leads to the internalisation of EGFR and its subsequent
   degradation by the Rab7-dependent transport to the lysosome [[127]54].

   Control and LMTK3 overexpressing cells were serum-starved overnight
   then treated with cycloheximide for one hour. They were then treated
   with EGF, and western blotting was used to assess EGFR levels at
   different timepoints. Our results demonstrated that EGFR levels within
   cells decrease rapidly after EGF treatment in control cells, however
   the rate of EGFR degradation was reduced in LMTK3 overexpressing cells
   (Fig. [128]3a-b), implicating LMTK3 in the fusion process of endosomes
   with lysosomes and the degradation of endocytic cargo.

Fig. 3.

   [129]Fig. 3
   [130]Open in a new tab

   Rab7 phosphorylation by LMTK3 increases MVB and ILV size. a. Western
   blot showing EGFR degradation in control and LMTK3 overexpressing T47D.
   Cells were collected at 0, 30, 60, 120 and 180 min after treatment with
   EGF. GAPDH is shown as a loading control (n = 3). b. Quantification of
   (a). *, p < 0.05, **, p < 0.01 (two-way ANOVA with Bonferroni’s
   multiple comparisons test). c. Representative electron micrographs of
   control and LMTK3 MDA-MB-231. Figure shows MVBs containing ILVs. Images
   taken at 10,000 × magnification. Scale bars = 200 nm, at least 50
   structures were measured per condition from at least 30 fields of view,
   3 biological replicates were performed. d,e. Quantification of (c). f.
   Western blot showing Rab7 knockout cells and re-expression of Rab7
   wild-type or Rab7 S72 A mutant (n = 3). g. T47D Rab7 wild-type or Rab7
   S72 A cells stably expressing CD63-RFP with and without LMTK3
   overexpression were imaged by confocal microscopy to assess changes in
   MVB diameter. Scale bars = 10 µm or 1 µm (zoom), n = 500 MVBs. h.
   Quantification of MVB diameter. Graphs show the mean ± SEM (*p < 0.05,
   **p < 0.01, ***p < 0.001, ****p < 0.0001; one-way ANOVA, followed by
   Tukey’s multiple comparisons test)

   Given that Rab7 is integral to the formation of MVBs, and Rab7 deletion
   has been linked to enlarged MVBs and changes in ILVs [[131]55,
   [132]56], we hypothesized that LMTK3-mediated phosphorylation and
   inactivation of Rab7 would lead to similar changes. TEM analysis of
   cell sections revealed that LMTK3 overexpression significantly
   increases both MVB and ILV size (Fig. [133]3c-e), supporting the
   hypothesis that LMTK3 alters EV biogenesis through disrupting MVB
   trafficking, therefore leading to a change in EV subpopulations
   released by cells. As ILVs are the precursors to exosomes, these data
   suggest that the regulation of Rab7 by LMTK3 increases EV size through
   the exosomal pathway.

   To further explore if the LMTK3-related changes in MVBs were dependent
   on Rab7, we used confocal microscopy to examine the impact of LMTK3
   overexpression on MVB size in cells with wild-type Rab7 and Rab7 S72A
   mutant, which cannot be phosphorylated at serine 72. Firstly, Rab7
   knockout cells were generated then either Rab7 wild-type or Rab7 S72A
   were re-expressed in the cells (Fig. [134]3f). Interestingly, in Rab7
   wild-type cells, LMTK3 overexpression caused a significant increase in
   average MVB size, however this effect did not occur in Rab7 S72A cells,
   indicating that LMTK3 disrupts MVB dynamics predominantly through
   phosphorylation of Rab7 at Ser72 (Fig. [135]3g-h).

   In summary, our findings demonstrate that LMTK3 overexpression in
   breast cancer cells promotes the phosphorylation of Rab7 at Ser72,
   which in turn disrupts late endosome trafficking and results in the
   formation of enlarged MVBs. This altered MVB biogenesis drives the
   production of enlarged EVs with distinct protein cargo, including
   increased packaging of PSAT1 and LDHB. The data suggest a novel
   mechanism by which LMTK3 regulates EV biogenesis and trafficking,
   contributing to its oncogenic potential through the modulation of Rab7
   activity.

LMTK3 EVs alter the proteome of monocytes and reduce their infiltration into
tumour spheroids

   The increase in EV size, change in subpopulation distribution, and
   altered protein cargo unveiled a novel role of LMTK3 in the regulation
   of EV biogenesis. As EVs influence the behaviour of surrounding cells
   and considering the contribution of the immune compartment in cancer
   progression, we next explored the impact of LMTK3 abundance on the BC
   immune microenvironment.

