Abstract

   Myeloproliferative neoplasms (MPNs) are challenging to treat due to the
   complex bone marrow (BM) microenvironment and lack of curative
   therapies. Current treatments fail to eliminate malignant clones and
   face issues like drug resistance. This study addressed these challenges
   by identifying USP5 as a critical regulator in JAK2^V617F-mutated
   mesenchymal stem cells (MSCs), which promotes proliferation by
   suppressing Caspase-3-mediated apoptosis. We developed engineered
   exosomes (USP5@Exosome-CP) co-expressing CXCR4 and a
   P-selectin-targeting peptide to enhance BM targeting. These exosomes,
   loaded with the USP5 inhibitor USP5-IN-1, demonstrated efficient BM
   homing and sustained drug release. In MPN mouse models, USP5@Exosome-CP
   significantly reduced MSC proliferation, extended survival, and showed
   minimal systemic toxicity. Transcriptomic analysis revealed that USP5
   knockdown activated apoptosis pathways and suppressed oncogenic
   signaling. Our results establish USP5 as a therapeutic target and
   validate the engineered exosome platform as a promising strategy for
   MPN treatment, offering a blueprint for targeting other hematologic
   malignancies. This approach combines USP5 inhibition with BM-targeted
   nanotechnology, providing a proof-of-concept for personalized MPN
   therapy with improved efficacy and reduced off-target effects.

Supplementary Information

   The online version contains supplementary material available at
   10.1186/s12951-025-03588-4.

   Keywords: Myeloproliferative neoplasms, Mesenchymal stem cells,
   Deubiquitinating enzyme, Bone marrow targeting

Introduction

   Myeloproliferative neoplasms (MPN) are a group of clonal hematopoietic
   stem cell disorders characterized by the overproduction of mature
   myeloid blood cells, including red blood cells, white blood cells, and
   platelets [[38]1, [39]2]. These disorders, which include polycythemia
   vera (PV), essential thrombocythemia (ET), and primary myelofibrosis
   (PMF), are driven by acquired somatic mutations in genes such as JAK2,
   CALR, or MPL, leading to hyperactivation of the JAK/STAT signaling
   pathway [[40]3, [41]4]. The JAK2^V617F mutation, in particular, is a
   prevalent driver mutation in Philadelphia chromosome-negative MPN (Ph^−
   MPN) and significantly accelerates disease progression by promoting
   clonal expansion of hematopoietic stem cells [[42]5, [43]6].

   Current treatment options for MPN are primarily palliative, aimed at
   managing symptoms and reducing thrombotic risks rather than curing the
   disease [[44]7]. Hydroxyurea, an oral chemotherapy agent, is widely
   used to reduce blood cell counts and alleviate symptoms [[45]8].
   Ruxolitinib, a JAK inhibitor, suppresses hyperactive JAK/STAT signaling
   and is particularly beneficial for patients with PV and ET [[46]9,
   [47]10]. However, these therapies fail to eliminate malignant clones,
   often lead to dose-limiting toxicities, and face challenges such as
   drug resistance and disease relapse [[48]11]. Moreover, the chronic
   nature of MPN and the complexity of its bone marrow (BM)
   microenvironment further complicate therapeutic interventions.

   The BM microenvironment plays a pivotal role in MPN pathogenesis,
   providing a niche that supports the survival and proliferation of
   mutant hematopoietic stem cells [[49]12]. Targeted drug delivery to
   this site remains a major hurdle due to physiological barriers and poor
   drug penetration [[50]13–[51]15]. The CXCR4-CXCL12 (SDF-1) axis, a key
   mediator of cell homing and retention in the BM, is critical for cancer
   cell proliferation, immune evasion, and stromal interactions [[52]16].
   CXCR4 is highly expressed on malignant cells in MPN, while CXCL12 is
   abundantly secreted by BM stromal cells, making this axis a promising
   target for disrupting pathological cell-matrix interactions [[53]17].

   To address these challenges, stem cell-derived exosomes have emerged as
   innovative drug delivery vehicles [[54]18]. Exosomes are natural
   nanoscale extracellular vesicles capable of transporting proteins,
   lipids, and nucleic acids between cells, offering advantages such as
   biocompatibility, low immunogenicity, and the ability to cross
   biological barriers [[55]19, [56]20]. Engineered exosomes can inherit
   targeting motifs from parental cells, enabling precise delivery to
   specific tissues while minimizing off-target effects [[57]21, [58]22].
   Recent studies highlight their potential in modulating inflammatory
   responses, promoting tissue repair, and enhancing drug bioavailability
   in hard-to-reach niches like the BM [[59]23, [60]24].

   To systematically identify JAK2^V617F-specific vulnerabilities within
   the ubiquitin-proteasome system, we employed comprehensive functional
   screening of 86 deubiquitinating enzymes (DUBs) in mutant versus normal
   mesenchymal stem cells (MSCs). This unbiased approach was critical
   because: (i) canonical DUBs (e.g., USP9X/USP15) show disease-context
   dysregulation not extrapolatable to JAK2^V617F MPN; (ii) the mutation
   remodels ubiquitin signaling in a target-specific manner, necessitating
   empirical discovery; and (iii) it circumvents presupposition bias to
   reveal novel targets with mechanistic and therapeutic uniqueness.
   Through transcriptomic and proteomic analyses, we discovered that USP5
   is markedly overexpressed in mutant MSCs (Mut-MSCs), where it
   suppresses Caspase-3-mediated apoptosis and drives pathological
   proliferation. Knockdown of USP5 or pharmacological inhibition using
   USP5-IN-1 restored apoptosis, normalized cell cycle progression, and
   significantly attenuated Mut-MSC proliferation in vitro and in vivo.
   These findings establish USP5 as a critical therapeutic target in MPN.

   To overcome the limitations of systemic drug delivery, we engineered
   MSC-derived exosomes (USP5@Exosome-CP) co-functionalized with CXCR4 and
   a P-selectin-targeting peptide (PSN). The dual-targeting design
   leverages the upregulated CXCL12 secretion in Mut-MSCs and the
   inflammatory overexpression of P-selectin on BM endothelial cells
   (BMEC) in MPN models. This strategy ensures efficient transendothelial
   transport and selective accumulation of exosomes in the BM niche.
   Loaded with USP5-IN-1, the engineered exosomes demonstrated sustained
   drug release, enhanced Mut-MSC targeting, and potent induction of
   apoptosis, culminating in prolonged survival of MPN model mice (Scheme
   [61]1). By integrating mechanistic insights into USP5-driven MPN
   progression with advanced nanocarrier engineering, this study provides
   a transformative therapeutic platform. Our approach not only validates
   USP5 as a biomarker and target but also pioneers the use of biomimetic
   exosomes for precision therapy in hematological malignancies, offering
   a blueprint for future applications in oncolog.

Scheme 1.