   A computational analysis was conducted to compare LMTK3 expression
   against the levels of immune cell infiltration, through deconvolution
   of publicly available BC tumours RNA datasets [[136]34–[137]36]. Our
   analysis showed significant alterations in different populations of
   immune cells based on LMTK3 abundance (Fig. [138]4a and Figure S3).
   Since monocytes and macrophages were among the significantly altered
   cell types and acknowledging that macrophage polarisation is a key
   factor in the BC TME, we next assessed various M1 (CD80, CD86 and
   HLA-DRA) and M2 (CD163, TGFB1 and STAT6) macrophage markers in the
   available clinical samples. Interestingly, M1 markers were reduced in
   tumours with high levels of LMTK3, while, TGFB1 and STAT6 M2 markers
   (but not CD163) were upregulated, also highlighting the plasticity and
   heterogeneity of different subsets of M1/M2-like macrophages [[139]57]
   (Fig. [140]4b). Given the key role of monocytes and monocyte-derived
   cells in BC, and the LMTK3-dependent changes in both monocyte
   infiltration and macrophage polarisation in clinical samples, we chose
   to focus exclusively on monocytes in the present study.

Fig. 4.

   [141]Fig. 4
   [142]Open in a new tab

   Immune cell infiltration is altered by LMTK3 abundance. a. RNA
   sequencing data and TIMER2.0 infiltration data were used to compare
   LMTK3 abundance with infiltration of different immune cell populations.
   Graph shows the mean ± SEM infiltration scores in LMTK3 low vs LMTK3
   high samples (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001;
   unpaired two-tailed Student’s t-test). b. Changes in the levels of
   three M1 and three M2 macrophage markers in breast cancer clinical
   samples, displayed as mRNA expression z-scores relative to all samples
   (log RNA Seq V2 RSEM). Q1 (lowest quartile of LMTK3 expression; blue)
   and Q4 (highest quartile of LMTK3 expression; red) are shown. Graphs
   show the mean ± SEM (*p < 0.05, **p < 0.01, ***p < 0.001, ****p <
   0.0001; two-tailed unpaired Student’s t-test), n = 4,715. c. Schematic
   diagram of the spheroid infiltration assay. d. Representative images of
   T47D tumour spheroids with infiltrating monocytes stained in green. e.
   Quantification of (d). Data are shown from three biological repeats.
   Graphs show the mean ± SEM of three independent experiments (*p < 0.05,
   **p < 0.01, ***p < 0.001, ****p < 0.0001; one-way ANOVA, followed by
   Tukey’s multiple comparisons test). f,g. PBMC-derived monocytes were
   treated with PBS, control EVs or LMTK3 EVs. They were then co-cultured
   with T47D tumour spheroids and imaged after 24 h. f. Representative
   images of tumour spheroids with infiltrating monocytes. g.
   Quantification of (f). h. THP-1 monocytes were treated with either
   control EVs or LMTK3 EVs. Proteomic analysis revealed changes in the
   protein expression of the monocytes. Figure shows a volcano plot of the
   proteomic analysis. X axis shows log[2](fold change) of LMTK3
   EV-treated THP-1/Control EV-treated THP-1. Significant upregulated hits
   are shown in red and downregulated hits in blue. Welch’s t test was
   used to calculate p values. Dotted line indicates p = 0.05. i. Volcano
   plot showing proteomics data of control and LMTK3 EV-treated monocytes
   displaying only proteins involved in M1 or M2 polarisation. Red points
   are associated with M2 polarisation and blue points are associated with
   M1 polarisation. j. RT-qPCR showing changes in macrophage markers
   following EV treatment. Graphs show the mean ± SEM of three independent
   experiments (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001;
   one-way ANOVA, followed by Tukey’s multiple comparisons test)

   Initially, we showed that LMTK3 overexpression in T47D breast cancer
   cells did not impact the uptake of EVs by monocytes (Figure S4a-S4d).
   Furthermore, the EVs had no effect on THP-1 proliferation or viability
   (Figure S4e-S4f). To examine the effect of LMTK3-derived EVs on
   monocytes in a physiologically relevant setting, a 3D co-culture assay
   was used. Briefly, either THP-1 or PBMC-derived monocytes were stained
   with a green, fluorescent lipophilic tracer and treated with EVs
   derived from either control or LMTK3-overexpressing T47D cells
   [[143]58]. High purity monocytes were obtained from PBMCs through
   plastic adhesion, followed by negative selection using PluriSpin Human
   Monocyte Enrichment solution (Figure S5a-S5b). A concentration of 10
   µg/mL of EV protein was used, which is within the typical range for
   such functional EV studies [[144]59, [145]60]. Following EV treatment,
   monocytes were co-cultured with T47D spheroids to allow infiltration
   into the tumours (Fig. [146]4c). Our data demonstrated that
   LMTK3-derived EVs reduce the infiltration of both immortalised (THP-1)
   and primary monocytes (PBMC-derived) into BC tumour spheroids
   (Fig. [147]4d-g). These results were also validated in MDA-MB-231
   spheroids using MDA-MB-231 EVs (Figure S6a-S6b). Additionally, a rescue
   experiment was carried out where LMTK3 overexpressing cells were
   treated with C28 and EVs were collected. This abrogated the ability of
   LMTK3 EVs to reduce monocyte infiltration, suggesting that the effect
   is specific to LMTK3 (Figure S6c-S6d).

   To elucidate the signalling pathways through which the EV-induced
   changes in monocyte infiltration occur, a proteomics experiment was
   conducted to assess changes in monocyte protein expression after
   treatment with LMTK3 EVs. The proteomic analysis identified 15,944
   proteins, of which, 1,441 were significantly altered between LMTK3 and
   control EV-treated monocytes (p < 0.05) (Fig. [148]4h, Figure S6e,
   Table S2). GO analysis showed enriched proteins are involved in amino
   acid metabolic processes, catabolic processes and pre-replicative
   complex assembly (Figure S5c). Among the top upregulated hits were
   PHGDH, DHX57, CNBP, FAH and REPIN1, and the top downregulated hits
   included S100A10, KLF13, MTG2, UCHL1, TRPC3 and PIP4P2. Transient
   receptor potential cation channel subfamily C member 3 (TRPC3) and
   Ubiquitin C-Terminal Hydrolase L1 (UCHL1) are both known to play roles
   in promoting M1 macrophage polarisation [[149]61, [150]62].
   Furthermore, S100A10 has been linked to increasing macrophage invasion
   in tissue [[151]63], therefore downregulation of S100A10 may reduce
   monocyte infiltration into tumours. In addition, PHGDH upregulation is
   strongly associated with an M2-like phenotype in macrophages [[152]25].

   Monocytes are a source of macrophages in cancer, and macrophages are
   known to possess several pro- and anti-tumour functions. Given this and
   the identified differential regulation of proteins involved in
   macrophage polarisation, we explored the effect of EVs from
   LMTK3-overexpressing cells on macrophage polarisation. To assess if the
   EVs were promoting polarisation of the monocytes to either M1 or M2
   macrophages, the proteomics data of monocytes treated with EVs was
   compared with a database of M1 and M2-related protein signatures in
   macrophages (Fig. [153]4i) [[154]49]. M1 and M2 polarisation represent
   the extremes of the macrophage polarisation spectrum, hence macrophages
   can acquire characteristics of both M1 and M2 macrophages.
   Nevertheless, this analysis showed that monocytes treated with LMTK3
   EVs are more enriched in M2-related proteins than monocytes treated
   with control EVs. To confirm that this translates into the promotion of
   M2 polarisation in macrophages, THP-1 monocytes were treated with PMA,
   triggering macrophage differentiation, then subsequently treated with
   EVs. RT-qPCR was then used to assess M1 and M2 markers in the treated
   macrophages (Fig. [155]4j). LMTK3-derived EVs promoted an M2 phenotype
   in macrophages with upregulated CD206 and downregulated CD86. Control
   EVs also led to an increase in CD206, however NOS2 was also upregulated
   and TGFB1 was downregulated. Furthermore, treating LMTK3 overexpressing
   breast cancer cells with C28 prior to EV collection resulted in a
   partial abrogation of the M2-promoting ability of the EVs and a full
   reversal of the inhibition of M1-like gene expression in macrophages
   (Figure S6f).

   Collectively, our data reveal a link between LMTK3 expression and M2
   polarisation in macrophages and suggest a possible role for LMTK3 in
   immunosuppression and evasion of anti-tumour immunity in BC.

LMTK3 EVs reduce monocyte infiltration in tumour spheroids through
PSAT1-PHGDH signalling

   Amongst the validated and most upregulated hits upon LMTK3 EV
   treatment, was phosphoglycerate dehydrogenase (PHGDH), which has been
   previously described as a metabolic checkpoint of macrophage M2
   polarisation [[156]25]. PHGDH plays an important role in serine
   biosynthesis which is a key factor in macrophage polarisation
   [[157]64]. Other studies have shown that an M2-like phenotype in
   macrophages is characterised by a reduction in adhesiveness which may
   explain the reduced rates of infiltration in the spheroid assay
   [[158]65].

   We validated the previous proteomics findings through western blotting
   to show PHGDH upregulation occurs in THP-1 monocytes following
   treatment with LMTK3 EVs but not control EVs (Fig. [159]5a-b). To
   investigate the potential involvement of PHGDH in monocyte
   infiltration, we initially established a THP-1 cell line stably
   overexpressing PHGDH (Fig. [160]5c) and examined its ability to
   infiltrate tumour spheroids compared to parental THP-1 cells.
   Consistent with our hypothesis, PHGDH overexpression caused a
   significant reduction in monocyte infiltration into breast tumour
   spheroids (Fig. [161]5d-e), indicating that the effect of LMTK3-derived
   EVs on monocyte infiltration can be attributed to PHGDH-mediated
   signalling. Next, to explore if PHGDH upregulation is sufficient to
   drive expression of M2-related genes, we performed RT-qPCR analysis on
   macrophages derived from parental and PHGDH overexpressing monocytes.
   Interestingly, PHGDH upregulation in monocytes was able to drive CD206
   upregulation and reduce CD86 expression consistent with an
   immunosuppressive phenotype and in agreement with previous studies
   [[162]28].