   [62]Scheme 1
   [63]Open in a new tab

   This study achieved MSCs-derived exosomes carrying CXCR4 proteins by
   means of genetic engineering, as well as adsorption of peptides
   targeting P-selectin through ligand-receptor-associated effects and
   loading of USP-5-IN-1, a drug that induces apoptosis in Mut-MSCs, via
   an electroporation strategy. The nanomedicine efficiently traversed the
   BMECs, targeted Mut-MSCs, and mediated their effective apoptosis, thus
   realizing the treatment of MPN model

Results

Screening of deubiquitinating enzymes for regulating aberrant proliferation
of MSCs

   Overproliferation of hematopoietic stem cells leads to rapid
   progression of MPN, concurrently shortening the therapeutic window for
   patients [[64]25]. Curtailing the clonal proliferation of hematopoietic
   stem cells in MPN and effectively controlling the disease’s progression
   are prerequisites for facilitating effective treatments. This study
   delves into the proliferation characteristics of MSCs carrying the
   JAK2^V617F mutation (Mut-MSCs), aiming to uncover deubiquitinating
   enzymes associated with the abnormal proliferation of these cells.
   Figure [65]1A illustrates the experimental workflow for screening
   deubiquitinating enzymes that are specifically overexpressed in
   Mut-MSCs with abnormal proliferation at the transcriptome level. We
   initially isolated MSCs from the BM of normal and JAK2^V617F
   mutation-bearing mice using a CD117 magnetic bead selection kit, and
   MSCs or Mut-MSCs occupied more than 95% of the isolated cells (Figure
   [66]S1, Supporting information). Total mRNA was extracted, and
   complementary DNA (cDNA) was synthesized from normal MSCs or Mut-MSCs.
   Subsequently, we utilized our self-constructed primer library,
   encompassing 86 deubiquitinating enzymes (Supporting information Table
   [67]1), to identify enzymes with differential expression between normal
   MSCs and Mut-MSCs. Quantitative PCR QPCR results presented in
   Fig. [68]1B revealed that USP5 and USP8 were most highly expressed in
   Mut-MSCs compared to normal MSCs, while USP13, USP42, and USP51 were
   significantly downregulated. Considering the potential for
   post-translational modifications, we employed western blotting
   experiment to compare the expression of USP5, USP8, USP13, USP42, and
   USP51 in normal and Mut-MSCs. The results in Fig. [69]1C confirmed that
   the mRNA expression levels of differentially expressed deubiquitinating
   enzymes in normal and Mut-MSCs were consistent with their protein
   expression levels.

Fig. 1.

   [70]Fig. 1
   [71]Open in a new tab

   Identification of deubiquitinating enzymes for regulating the
   proliferative properties of MSCs obtained from MPN mouse model. (A)
   Schematic diagram of the experimental procedure for obtaining cDNA from
   mouse BM hematopoietic stem cells. (B) Differential expression
   statistics of the deubiquitinating enzyme family at the mRNA level in
   normal MSCs and in MSCs sorted from the BM of mice harboring a mutation
   in the JAK2^V617F gene. (C) Western blotting experiments to verify
   differences in protein expression levels of differentially expressed
   deubiquitinating enzymes screened from (B). (D) Western blotting
   experiments to verify the effect of knockdown or overexpression of
   differentially expressed deubiquitinating enzymes on caspase-3
   expression. (E) Flow cytometry assay to verify the effect of
   overexpression or knockdown of differentially expressed
   deubiquitinating enzymes on the proliferative activity of MSCs derived
   from the MPN mouse model. (F) Volcano plot of differential gene
   expression in the PBS group and sgUSP5 group. (G) Heat map of
   clustering analysis of differential gene in sgUSP5 treated MSCs. (H)
   KEGG pathway enrichment analysis in sgUSP5 treated MSCs. (I) Enrichment
   analysis of GSEA gene set in sgUSP5 treated MSCs. (J) Representative
   flow cytometry plots and quantitative analysis of Annexin V positive
   cells after 24 h of different treatment in MSCs. Data are presented as
   the mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, and NS: not
   significant. N = 3 experimental replicates

   Given the correlation between cellular proliferation characteristics
   and the expression of apoptotic proteins, we compared the expression of
   caspase-3, a protein closely associated with apoptosis, to determine
   its correlation with the differentially expressed deubiquitinating
   enzymes. We overexpressed USP13, USP42, and USP51 in Mut-MSCs or
   knocked down the expression of USP5 and USP8, and then assessed
   caspase-3 expression in Mut-MSCs following these treatments. Western
   blotting results in Fig. [72]1D showed that Mut-MSCs expressed lower
   levels of caspase-3 protein compared to normal MSCs. Overexpression of
   USP13, USP42, and USP51 did not affect caspase-3 expression in
   Mut-MSCs, but knocking down USP5 or USP8 significantly enhanced
   Caspase-3 expression to levels comparable to those in normal MSCs.
   Additionally, we verified the proliferation of Mut-MSCs following these
   treatments. Flow cytometry results in Fig. [73]1E confirmed that while
   overexpression of USP13, USP42, and USP51 did not influence caspase-3
   expression, it moderately inhibited the proliferation of Mut-MSCs. In
   contrast, knocking down USP5 or USP8 nearly abrogated MSC
   proliferation. These findings confirm that USP5 or USP8 can
   specifically regulate caspase-3 expression in Mut-MSCs and serve as
   effective targets to inhibit their proliferation. However, since the
   efficacy of USP8 has been reported in other studies [[74]26], this
   research focuses on modulating USP5 activity to achieve effective
   treatment of MPN models.

   To demonstrate the mechanisms of USP5 regulating the proliferation of
   Mut-MSCs, we performed bulk RNA sequencing in Mut-MSCs after treatment
   with or without knock outing of USP5 gene. The volcano plot of
   differentially expressed genes and clustering results were shown in the
   Fig. [75]1F-I. After treatment with knock outing of USP5 gene,
   expression levels of oncogenes such as Idh2 and Slc6a2 were
   significantly downregulated, while levels of tumor suppressor genes
   such as Ptpn2 and Runx3 were significantly upregulated (Fig. [76]1F-G).
   KEGG enrichment analysis showed that pathways related to apoptosis were
   markedly upregulated (Fig. [77]1H). Particularly, GESA pathway analyzes
   revealed that cell differentiation, apoptosis, negative regulation of
   cell cycle and negative regulation of TOR signaling were enriched in
   knock outing of USP5 gene group (Fig. [78]1I), which was in line with
   our flow cytometry results (Fig. [79]1J). We hypothesize that USP5 may
   deubiquitinate and stabilize anti-apoptotic proteins (e.g., XIAP),
   leading to Caspase-3 suppression, though direct targets require future
   investigation [[80]26]. Overall, these data indicate that knockdown of
   USP5 gene in Mut-MSCs induces apoptosis and differentiation and reduces
   cell cycle related and other oncogenic pathways.

   USP5-IN-1 was reported to efficiently inhibit USP5 activity [[81]27],
   therefore, we used this inhibitor to verify the killing effect on
   Mut-MSCs or MSCs. The results of Figure [82]S2 (Supporting information)
   showed that low concentrations of USP5-IN-1 exhibited a higher killing
   effect on Mut-MSCs and a weaker killing effect on MSCs, but if
   knockdown of USP5 expression in Mut-MSCs, it was able to reverse
   Mut-MSCs sensitivity, demonstrating that USP5-IN-1 can specifically
   inhibit the function of USP5 protein. In addition, we also tested the
   effects of USP5-IN-1 or knockdown of USP5 on Mut-MSCs-mediated MPN
   model. To construct the MPN model, we obtained Mut-MSCs from the BM of
   JAK2^V617F mutated mice and constructed luciferase-expressing Mut-MSCs
   by retroviral system, and subsequently injected these Mut-MSCs into
   normal C57 mice that had undergone BM cleaning with X-ray via tail
   vein. The results in Figure [83]S3 (Supporting information) showed that
   USP5 knock down or USP5-IN-1 therapy could effectively inhibit the
   generation of MPN models, whereas Mut-MSCs alone were able to rapidly
   form MPN models. The above results confirm the importance of USP5 in
   mediating the formation of MPN models in MSCs.