Fig. 5.

   [163]Fig. 5
   [164]Open in a new tab

   LMTK3 EVs promote PSAT1 packaging in EVs, causing an upregulation of
   PHGDH in monocytes. a. Western blot of monocytes treated with PBS,
   control EVs or LMTK3 EVs. PHGDH was confirmed as upregulated in
   monocytes treated with LMTK3 EVs. GAPDH is used as a loading control (n
   = 3). b. Quantification of (a). c. Western blot showing PHGDH
   overexpression in THP-1 monocytes compared to control monocytes (n =
   3). d. T47D tumour spheroid infiltration assay using control and PHGDH
   overexpressing monocytes. Monocytes were stained with a green
   lipophilic tracer. Representative images are shown. e. Quantification
   of (d). Graphs show the mean ± SEM of three independent experiments (*p
   < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; as measured by
   two-tailed unpaired Student’s t-test). f. RT-qPCR of PHGDH
   overexpressing monocytes differentiated into macrophages showing
   upregulation of CD206 and downregulation of CD86 compared to control
   cells. g. A plasmid containing doxycycline-inducible PSAT1 was
   transfected into MDA-MB-231 and a stable cell line was generated. EVs
   were collected from the cells with and without doxycycline (24 h). The
   figure shows a western blot for PSAT1 in the cell lysates and EVs for
   PSAT1. CD81 and CD9 are used as positive EV markers and an EV loading
   control. GM130 is used as a negative EV marker and cell lysate loading
   control (n = 3). h. THP-1 monocytes were treated with EVs from cells
   with Dox-inducible PSAT1 overexpression. EVs from MDA-MB-231 cells
   without doxycycline were used as a control treatment. Vinculin was used
   as a loading control. Figure is representative of three biological
   repeats (n = 3). i. Quantification of (h). j. T47D tumour spheroid
   infiltration assay was carried out using monocytes treated with either
   PBS, EVs from Dox-inducible PSAT1 MDA-MB-231 without doxycycline or EVs
   from Dox-inducible PSAT1 MDA-MB-231 with doxycycline. Figure shows
   representative images from the spheroid infiltration experiment using
   monocytes treated with EVs from PSAT1 overexpressing cells with and
   without doxycycline, or PBS. k. Quantification of (i). Graphs show the
   mean ± SEM of three independent experiments (*p < 0.05, **p < 0.01,
   ***p < 0.001, ****p < 0.0001; one-way ANOVA, followed by Tukey’s
   multiple comparisons test)

   Given the downstream serine biosynthesis enzyme PSAT1 was one of the
   top upregulated proteins in LMTK3 EVs, we hypothesised that transfer of
   PSAT1 from LMTK3 BC cells to monocytes alters the balance of the
   intermediary serine biosynthesis metabolites (3-phosphohydroxypyruvate
   and 3-phosphoserine), triggering upregulation of PHGDH. To test this,
   PSAT1 was stably overexpressed in BC cells using a
   doxycycline-inducible overexpression vector, and the cells were
   expanded for EV collection. The cells were then treated with
   doxycycline and western blotting was used to confirm overexpression of
   PSAT1 in the cells and EVs (Fig. [165]5g). We next explored the ability
   of PSAT1-overexpressing EVs to promote PHGDH upregulation in monocytes.
   Conditioned media was collected from the PSAT1 overexpressing BC cells
   with and without the addition of doxycycline, and the EVs were
   separated and used to treat THP-1 monocytes. Western blotting analysis
   confirmed that PSAT1 abundance in EVs can significantly increase PHGDH
   protein levels in monocytes (Fig. [166]5h-i). Moreover, we assessed the
   potential of EVs with upregulated PSAT1 to decrease tumour spheroid
   infiltration. As anticipated, the spheroid infiltration assay showed
   that BC EVs with upregulated PSAT1 levels can reduce monocyte
   infiltration in spheroids (Fig. [167]5j-k), to a comparable degree to
   that induced via treatment with LMTK3-derived EVs, resulting in a
   similar phenotype.

   Overall, these data show that BC cells-derived EVs containing PSAT1 are
   sufficient to induce paracrine upregulation of PHGDH in monocytes,
   which ultimately leads to reduced infiltration of monocytes in tumour
   spheroids.