Bioinformatic analysis of USP5 expression correlates with disease progression
in acute myeloid leukemia patients (AML)

   To verify whether USP5 is associated with survival and the immune
   microenvironment in MPN patients, we analyzed USP5 expression using a
   database of AML patients associated with the MPN disease model.
   Therefore, we first evaluated the expression of USP5 in AML from mRNA
   level. The results of Fig. [84]2A showed that the mRNA of USP5 was
   highly expressed in AML patient, and the high expression of USP5 was
   present at different stages of AML progression (Fig. [85]2B). This
   demonstrates that the high expression of USP5 is closely related to AML
   disease progression. Meanwhile, ALM patients with high USP5 expression
   had significantly lower survival than AML patients with low EIF2S2
   expression (Fig. [86]2C). In addition, we also simultaneously analyzed
   the correlation of USP5 expression with the expression of
   immunomodulatory molecules and immune cell infiltration (Fig. [87]2D),
   and the results showed that highly USP5 expression promotes high
   expression of relevant immunosuppressive molecules such as PD-L1, PD-1,
   CTLA4, LAG3 and TIGIT in the tumor microenvironment, which in turn
   limits anti-tumor immunity. Furthermore, highly USP5 expression in AML
   patient is also significantly positively correlated with a large
   infiltration of immunosuppressive cells, such as myeloid
   immunosuppressive cells and regulatory T cells (Fig. [88]2E). By
   analyzing the single-cell database of AML (Fig. [89]2F-G), we found
   that USP5 was predominantly expressed in MSCs with high expression of
   CD34 and CD117. These results demonstrated that USP5 gene mainly
   promote the development of AML. Although AML is distinct from MPN, we
   analyzed USP5 in AML due to (1) shared molecular features in bone
   marrow dysfunction, and (2) limited availability of MPN patient
   databases with transcriptomic data. Our findings suggest USP5’s broader
   role in myeloid malignancies, warranting future validation in MPN
   cohorts.

Fig. 2.

   [90]Fig. 2
   [91]Open in a new tab

   Bioinformatic analysis of USP5 expression correlates with disease
   progression in acute myeloid leukemia patients (AML). (A) Analysis of
   USP5 mRNA expression in different cancer species based on TCGA
   database. (B) Analysis of USP5 mRNA expression in AML patients with
   different disease stage. (C) Survival analysis of AML patients with
   high expression of USP5 versus low expression of USP5. (D) Analysis of
   USP5 expression in tumor tissue in relation to the expression of
   different immunoregulatory proteins in TCGA database. (E) Analysis of
   USP5 expression in AML BM in relation to the infiltration of different
   immune cells in TCGA database. (F) Unsupervised clustering algorithm to
   analyse different cell subpopulations in single-cell data from GSEA AML
   patients. (G) Analysis of different cell subpopulations expressing USP5
   in AML patients

Preparation and characterization of USP5@exosome-CP

   Although the genetic silencing of USP5 in Mut-MSCs can efficiently
   inhibit their abnormal proliferation, in vivo gene editing of target
   cells remains challenging. USP5-IN-1 has been reported to potently
   inhibit USP5 activity at low concentrations. However, its hydrophobic
   nature limits its in vivo application and clinical translation.
   Additionally, conventional in vivo drug administration of a single
   agent could not ensure effective BM infiltration and targeted
   regulation of Mut-MSCs. To address these challenges, some studies have
   focused on constructing bone marrow homing nanocarriers for targeting
   the bone marrow microenvironment [[92]13, [93]15], where we propose to
   employ engineered stem cell-derived exosomes (MSCs) loaded with small
   molecule drugs such as USP5-IN-1 to achieve efficient in vivo targeting
   of BM Mut-MSCs.

   To facilitate exosome targeting to BM-derived MSCs (BMSCs), we first
   characterized the BM microenvironment. Exosomes need to cross the
   barrier of BM vascular endothelial cells (BMEC) before targeting BMSCs,
   whereas the exosome itself is weakly characterized for crossing
   vascular endothelial cells. To promote the specific targeting of
   exosome to BMEC, we first analyzed the expression of p-selectin in
   BMEC, as p-selectin has been reported to be expressed predominantly on
   the inflammatory vascular endothelium and most of the immune cells can
   cross the vascular endothelium via P-selectin [[94]24, [95]28].
   Immunohistochemical results (Fig. [96]3A) showed that the expression of
   p-selectin in BM from the MPN model was much higher than that in BM
   from normal mice, and the confocal imaging results (Fig. [97]3B-C) also
   showed that the expression of p-selectin in BMEC from the MPN model was
   much higher than that in BMEC from the normal mouse model. The above
   results confirmed that p-selectin can serve as a good target for
   exosomes to cross the vascular endothelial cells. Moreover, to further
   promote the targeting properties of exosomes to mutant MSCs, we
   analyzed the expression of different chemokines in BMEC from MPN model.
   The results of multi-cytokine assay (Fig. [98]3D) showed that the
   expression of CXCL12 was the highest in the BM of MPN and was
   significantly different from that of normal BM. The results of Elisa
   assay (Fig. [99]3E) further confirmed that the secretion of CXCL12 in
   Mut-MSCs was also significantly higher than that in the MSCs cells, and
   the above results suggested that we could target Mut-MSCs of MPN model
   through the interactions such as CXCR4/CXCL12.

Fig. 3.

   [100]Fig. 3
   [101]Open in a new tab

   Preparation and characterization of USP5@exosome-CP. (A)
   Immunohistochemical assessment of P-selectin expression in bone marrow
   of MPN model mice. Scale bar is 100 μm. (B) Confocal imaging to assess
   bone marrow-derived vascular endothelial cells (BMEC) in
   inflammatory-stimulated high expression of P-selectin. Scale bar is
   10 μm. (C) Flow cytometry to statistic the relative fluorescent
   intensity (RFI) of P-selectin in BMEC or BMEC infused with TNF-α. (D)
   Multi-cytokine detection kit to characterize the differences in the
   expression of different chemokines in the bone marrow microenvironment
   between MPN and normal mice. N = 3 experimental replicates. (E) Elisa
   assay to detect differences in CXCL12 secretion between normal MSCs and
   Mut-MSCs. N = 4 experimental replicates. (F) Schematic of Exosome-CP
   preparation. (G) Statistics of fluorescence intensity of engineered
   exosomes incubated with different mass ratios of FITC-labeled fusion
   peptide CP05-PSN (CP) for 24 h by using flow cytometry (The quality of
   the engineered exosomes remains unchanged, and the optimal amount of CP
   that can be loaded into the engineered exosomes is determined by
   adjusting the mass ratio of CP to engineered exosomes during
   incubation) (N = 3). (H) Statistics on the concentration of engineered
   exosomes loaded with USP5 inhibitor (USP5-IN-1) under different
   electrical stimulation conditions (n = 3). (I) HPLC to be used to
   identify the drug loading characteristics of USP5@Exosome-CP. (J)
   Statistics on zeta potential of different indicated engineered exosomes
   (n = 3), including Exosome, Exosome containing CXCR4 (Exosome-C),
   Exosome cotaining CP05-PSN (Exosome-P), Exosome containing CXCR4 and
   CP05-PSN (Exosome-CP), Exosome containing USP5-IN-1 (USP5@Exosome) and
   Exosome containing USP5-IN-1, CXCR4 and CP05-PSN (USP5@Exosome-CP). (K)
   Statistics on the range of particle size distribution of indicated
   engineered exosomes by using DLS and TEM (scale bar is 200 nm). (L)
   Validation of different types of exosomes expressing universal exosome
   markers or CXCR4/Lamp2b-CXCR4 by western bloting. Data are presented as
   the mean ± SD. *P < 0.05,**P < 0.01, ***P < 0.001, and NS: not
   significant

   To ensure that CXCR4 can be efficiently secreted into exosomes, we
   first express a fusion protein, Lamp2b-CXCR4, in MSCs derived from
   normal mouse BM (Figure [102]S4, Supporting information), which is
   actively secreted into exosomes, thereby equipping MSC-derived exosomes
   with the CXCR4 protein. In addition, we embedded the fusion peptide
   CP05-PSN (peptide sequence is CRHSQMTVTSRLGGGGSDAEWVDVS), which can
   target p-selectin, in the exosomes loaded with CXCR4 protein (CP05 is a
   peptide targeting CD63 protein on exosomes [[103]29], and PSN can
   target p-selectin, and auto-loading of the fusion peptide on exosomes
   can be facilitated by linker linking P-selectin targeting motif to
   CP05, such as GGGGS). The schematic diagram of exosomes extraction was
   in depicted in Fig. [104]3F.

   To verify how much CP05-PSN (CP, a peptide fusion of CP05 and
   P-selectin targeting motif) peptide can be effectively loaded on the
   exosomes outer surface, we incubated different concentrations of CP
   peptide modified with FITC dye with equal protein concentration of
   exosomes (Exosome-CP). The statistics results in Fig. [105]3G by using
   flow cytometry showed that the fluorescence intensity of exosomes was
   saturated when the ratio of the protein concentration of CP peptide to
   exosomes was 2:5. In addition, to determine the stability of CP peptide
   bound to exosomes, we incubated it with 10% mouse plasma for different
   times, and the results of naive SDS-PAGE gels (Figure [106]S5,
   Supporting information) showed that the CP peptide was able to be
   stably maintained on the surface of the Exosome-CP for about 36 h.

   To verify the drug loading ability of Exosome-CP, we also explored the
   parameters of the electroporator for the exosomes to be effectively
   loaded with USP5-IN-1, and the results of Fig. [107]3H showed that when
   the voltage parameter of the electroporator was set to 500 V, and the
   capacitance parameter was 500µF, the exosomes could be loaded with the
   largest amount of USP5-IN-1 molecular. The HPLC results (Fig. [108]3I)
   confirmed that the peak times of the small molecules extracted from
   USP5@Exosome-CP coincided with those of free USP5-IN-1, proving that
   USP5@Exosome-CP indeed contains a large amount of USP5-IN-1, and does
   not change its chemical properties. To verify the stability of
   USP5-IN-1 in USP5@Exosome-CP, we also used HPLC to characterize the
   amount of USP5-IN-1 remaining in USP5@Exosome-CP after incubation with
   10% mouse plasma for different times. The results (Figure [109]S6,
   Supporting information) showed that the half-life (t1/2) of USP5-IN-1
   in USP5@Exosome-CP under physiological conditions (10% mouse serum) was
   approximately 48 h, with ~ 60% remaining after 72 h, demonstrating the
   hydrophobicity of USP5@Exosome-CP-loaded stability of the hydrophobic
   drug. We also measured the zata potential and particle size of exosomes
   processed in various ways and found that the loading of peptide or drug
   did not significantly affect the zata potential of exosomes
   (Fig. [110]3J), whereas the particle size of exosomes loaded with drug
   increased to a certain extent (Fig. [111]3K). The results in Figure
   [112]S7 (Supporting information) showed that the loading of CP peptide
   and USP5-IN-1 agents did not influence the structure of the exosomes.

   Western blot was used to characterize whether CXCR4 is efficiently
   loaded into exosomes. Figure [113]3L demonstrated that all the exosomes
   were rich in exosome-associated proteins such as CD63 and CD81, whose
   expression was not influenced by encapsulation of CP peptide and
   USP5-IN-1 agents, and CXCR4 was also efficiently loaded into engineered
   exosomes. Thus, the exosomes that we extracted had the typical
   characteristics of small extracellular vesicles. In summary, these data
   confirmed the successful construction of a biomimetic nanocarrier with
   optimized drug loading. The simple synthesis strategy used here
   maximizes drug delivery and target motif loading, demonstrating the
   potential of engineered MSCs-derived exosomes for further nano-drug
   development.

Validation of USP5@Exosome-CP targeting inflammatory BM-derived endothelial
cells and Mut-MSCs

   In order to validate the specific targeting capacity of engineered
   exosomes (with single or dual targeting properties) towards
   inflammatory BM-derived endothelial cells (BMEC) and MSCs. Systematic
   evaluation of targeting characteristics was performed using confocal
   microscopy and flow cytometry for various engineered exosomes (Exosome,
   Exosome-CP, Exosome-C, Exosome-P), with a focus on their efficiency in
   targeting normal/inflammatory BMECs and Mut-MSCs. As demonstrated in
   Fig. [114]4A, confocal imaging revealed low uptake efficiency of both
   P-selectin-targeting peptide modified and CXCR4-functionalized exosomes
   by normal BMECs. However, P-selectin modified exosomes (Exosome-P)
   exhibited significantly enhanced endocytosis in inflammatory BMECs.
   Quantitative flow cytometry analysis (Fig. [115]4B) further confirmed
   this observation, highlighting the critical role of
   P-selectin-targeting peptides in mediating exosome transendothelial
   transport across the inflammatory vascular barrier in MPN models. We
   also elucidated the mechanism of CXCR4 modules in targeting
   CXCL12-hypersecreting Mut-MSCs, various engineered exosomes were
   co-cultured with normal MSCs and Mut-MSCs for 24 h. Figure [116]4C-D
   demonstrated that CXCR4-modified exosomes (Exosome-C) significantly
   enhanced targeting efficiency towards MSCs, with particular specificity
   for Mut-MSCs, confirming the pivotal role of the CXCR4/CXCL12 axis in
   Mut-MSC targeting.

Fig. 4.