LMTK3 promotes an immunosuppressive phenotype in macrophages in vivo

   Following the in vitro studies demonstrating that LMTK3-dependent EV
   cargo promotes M2-like polarisation in macrophages and exclusion of
   monocytes in breast tumour spheroids, we sought to investigate the
   effects of LMTK3 overexpression in an in vivo syngeneic model, using
   the immunodeficient mouse strain Rag1^−/− C57Bl/6. These mice are
   unable to produce mature adaptive immune cells, including T cells due
   to impaired V(D)J recombination, inhibiting T and B cell development,
   however, innate immune cells are unaffected and develop normally,
   preserving monocyte and macrophage function. We initially established a
   mouse mammary carcinoma cell line (MMTV-Neu) stably overexpressing
   LMTK3 (Fig. [168]6a-b) and next, EVs were collected from these cells
   and compared to the control cells to assess LMTK3-dependent changes in
   EV protein cargo. Based on our aforementioned results and proposed
   mechanism, we examined the expression of PSAT1 in MMTV-Neu cells and
   confirmed its elevated packaging into EVs following LMTK3 upregulation
   (Fig. [169]6c).

Fig. 6.

   [170]Fig. 6
   [171]Open in a new tab

   LMTK3 alters monocyte behaviour and macrophage polarisation in vivo. a.
   Schematic diagram showing the experimental procedure where mice were
   injected with breast cancer cells that were allowed to form tumours.
   Tumours were then measured and analysed by flow cytometry and IHC. b.
   Western blot showing LMTK3 overexpression in MMTV-Neu mouse mammary
   carcinoma cells. GAPDH is shown as a loading control (n = 3). c.
   Western blot showing overexpression of LMTK3 in MMTV-Neu cells results
   in upregulation of PSAT1 in EVs. HSP70 was used as a loading control (n
   = 3). d. Tumours were measured twice per week with callipers. Tumour
   volume is displayed in mm^3 (two-way ANOVA with Bonferroni’s multiple
   comparisons test). e. Tumour weights on day 32. Graphs show the mean
   ± SEM *p < 0.05, **p < 0.01, ***p < 0.001; two-tailed unpaired
   Student’s t-test. n = 8 tumours (control) and n = 9 tumours (LMTK3). f.
   IHC analysis of ki-67 positive nuclei. Representative images are shown.
   Scale bar = 100 µm. g. Quantification of (f). Graph shows the mean
   ± SEM as measured by two-tailed unpaired Student’s t-test. h. Tumours
   were dissociated and stained for CD45, CD11b and F4/80. Triple-positive
   cells were counted and normalised to tumour weights. Graph shows number
   of macrophages per gram of tumour. Graph shows the mean ± SEM;
   two-tailed unpaired Student’s t-test. n = 8 control and 9 LMTK3
   tumours. i. Representative flow cytometry of cells indicating M1 and
   M2-like macrophages in dissociated tumours. j,k. Flow cytometry showing
   percentage of M1- (j) and M2-like (k) macrophages in tumours. Graphs
   show the mean ± SEM *p < 0.05, **p < 0.01, ***p < 0.001; two-tailed
   unpaired Student’s t-test

   Parental and LMTK3-overexpressing MMTV-Neu cells were then injected
   into Rag1^−/− C57Bl/6 mice and allowed to form tumours in both flanks
   of the mice and the tumours were measured twice per week. Consistent
   with previous studies [[172]1, [173]5, [174]7, [175]8], overexpression
   of LMTK3 in the MMTV-Neu cells resulted larger tumours
   (Fig. [176]6d-e). Furthermore, the LMTK3 tumours displayed higher
   levels of ki-67 compared to control tumours, providing additional
   evidence that LMTK3 promotes proliferation of BC cells
   (Fig. [177]6f-g).

   Following dissociation of the tumours, we assessed the infiltration of
   innate immune cells (monocytes and monocyte-derived macrophages) into
   the tumours via flow cytometry analysis of CD45, CD11b and F4/80
   positive cells (monocytes/macrophages). Our data showed that LMTK3
   overexpressing tumours have a reduced level of monocyte infiltration in
   vivo, in agreement with the clinical analysis of human cancers and our
   in vitro data demonstrating that LMTK3 EVs exclude monocytes from
   tumour spheroids (Fig. [178]6h). In addition, these tumours had
   increased levels of M2-like macrophages and decreased levels of M1-like
   macrophages compared to control tumours (Fig. [179]6i-k). These data
   further validate the tumour promoting role of macrophage PHGDH, as
   previously described [[180]25, [181]28] and a novel tumour promoting
   role of LMTK3 by manipulating the TME and favouring immunosuppression
   through EV-dependent upregulation of PHGDH in monocytes. These results
   were also replicated in a separate syngeneic female mouse cohort using
   orthotopic injection of tumour cells, confirming the observed
   immunosuppressive and growth promoting effects of LMTK3 overexpression
   (Figure S7).

LMTK3 EVs directly contribute to immunosuppressive macrophage phenotype in
vivo

   To assess the contribution of LMTK3 EVs specifically to tumour growth
   and macrophage polarisation, we conducted an in vivo model where 4T1
   cells were used to form tumours in BALB/c mice (Fig. [182]7a). The mice
   were subsequently treated with intratumoural injections of T47D- or
   MDA-MB-231-derived EVs from either control or LMTK3 overexpressing
   cells. Tumours were measured three times per week and were used for
   flow cytometry to assess infiltrating macrophage polarisation.

Fig. 7.

   [183]Fig. 7
   [184]Open in a new tab

   LMTK3 EVs increase tumour growth and promote an immunosuppressive
   phenotype in infiltrating macrophages. a. Schematic diagram showing
   mice were injected with 4T1 cells to form tumours. Mice were then
   randomised and treated with 50 μg of EVs three times per week. Tumours
   were analysed with flow cytometry. b,c. Tumour growth curves showing
   tumour volume in mm^3 for mice treated with T47D EVs (b) and MDA-MB-231
   EVs (c) (two-way ANOVA with Bonferroni’s multiple comparisons test).
   d,e. Graphs show the final tumour weights (grams) for 4T1 mouse models
   treated with T47D (d) or MDA-MB-231 (e) EVs. Graphs show mean ± SEM
   *p < 0.05, **p < 0.01; two-tailed unpaired Student’s t-test. n = 8
   tumours per condition. f,g. IHC analysis of 4T1 tumours for ki-67
   positive nuclei. Representative images are shown. Scale bar = 50 µm.
   h,i. Quantification of (f,g). Graph shows the mean ± SEM, *p < 0.05 as
   measured by two-tailed unpaired Student’s t-test. j-m. Flow cytometry
   showing percentage of M1- (j,k) and M2-like (l,m) macrophages in
   tumours. Graphs show the mean ± SEM *p < 0.05; two-tailed unpaired
   Student’s t-test. n. Proposed model showing LMTK3 overexpression in
   breast cancer cells results in increased serine 72 phosphorylation of
   Rab7. This alters MVB trafficking, resulting in increased packaging of
   PSAT1 into EVs. These EVs are taken up by monocytes, causing an
   upregulation of PHGDH, resulting in an M2-like phenotype in
   monocyte-derived macrophages and reduced infiltration

   Our results showed that LMTK3 EVs from both cell lines (T47D and
   MDA-MB-231) increase tumour growth (Fig. [185]7b-i), albeit to a lesser
   extent, as expected, than LMTK3 overexpression does in our previous
   model (Fig. [186]6d). This is consistent with the various additional
   oncogenic roles of LMTK3 apart from regulating EV biogenesis.
   Furthermore, LMTK3 EVs were sufficient to decrease M1-like gene
   expression (MHC-II) and increase M2-like gene expression (CD206) in
   tumour-infiltrating macrophages (Fig. [187]7j-m).

   Taken together, these data show a direct link between LMTK3-dependent
   EVs and tumour growth and demonstrate that LMTK3 EVs promote an
   immunosuppressive M2-like phenotype in macrophages (Fig. [188]7n).

Discussion

   In recent years, EVs have gained considerable interest for their role
   in cellular communication, particularly in cancer [[189]41, [190]66].
   EV-mediated transport of proteins, lipids, and nucleic acids enables
   cancer cells to influence numerous physiological processes in a variety
   of cell types across distances. Given the strong involvement of EVs in
   cancer biology, understanding the mechanisms governing their
   biogenesis, cargo selection, release and uptake has become an important
   goal in cancer research. Despite numerous advances in knowledge of how
   EV biogenesis occurs, many questions remain, particularly regarding
   selective loading mechanisms of EVs and how these processes could be
   manipulated for therapeutic purposes. As EVs are increasingly
   recognised as key regulators of oncogenic features in the TME,
   investigating regulatory mechanisms governing their biogenesis and
   specific pathways influencing their size, cargo and subpopulations is
   critical.

   Among the many impacts of EVs, one of the most significant may be their
   ability to facilitate tumour progression through manipulation of the
   TME. A wealth of research has shown that EVs play a central role in TME
   modulation through the transfer of bioactive cargo. EVs have been
   implicated in several different oncogenic processes within the TME,
   including cancer-associated fibroblast induction, changes in
   vasculature, and metastasis. Furthermore, EVs have several
   immune-regulating functions where they suppress anti-cancer immunity,
   promote immune tolerance and support cancer escape from immune
   surveillance. Recent studies have shown that EVs can impact the
   composition of tumours, recruiting pro-tumourigenic cells and excluding
   or suppressing cytotoxic anti-tumour cells [[191]66]. In addition,
   several studies have also explored the link between EVs and macrophage
   polarisation in cancer [[192]67]. Given the recent advancements in
   immunotherapies, including checkpoint inhibitors and CAR T-cell
   therapies, it is increasingly important to further understand the
   factors governing immune regulation in cancer.