   [117]Fig. 4
   [118]Open in a new tab

   Characterization of different engineered exosomes targeting
   inflammatory endothelial cells and BMSCs. (A) Confocal imaging
   characterization of different engineered exosomes targeting normal
   BMECs and inflammatory BMECs (infused with TNF-α). Scale bar is 20 μm.
   (B) Flow cytometry characterization of different engineered exosomes
   targeting normal BMEC and inflammatory BMEC (infused with TNF-α). (C)
   Confocal imaging characterization of different engineered exosomes
   targeting normal MSCs and Mut-MSCs. Scale bar is 20 μm. (D) Flow
   cytometry characterization of different engineered exosomes targeting
   normal MSCs and Mut-MSCs. (E) Schematic diagram of the Transwell model
   validating different engineered exosomes across normal BMEC or
   inflammatory BMEC. (F) Flow cytometry statistics on the properties of
   different engineered exosomes targeting normal MSCs and Mut-MSCs after
   crossing normal BMEC. (G) Flow cytometry statistics on the properties
   of different engineered exosomes targeting normal MSCs and Mut-MSCs
   after crossing an inflammatory BMEC. Data are presented as the
   mean ± SD. *P < 0.05,**P < 0.01, ***P < 0.001, and NS: not significant.
   N = 3 experimental replicates

   Building upon these findings, we established a Transwell model
   (Fig. [119]4E) to simulate exosome transendothelial delivery. The upper
   chamber contained normal/inflammatory BMEC monolayers, while the lower
   chamber housed a co-culture system of CFSE-labeled Mut-MSCs and normal
   MSCs. Following 24-hour incubation with DiD-labeled exosomes in the
   upper chamber, flow cytometric analysis of lower chamber cells
   (Fig. [120]4F-G) showed that dual-modified exosomes (Exosome-CP)
   demonstrated superior targeting efficiency towards Mut-MSCs after
   traversing inflammatory BMEC layers, with mean fluorescence intensity
   (MFI) approximately 5-fold higher than those of Exosome-C and
   Exosome-P. This enhanced performance suggests that Exosome-CP’s
   superior penetrability and targeting specificity may originate from the
   selective binding capacity of its surface P-selectin-targeting peptide
   within inflammatory microenvironments. Collectively, Exosome-CP
   exhibits remarkable advantages in targeting inflammatory endothelia and
   Mut-MSCs, providing a novel delivery platform for precision therapy in
   inflammation-related diseases and MPN models. In addition, to confirm
   whether Exosome-CP carrying USP5-IN-1 could efficiently induce
   apoptosis in Mut-MSCs, we used engineered exosomes of different
   designated treatment groups to co-incubate with Mut-MSCs in vitro for
   24 h, and then observed the death ratio of Mut-MSCs in different groups
   by staining PI dye. The flow cytometry statistical results in Figure
   [121]S8 (Supporting information) confirmed that USP5@Exosome-CP was
   able to induce significant death of Mut-MSCs, which was consistent with
   the effect of incubating USP5-IN-1 alone, but much more effective than
   that of Exosome-CP as well as exosomes. The above results confirmed
   that USP5@Exosome-CP had both targeting and therapeutic effect.

Biodistribution of USP5@exosome-CP

   To assess the BM-homing capability of the USP5@Exosome-CP nanosystem,
   different engineered exosomes of equal protein concentration after DiD
   staining were injected via tail vein into MPN model mice carrying the
   JAK2^V617F mutation. The bioluminescence imaging results in
   Fig. [122]5A showed that we successfully constructed a
   luciferase-expressing MPN model. We performed fluorescence imaging of
   mice injected with different specified treatments at different time
   points, and the results in Fig. [123]5B showed that the unengineered
   exosomes themselves had the ability to target the BM, and the
   Exosome-CP carrying CXCR4 and CP peptide was able to further promote
   their accumulation in the BM of mice. In addition, carrying USP5-IN-1
   did not affect the targeting properties of engineered exosomes in BM.
   The fluorescence signal of USP5@Exosome-CP or Exosome-CP appeared at
   3 h and peaked at 12 h post-injection. By contrast, the DiD groups
   underscored the minimal BM-homing capabilities. The fluorescence
   signals of dissected bone tissue also revealed significantly stronger
   fluorescence intensity of USP5@Exosome-CP or Exosome-CP in comparison
   with exosomes or USP5@Exosome, with no detectable fluorescence of the
   DiD groups. Specifically, USP5@Exosome-CP exhibited a 2.2-fold increase
   in fluorescence intensity compared to unmodified Exosome (P < 0.001)
   and a 2.1-fold increase compared to USP5@Exosome (P < 0.01)
   (Fig. [124]5C-D). In addition, fluorescence imaging maps of the major
   organs of mice in the different designated treatment groups showed that
   the engineered or unengineered exosomes accumulated mainly in the liver
   and spleen, with lower levels of accumulation in other organs
   (Fig. [125]5E). To further evaluate the targeting properties of
   USP5@Exosome-CP on in vivo Mut-MSCs, we observed the BM of different
   groups of mice using immunofluorescence. The results in Fig. [126]5F
   showed that Exosome and USP5@Exosome infiltrated into the BM with weak
   fluorescence intensity and co-localized weakly with Mut-MSCs cells
   (c-kit^+), whereas Exosome-CP and USP5@Exosome-CP infiltrated into the
   BM with significantly enhanced ability and had strong fluorescence
   co-localization with Mut-MSCs. The above results confirmed that
   USP5@Exosome-CP could efficiently traverse BMEC in vivo and could
   highly target Mut-MSCs cells in the BM.

Fig. 5.

   [127]Fig. 5
   [128]Open in a new tab

   Biodistribution and BM-targeting ability of USP5@Exosome-CP in vivo.
   (A) Whole body imaging to verify the successful MPN modeling in
   different treatment group. (B) Radiant efficiency-based fluorescence
   imaging of C57BL/6J mice at various time points post-intravenous
   injection of DiD labeled Exosome, DiD labeled USP5@Exosome, DiD labeled
   Exosome-CP and DiD labeled USP5@Exosome-CP. There were N = 3 mice per
   group. The BMs are indicated by blue ovals. (C) Fluorescence images of
   bone tissues from mice injected with indicated treatment. (D)
   Region-of-interest analysis of fluorescent intensities from bone
   tissues. N = 6 (three mice with each mouse per group captured twice).
   (E) The distribution of indicated treatment in various organs. (F)
   Immunofluorescence was used to assess the ability of exosomes from
   different designated treatment groups to target BMSCs in the MPN model.
   DAPI (blue) was used to stain nuclei, C-kit (Green) was used to stain
   MSCs, and DiD was used to trace exosomes. Scale bar is 100 μm and the
   scale bar of the zoom image is 30 μm. The units of the colour bars in
   (A), (B) and (D) are ps^–1 cm^–2 sr^–1/µW cm^–2. Data are presented as
   the mean ± SD. *P < 0.05,**P < 0.01, ***P < 0.001, and NS: not
   significant