   LMTK3 is a well-established oncogenic protein kinase with known roles
   in promoting BC growth, metastasis and therapy resistance
   [[193]3–[194]5, [195]7, [196]8]. However, its involvement in EV
   biogenesis and its contribution in the TME have not previously been
   explored. Preliminary evidence suggests LMTK3 may have a role in
   endosomal trafficking, however, this is limited to studies in normal
   physiology with much less evidence shedding light on its roles in
   trafficking and immune regulation in cancer [[197]8]. Considering that
   previous studies have mainly linked the upregulation of LMTK3 protein
   levels with cancer development (instead of LMTK3 mutations, which are
   in general very rare in different malignancies) [[198]3–[199]5, [200]7,
   [201]8], we explored the impact of LMTK3 upregulation in this study in
   order to assess its role in EV biogenesis.

   The current study is the first to demonstrate that LMTK3 upregulation
   in cells impacts EV size, protein cargo and subpopulation distribution,
   revealing a novel mechanism whereby LMTK3 modulates EV biogenesis
   through Rab7 regulation. Furthermore, we present evidence that links
   LMTK3 expression with an immunosuppressive TME. Overexpression of LMTK3
   in a panel of BC cell lines led to a significant increase in the
   average size and subpopulation distribution of EVs, highlighting its
   role in a fundamental aspect of EV biogenesis. As EV size is linked to
   properties such as release, internalisation, diffusion and clearance,
   this alteration may have functional consequences in cancer. For
   instance, these larger EVs may exhibit different rates of
   internalisation or biodistribution as EV size is known to alter these
   properties; however, this requires further exploration [[202]68,
   [203]69]. Moreover, LMTK3-altered proteomic cargo of EVs, including
   proteins involved in amino acid transport, can influence the
   intercellular communications within the TME. Interestingly, the
   identification of LMTK3 as a cargo protein inside the EVs suggests
   another mechanism by which LMTK3 may propagate its oncogenic effects to
   nearby cells. Additionally, LMTK3-mediated downregulation of
   CD63-positive EV particles, which are traditionally associated with EVs
   generated through the tetraspanin-dependent exosome biogenesis pathway
   [[204]70], suggests that LMTK3 mainly influences the biogenesis of
   exosomes rather than other subtypes of EVs.

   The subcellular localisation of LMTK3 to the late endosomal membrane
   further suggests that this altered biogenesis involves the exosomal
   pathway of EV biogenesis. Rab7 is known to regulate MVB trafficking and
   phosphorylation of Rab7 at Ser72 is linked to changes in GTPase
   activity and decreased association with the Rab7 effector, RILP
   [[205]71]. Rab7 has a complex role in vesicle trafficking. One study
   showed that Rab7 positively regulates EV secretion [[206]72], whereas
   others have shown the opposite effect [[207]71]. Here, we identified
   Rab7 as a novel phosphorylation substrate of LMTK3 and, via gain- and
   loss- of function experiments we demonstrated that LMTK3-mediated
   phosphorylation of Rab7 at Ser72 is sufficient to cause the increase in
   EV size and altered protein cargo. In addition to mediating
   intercellular communication, EVs are thought to be an energetically
   favourable mechanism allowing cells to expel waste, particularly for
   proteins that are hard to degrade. This regulation of Rab7 may
   therefore represent a switch from lysosomal degradation to expulsion of
   waste through EVs, however, further investigation is needed into this.

   Our findings have important implications for understanding BC
   progression and EV-mediated intercellular communication. The ability of
   LMTK3 to alter EV size, cargo, and subpopulation distribution suggests
   a critical role in modifying the TME and promoting tumour progression.
   As EVs are known to alter many processes, including angiogenesis,
   invasion and immune evasion, LMTK3 may enhance these pro-tumourigenic
   interactions, promoting the progression of BC through a novel process.
   Targeting LMTK3 or its downstream effector Rab7 may therefore reduce
   the oncogenic potential of BC EVs, impairing the ability of the TME to
   promote growth. Therefore, the subsequent exploration of the role of
   LMTK3-dependent EVs on the TME was essential.

   Deconvolution of RNA sequencing data from BC patients identified
   monocytes being downregulated in LMTK3 overexpressed BCs. It is unclear
   why monocytes are excluded from the tumours, however, in early BC, the
   microenvironment is pro-inflammatory, and cancers must evade immune
   detection to survive [[208]72, [209]73]. The exclusion of monocytes may
   represent a strategy where the cancer cells delay the infiltration of
   myeloid cells until a pro-tumourigenic microenvironment is
   well-established. Still, this is yet to be determined and analysis of
   clinical data from BC patients at different stages would provide
   further information on this matter. Additionally, M2 macrophage markers
   were increased and M1 macrophage markers were decreased in LMTK3
   upregulated BC patients. In accordance with the many tumour-promoting
   roles of LMTK3, the polarisation of macrophages to an M2-like phenotype
   may also contribute to the poorer survival in patients with higher
   levels of LMTK3. In addition to monocytes, other immune cell
   populations were also affected. Interestingly, some populations of T
   cells were downregulated in tumours expressing high LMTK3 levels, as it
   is expected in response to an increase of immunosuppressive
   macrophages. This finding further supports the hypothesis that LMTK3
   upregulation contributes to an immunosuppressive microenvironment.