Anti-MPN effects of USP5@Exosome-CP

   The therapeutic efficacy of USP5@Exosome-CP was assayed in MPN mice
   model (Fig. [129]6A), and the therapeutic process is depicted in
   Fig. [130]6B. The therapeutic concentration used in in vivo experiments
   was 5 mg/kg USP5-IN-1 (equivalent to 50 mg/kg exosome particles per
   mouse). The bioluminescence signals of the different designated
   treatment groups were obtained on days 10, 20 and 30. The results in
   Fig. [131]6C showed that tail vein injection of exosome or Exosome-CP
   did not demonstrate a therapeutic advantage in the MPN model, whereas
   USP5-IN-1 administered intraperitoneally alone was able to inhibit the
   development of MPN to a certain extent, but the effect was not
   significant, which was mainly related to its effective concentration of
   the drug in the BM, whereas USP5@Exosome was able to inhibit the MPN
   development further in comparison with the administration of USP5-IN-1
   alone development, and USP5@Exosome-CP with targeted homing properties
   demonstrated the highest degree of therapeutic effect on MPN.
   Figure [132]6D also demonstrates that the mice treated with
   USP5@Exosome-CP manifested more potent eradication of Mut-MSCs compared
   with other groups. The survival rate after 80 days in mice receiving
   USP5@Exosome-CP was approximately 30% (Fig. [133]6E). These findings
   indicated that USP5@Exosome-CP may have a positive impact on BM
   function and contribute to the recovery of MPN model. To assess the
   biosafety of USP5@Exosome-CP, H&E staining was used to evaluate the
   potential toxicities in different designated treatment groups.
   Figure [134]6F shows immunohistochemical staining of c-kit^+ cells in
   bone marrow, revealing reduced Mut-MSCs in USP5@Exosome-CP-treated
   mice. Figure [135]6G shows H&E-stained organs without pathological
   changes. These results confirm that USP5@Exosome-CP can serve as a
   superior therapeutic strategy for MPN models.

Fig. 6.

   [136]Fig. 6
   [137]Open in a new tab

   Anti-MPN effects of USP5@exosome-CP in vivo. (A) Schematic of the MPN
   modeling and treatment protocol. There were N = 10 mice per group. The
   therapeutic concentration used in in vivo experiments was 5 mg/kg
   USP5-IN-1 (equivalent to 50 mg/kg exosome per mouse). (B) Schematic
   illustration of the therapeutic process of USP5@exosome-CP. (C)
   Bioluminescence images of mice treated with groups. The units of the
   colour bars are counts. (D) Flow cytometry analysis of MSCs in BM. (E)
   Survival curves of MPN mouse model treated with different formulations.
   the survival statistical significance was analysed via a log-rank
   (Mantel-Cox) test. (F) Immunohistochemical assessment of the expression
   intensity of c-kit^+ positive cells in bone marrow in different
   designated treatment groups. Scale bar is 200 μm. (G) Representative
   histological examinations of H&E stained organs from mice receiving
   indicated treatment. Scale bar is 200 μm

Discussion

   In this study, we have successfully identified USP5 as a critical
   regulator of MSCs proliferation in MPN driven by the JAK2^V617F
   mutation. Our findings demonstrate that USP5 overexpression in MSCs
   contributes to the aberrant proliferation observed in MPN by modulating
   the apoptosis pathway, specifically through the regulation of
   Caspase-3. This discovery underscores the importance of targeting USP5
   as a therapeutic strategy for MPN, highlighting its potential as a
   novel biomarker and therapeutic target.

   We have further developed an innovative approach to deliver USP5
   inhibitors directly to the BM microenvironment using engineered stem
   cell-derived exosomes. Unlike conventional therapies such as
   hydroxyurea and ruxolitinib, which often face challenges of systemic
   toxicity and limited bone marrow penetration, our engineered exosomes
   enable precise delivery to the bone marrow niche, thereby enhancing
   efficacy while minimizing off-target effects. These exosomes,
   co-expressing CXCR4 and a P-selectin-targeting peptide, efficiently
   home to BM MSCs, overcoming the challenges of traditional drug delivery
   methods. The encapsulation of USP5-IN-1 within these exosomes ensures
   sustained release and effective inhibition of USP5 activity, leading to
   significant suppression of MSCs proliferation and induction of
   apoptosis. While our engineered exosomes show promise, scaling
   production and ensuring batch-to-batch consistency remain challenges
   for clinical translation. Future studies will focus on optimizing
   GMP-compliant manufacturing processes [[138]20].

   Our findings establish USP5 as a critical therapeutic target in MPN,
   with dual-targeted exosomes offering a promising strategy for clinical
   translation by overcoming the limitations of current therapies. While
   USP5-IN-1 at high concentrations (> 5 µM) induced death in normal MSCs,
   our targeted delivery system (USP5@Exosome-CP) achieved selective
   killing of Mut-MSCs at therapeutic doses (1 µM) without affecting
   normal MSCs (Fig [139]S2, [140]S8). This specificity is attributed to:
   (1) higher USP5 dependency in Mut-MSCs, and (2) exosome-mediated
   targeting reducing systemic exposure. The use of engineered exosomes as
   nanocarriers represents a significant advancement in targeted drug
   delivery, potentially improving treatment efficacy and reducing
   systemic toxicity [[141]30–[142]32]. This approach could be further
   explored for other hematological malignancies and solid tumors, where
   precise targeting of the tumor microenvironment is crucial for
   effective therapy. Overall, our research opens new avenues for the
   development of personalized and targeted therapies for MPN,
   contributing to the broader field of cancer research and treatment.

Materilas and method

Materials

   Fetal Bovine Serum (FBS) was obtained from Zhejiang Tianhang
   Biotechnology Co., Ltd. (Huzhou, China). Plasmocin was bought from
   InvivoGene (Toulouse, France) and penicillin/streptomycin was obtained
   from Biosharp (Hefei, China). Sterile 1×phosphate bufered saline was
   purchased from Gibco Life Technologies, Inc. (Grand Island, NY, USA).
   CXCL12 chemokine were purchased from Biolegend (San Diego, CA, USA).
   USP5-IN-1 was bought from MCE (cat number: HY-139979). Acetonitrile,
   methanol and chloroform for High Performance Liquid Chromatography
   (HPLC) were all purchased from Termo Fisher Scientifc (Waltham, MA,
   USA) and their purity was more than 99%. The fluorescence dye DiD or
   CFSE (5,6- carboxyfluorescein diacetate, succinimidyl ester) were
   obtained from MedChemExpress (NJ, USA). Radioimmunoprecipitation assay
   buffer and the inhibitors of protease and phosphatase were obtained
   from Beyotime (Shanghai, China). For western blot, primary antibodies
   CD63, CD81 and Alix were obtained from Proteintech Group, Inc.
   (Chicago, IL, USA). Secondary antibodies goat anti-mouse IgG H&L-HRP
   conjugated and goat anti-rabbit IgG H&L-HRP conjugated were bought from
   Abcam (Cambridge, UK). All the antibodies for fow cytometry and
   immunofluorescence were bought from Biolegend (San Diego, CA, USA).
   CP05-PSN peptide (CRHSQMTVTSRLGGGSDAEWVDVS) or FITC labeled CP05-PSN
   peptide was purchased from Shanghai Apeptide Co., Ltd. (Shanghai,
   China).

Cell lines

   Mouse bone marrow-derived stem were generated as previously described
   [[143]33]. Murine MSCs were purchased from Pro-cell (Cat number #
   CP-M131) and cultured in Complete medium (specifically designed for
   mouse MSCs, purchased from Pro-cell) containing 10% (v/v) FBS and 1%
   (v/v) penicillin/streptomycin. All the mediums were added with 10%
   (v/v) FBS and 100 µg mL^− 1 penicillin/streptomycin.