   Our in vitro work showed that LMTK3 EVs are not taken up at a different
   rate compared to control EVs and they do not impact THP-1 monocytes
   proliferation or viability. This shows that the effects observed are
   specific to changes in monocyte function, rather than altered uptake,
   cytotoxicity or effects on proliferation. LMTK3 EVs reduced monocyte
   infiltration into tumour spheroids, mirroring the clinical data and
   suggesting that LMTK3-regulated EVs are responsible to an extent for
   this effect, instead of cytokines or other cell–cell signalling
   mechanisms. To further explore this in our models, we analysed the
   proteomic content of monocytes and revealed that LMTK3 EVs can
   differentially affect monocyte protein expression. Notably, PHGDH, a
   key protein involved in serine biosynthesis and macrophage M2
   polarisation [[210]25, [211]28], was amongst the main hits upregulated
   in monocytes treated with LMTK3 EVs. We showed that PHGDH
   overexpression is sufficient for reducing their infiltration into
   tumour spheroids, showing that the effects of LMTK3-derived EVs can be
   attributable PHGDH signalling. Furthermore, we provide evidence that
   LMTK3 EVs drive M2 polarisation in monocyte-derived macrophages. M2
   macrophages perform a wide range of pro-tumourigenic functions,
   promoting tumour growth, metastasis and drug resistance [[212]57]. We
   have also shown that LMTK3 overexpression in breast cancer causes the
   release of EVs that promote M2 polarisation in macrophages lending
   further evidence to the tumour-promoting role for LMTK3 in breast
   cancer.

   The ability of LMTK3 EVs to increase the rate of tumour growth and
   promote M2-like macrophage polarisation suggests that LMTK3-dependent
   changes in EV protein cargo contribute, at least partly, to these
   pro-tumourigenic effects. As anticipated, the magnitude of the effects
   was stronger in the mouse model comparing control and LMTK3
   overexpressing cells than in the EV mouse model. This difference may be
   attributable to several factors, including differences in the cancer
   cell lines used, differences in the mice or the dose of EVs, as
   continuous EV secretion by tumour cells cannot be perfectly replicated
   by intermittent intratumoural injections. Nevertheless, both models
   clearly demonstrate that LMTK3-modulated EVs have both pro-tumourigenic
   and immunomodulatory properties.

   Examining the upstream regulation of PHGDH upregulation in monocytes
   revealed that PSAT1 is an important EV cargo protein that promotes
   PHGDH upregulation in monocytes and subsequent decrease in monocytes
   infiltration into tumours. This effect could be mediated via increased
   flux through the serine biosynthesis pathway, thereby increasing
   metabolic states in the cells. As PSAT1 and PHGDH are often
   co-upregulated in cells during surges in metabolic demand, it is
   possible that the increase in EV PSAT1 and transfer to monocytes is
   sufficient to induce PHGDH upregulation, triggering the observed
   changes in the phenotype. In addition to PHGDH upregulation, S100A10
   was also downregulated in monocytes treated with LMTK3 EVs. S100A10 is
   a known regulator of macrophage invasion, therefore it may also play a
   role in the process [[213]63]. Whilst clarifying the exact mechanism of
   action was beyond the scope of the present study, future work could
   ultimately serve to elucidate this further. Previous studies have also
   implicated other hits in macrophage polarisation, including TRPC3 and
   UCHL1, which were among the top hits altered in monocytes treated with
   LMTK3 EVs, therefore, additional changes may also contribute to the M2
   phenotype observed [[214]61, [215]62].

   The in vivo mouse models also supported the in vitro data,
   demonstrating that LMTK3 upregulation in BC is tumourigenic.
   Furthermore, the flow cytometry results showed that M2-like macrophages
   are upregulated, and M1-like macrophages are downregulated in LMTK3
   overexpressing tumours. Future work using cell type-specific deletions
   of PSAT1 and PHGDH would be useful to provide further evidence to this
   mechanism.

   Overall, our experiments support that LMTK3 EVs are pro-tumourigenic in
   the BC immune TME. The alterations in EV protein cargo appear to be
   crucial in modulating the immune TME through regulation of monocytes
   and macrophage polarisation, promoting a pro-tumourigenic environment.
   Further research into their action on other cell types in the TME will
   be crucial to gaining a full understanding of the LMTK3-dependent
   changes in the TME and how this could be used to inform clinical
   decisions.

Supplementary Information

   [216]Supplementary Material 1.^ (5.6MB, zip)

Acknowledgements