BMECs isolation

   The BMECs of mice were acquired using a previously described approach
   [[144]34]. In brief, the epiphyses of bones were removed and the marrow
   was flushed out and cultured overnight with Dulbecco’s modified Eagle’s
   medium (DMEM) supplemented with 20% FBS. Then, the floating cells were
   discarded. The adhered cells were digested with trypsinization and
   selected using CD31 microbeads to isolate BMECs. The obtained BMECs
   were grown in a culture flask pretreated with rat-tail collagen type 1.
   The BMECs were cultured with a ECM kit (Oricell) that contained growth
   factors for ECs. FCM and tube formation assays were conducted for the
   characterization of the isolated BMECs.

Reverse transcription quantitative PCR (RT-qPCR) analysis

   Total RNAs were extracted using the TRIzol reagent (Invitrogen), and
   reverse transcription reactions were performed using the PrimeScript RT
   reagent kit (TaKaRa, Cat. No. RR470A) with a mix of random 6 mers and
   oligo (dT) primers. After mixing well generated cDNA templates with
   primers/probes and PerfectStart Green qPCR SuperMix (Transgen), RT-qPCR
   was performed with the Bio-Rad CFX Connect Real-Time PCR Detection
   System (Bio-Rad). The housekeeping gene, GAPDH, was used as a loading
   control. Primers were listed in Table [145]2 (Supporting information).

Western blotting

   All the exosomes and cells were lysed by RIPA bufer with the inhibitors
   of protease and phosphatase at 4 °C for 30 min, and then centrifuged at
   12,000 g for 30 min at 4 °C. Te mass of the sample loading was adjusted
   to the same according to their protein concentrations that were
   detected by BCA Protein Assay Kit. The samples were separated by
   SDS-PAGE and transferred to polyvinylidene difuoride membrane after
   boiled for 5 min. The membranes block by 5% not-fat milk at room
   temperature for 1 h and incubated with related primary antibodies at
   4 °C overnight. With several wash by Tris-bufered saline with 0.05%
   Tween-20, secondary antibodies incubated with the membranes at room
   temperature for 1 h. NcmECL Ultra ([146]P10100, NCM Biotech) was
   applied for chemiluminescent exposure of the blot.

Lentivirus construction and MSCs transfection

   The peptide sequences for Lamp2b-CXCR4 expression are listed in the
   supporting information. Lentivirus containing a purinomycin resistance
   gene, cas9 protein gene, luciferase gene, sgRNA for USP5/USP8 (sgRNA
   sequence for USP5 is 5’-GTGTACTATACGGGAAACAGCGG-3’, sgRNA sequence for
   USP8 is 5’-TTTCAACCGAAACTGCTACCAGG) or gene for USP13, USP42 or USP 51
   was constructed by Shanghai Jikai Biotechnology Co., LTD. The virus
   titer used to transfect MSCs was 1 × 10^7.

Cell proliferation detection

   Cells were stained using a dye such as CFSE at a working concentration
   and then co-incubated with the different designated treatment for 24 h,
   after which the proliferation of the cells was observed using flow
   cytometry, where cells with attenuated CFSE fluorescence intensity were
   considered to be proliferating cells.

Bioinformatics analysis

   The mRNA expression data of USP5 in different cancer types, different
   stage and corresponding AML patients were obtained from the online
   analysis tool of Timer 2.0 database ([147]http://timer.cistrome.org/)
   and the online analysis tool of the UALCAN database
   ([148]https://ualcan.path.uab.edu/index.html). The surival data of AML
   patients were obtained from the online analysis tools of GEPIA 2
   database ([149]http://gepia2.cancer-pku.cn/#survival).

Pre-processing of scRNA-seq data

   The scRNA-seq data for AML samples (accession number [150]GSE235923)
   were obtained from the GEO database
   ([151]http://www.ncbi.nlm.nih.gov/geo). The raw output data were
   processed with the “Seurat” R package (version 4.3.0;
   [152]http://satijalab.org/seurat/) for each individual sample. In each
   sample, low-quality cells that did not meet the following criteria were
   excluded: (1) 250 ≤ nCounts; (2) 200 ≤ nFeatures; (3) percentage of
   mitochondrial genes ≤ 20. Also, to avoid the effect of doubled cells,
   we excluded from the data doubled cells identified using “scDblFinder”
   (version 1.12.0; [153]https://plger.github.io/scDblFinder/) R package.

Data integration and the dimensionality reduction

   All samples were merged with the “AddSamples” function into one Seurat
   object. The merged Seurat object was normalized and scaled by
   regressing out UMI count and percentage of mitochondrial genes. For
   dimensionality reduction, the top 3000 highly variable genes (HVGs)
   were determined using the “FindVariableGenes” function. Dimensionality
   reduction was then performed using PCA and UMAP plots were generated by
   the “RunUMAP” function with the first 100 PCs as input, determined by
   visualizing the drop off in PC variance explained using the “ElbowPlot”
   function in Seurat. The batch effects were removed by the “Harmony” R
   package (version 1.2.0; [154]https://github.com/immunogenomics/harmony)
   based on the top 50 PCA components identified.

Cell-clustering and annotation

   The clustering analysis was performed based on the integrated joint
   embedding produced by Harmony with the Louvain algorithm after
   computing a shared nearest-neighbor graph with the Louvain algorithm
   that was implanted in the “FindClusters” function of the Seurat
   package. The identified clusters were visualized on the 2D map produced
   with the UMAP method. To annotate the cell clusters, DEGs with high
   discrimination abilities between the groups were identified with the
   “FindAllMarkers” function in Seurat using the default non-parametric
   Wilcoxon rank sum test with Bonferroni correction. The cell groups were
   annotated based on the DEGs and the well-known cellular markers from
   the literature. We then annotated clusters based on the average
   expression of genes in the following major cell types: MSCs (CD34),
   myeloid cells (CD68), T cells (PTPRC and CD3D), B cells (MS4A1), and
   fibroblasts (PDGFRB), endothelial cells (EPECAM1), and plasma cells
   (JCHAIN).

RNA sequencing (RNA-seq) and bioinformatic analysis

   Total RNAs were isolated from 1 × 10^6 either the Mut-MSCs or Mut-MSCs
   with knock down USP5 gene by using TRIzol Reagent following the
   manufacturer’s instructions (Invitrogen). RNA libraries were were
   prepared and sequenced by the Illumina NovaSeq 6000. For analysis of
   RNA-seq results, RNA-seq reads quality was evaluated using FastQC
   (v0.11.9,
   [155]https://www.bioinformatics.babraham.ac.uk/projects/fastqc/) and
   aligned to the mouse genome GRCm38 by HISAT2 (v2.2.1,
   [156]https://daehwankimlab.github.io/hisat2/). FeatureCounts (v2.0.1,
   [157]http://subread.sourceforge.net/) was used to quantitate the
   transcriptome using the GTF annotation files. Differential analyses
   were performed to the count files using the R packages DESeq2 (v1.28.1,
   [158]https://bioconductor.org/packages/release/bioc/html/DESeq2.html),
   following standard normalization procedures. The differentially
   expressed genes (DEGs) were identified with adjusted p value < 0.05 and
   absolute log2 fold-change > 1 and plotted with R packages ggplot2
   (v3.3.2, [159]https://cran.r-project.org/web/packages.

   /ggplot2/index.html). Heatmaps were generated using pheatmap package
   (v1.0.12,
   [160]https://cran.r-project.org/web/packages/pheatmap/index.html) in R
   (v4.0.2, [161]https://cran.r-project.org). Gene-set enrichment analysis
   was performed using the GSEA software (v4.1.0,
   [162]https://www.gsea-msigdb.org/gsea/index.jsp).

Preparation of exosomes and other engineered Rexo

   In 10 cm cell culture dishes, 6 × 10^6 MSCs were planted. Exosomes were
   isolated using differential ultracentrifugation with a Beckman Coulter
   Allegra X-15R (rotor SX4750) at 1,000 × g for 10 min to remove cells,
   followed by 14,000 × g for 30 min to eliminate debris, and finally
   pelleted at 100,000 × g for 70 min (Beckman Optima XE-90, Type 70 Ti
   rotor). Washed exosomes were resuspended in sterile PBS for downstream
   applications. To prepare Exosome-CP, the exosomes were infused with
   CP05-PSN peptide in a 2:5 ratio of mass for overnight, Overnight
   incubation allows CP05-mediated anchoring to exosomal CD63 via
   high-affinity binding, as previously described [[163]29]. Exosomes-CP
   were gained from the supernatant via 100,000 g for further 60 min at
   4 °C and washed with sterile 1×PBS for 2 times. At last, The
   Exosomes-CP were resuspended with 1×PBS for subsequent experiments.

Quantification of exosomes and other engineered Rexo

   The protein concentrations of exosomes and other engineered exosomes
   were measured. After washing, indicated engineered exosomes were lysed
   with radioimmunoprecipitation assay (RIPA) buffer at 4 °C for 30 min
   and then centrifuged for 30 min at 12,000 g at 4 °C. The supernatant
   containing the total protein was transferred to a new centrifuge tube.
   Protein was quantified using the BCA Protein Assay Kit (Thermo Fisher
   Scientific) in accordance with the manufacturer’s protocol.

Exosomes encapsulated with USP5-IN-1 via electroporation

   USP5-IN-1 was dissolved in ddH[2]O with 0.1% DMSO to aid solubility,
   and the drug concentration was 10 mg mL^− 1. The drug was mixed with
   exosomes in a 1:1 ratio of mass in 400 mM sucrose solution. By an
   electroporation system (Gene Pulser X cell, Bio Rad, USA), 400 µL
   mixture was electroporated in 0.2 cm cuvettes via exponential pulse
   (voltage: 500 V; capacitance: 250 µF; pulse duration: 10 ms). The
   loading efficiency (LE) was 12.5 ± 2.1%, and drug content (DC) was
   0.12 mg USP5-IN-1 per mg exosome protein, determined by HPLC.

Characterization of exosome size and transmission electron microscopy (TEM)

   One milliliter of 30 ng mL^− 1 exosomes were taken for the measurement
   of the particle size and polydispersity index by Malvern laser particle
   size analyzer (Zetasizer Nano.

   ZSP). For further identifcation of the sizes and morphology of exosomes
   were washed by ddH[2]O, deposited on copper mesh and then observed by
   TEM (HT7700-SS/FEI Tecnai G20 TWIN).

Analysis of USP5-IN-1 loaded in USP5@Exosome-CP in vitro by HPLC

   Ultrasound of USP5@Exosome-CP was performed in methanol solution
   following centrifugation (11 0000 g, 30 min). Then, the supernatants
   were filtered (0.2 μm filters) for HPLC (LC-2030 C Plus, designed by
   Shimadzu Corporation in Japan). A C18 (250 × 4.6 mm, 5 μm particle
   size) HPLC packed column was used as the chromatographic column. The
   mobile phase was CH3OH 0.5%TFA/H[2]O 0.5% TFA (1:1, V/V), the flow rate
   was 1.0 mL min^− 1, and the detection wavelength was 254 nm.

Immunofluorescence and confocal microscopy

   Slice or cells were fixed in 4% paraformaldehyde for 15 min and then
   kept in a blocking buffer for 1 h. Then, the cells were incubated with
   primary antibodies overnight at 4 °C, followed by staining with
   secondary antibodies. Nuclei were stained with DAPI. After that, the
   cells were visualized using an LSM710 laser scanning confocal
   microscope.

Mice

   C57BL/6J female mice were obtained from Hunan Slyke Jingda Laboratory
   Animal Co. Ltd. in Hunan, China, and JAK2^V617F genetic mutation mice
   were obtained from Yuan Zhou laboratory. These mice were bred and
   maintained in a specific pathogen-free (SPF) barrier facility. All
   animal studies were conducted under the approval of the Hubei
   Provincial Animal Care and Use Committee and followed the experimental
   guidelines established by the Animal Experimentation Ethics Committee
   of Huazhong University of Science and Technology. This study was
   approved by the Ethics Committee of Huazhong agriculture University
   (HZAUMO-2024-0334).

Construction of a mouse model of myeloproliferative neoplasms

   JAK2^V617F genetic mutation mice BM c-kit^+ cells were obtained: mice
   were killed by cervical dislocation, femur, tibia and ilium were
   isolated, BM cells were collected, and c-kit + cells were sorted out by
   incubation with CD117 magnetic beads. Construction of primary mouse
   mutant cells: luciferase-labeled gene vector was constructed using
   retroviral system, and the retroviral vector was packaged in HEK293T
   cells, and then the viral supernatant was obtained and infected with
   the BM c-kit + cells of mice. To construct a mouse model for MPN
   transplantation, primary c-kit + cells containing the luciferase gene
   were collected from infected mice and implanted into homozygous female
   recipients irradiated with a lethal dose of γ-rays (8.0 Gy) via a tail
   vein. The MPN transplanted mice were evaluated for the changes in
   peripheral blood counts by tail vein blood sampling at regular
   intervals after transplantation, and the size of the spleen and the
   degree of myelofibrosis in the mice were observed.

Bioluminescence and fluorescence imaging

   After anesthetizing C57BL/6J female mice (6 ~ 8 weeks old) with 1%
   pentobarbital sodium, they were intraperitoneally injected with firefly
   luciferin (150 mg kg − 1; Sigma-Aldrich; CAS: 103404-75-7). After
   15 min, mice were imaged using the Bruker In Vivo MS FX PRO Imager with
   3 min exposure times for acquiring luminescent or fluorescent images.

Histopathological assays

   The organs from mice were collected, immediately fixed in 10% (m/V)
   neutral-buffered formalin, and embedded in paraffin. Sections
   approximately 5 μm thick were cut and mounted on slides, and were
   examined using H&E staining and visualized under a light microscope.

Data statement

   Sample sizes were predetermined based on previous experience, using at
   least three groups of mice. All experiments were replicated at least
   twice to confirm findings. Statistical analyses were conducted using a
   two-tailed unpaired t-test or one-way ANOVA as appropriate. Mice were
   randomly assigned to treatment groups, and, where possible, treatment
   groups were blinded until statistical analysis. No animals or potential
   outliers were excluded from the data sets presented in this study. Data
   used and analyzed during the study are available from the corresponding
   author on reasonable request.

Electronic supplementary material

   Below is the link to the electronic supplementary material.
   [164]Supplementary Material 1^ (4.2MB, docx)

Acknowledgements