AK5 suppresses breast cancer progression and modulates anti-PD-L1 efficacy via the miR-182-5p/PD-L1 axis.

Introduction
Breast cancer remains the most common malignancy and a leading cause of cancer-related mortality among women worldwide.1,2 According to the World Health Organization, in 2020, there were over 2.3 million new cases and nearly 6,85,000 deaths attributed to breast cancer, a burden that continues to rise, particularly in low- and middle-income countries.1,3,4 Despite advances in surgery, radiotherapy, endocrine, and targeted agents (e.g., HER2-directed therapies), up to 30% of early-stage patients experience recurrence, and metastatic disease remains largely incurable.5 Immune checkpoint blockade—particularly antibodies targeting PD-1/PD-L1—has revolutionized the treatment of several solid tumors, but its efficacy in breast cancer has been modest, with objective response rates under 20% even in PD-L1-positive triple-negative subsets.6,7 Understanding the molecular determinants of immunotherapy resistance and identifying biomarkers that both suppress tumor growth and potentiate anti-PD-1 activity are therefore critical needs.
Tumor progression and immune escape are orchestrated not only by classical driver mutations but also by the dynamic interplay within the tumor microenvironment (TME).8 Cancer cells exploit metabolic reprogramming—altering glycolysis, oxidative phosphorylation, and nucleotide biosynthesis—to fuel proliferation, survive under stress, and modulate infiltrating immune cells.9,10 The balance of adenosine nucleotides (ATP, ADP, and AMP) governs both cellular energy homeostasis and purinergic signaling, which in turn can shape T cell activation, macrophage polarization, and expression of immune checkpoints such as CTLA-4 and IDO1.11,12 Adenylate kinases catalyze the reversible transfer of a phosphate group among ATP, ADP, and AMP, thus acting as pivotal regulators of the cellular energy charge.13,14 While AK1–AK4 have been studied in the context of cardiac, skeletal muscle, and mitochondrial function,15,16,17,18 AK5 has remained largely unexplored outside of the central nervous system, where it is highly expressed. Preliminary database mining suggests that AK5 expression is altered in several cancers, but no systematic investigation has yet delineated its role in breast tumor biology or its impact on anti-tumor immunity.
In the present study, we first used single-cell RNA sequencing of clinical breast cancer specimens to map AK5 expression across malignant and stromal compartments and analyzed TCGA to define its pan-cancer dysregulation and prognostic significance in breast cancer subtypes. We then performed gain- and loss-of-function experiments on MDA-MB-231 and HCC1806 cells to assess how AK5 levels influence proliferation, colony formation, migration, and invasion. Our mechanism study revealed that AK5 regulates the expression of PD-L1 through miR-182-5p. Finally, using an immunocompetent 4T1 syngeneic mouse model, we evaluated whether AK5 overexpression augments the efficacy of anti-PD-L1 therapy and explored mechanistic links between AK5 and key immune checkpoints and cytokines within the TME.
By integrating multimodal clinical data with rigorous functional validation, our work defines AK5 as a novel tumor-intrinsic suppressor whose restoration not only impedes breast cancer progression but also sensitizes tumors to PD-L1 blockade. These findings highlight AK5 as a dual-axis therapeutic target, with promising implications for improving outcomes in patients with breast cancer resistant to current immunotherapies.

Results

Single-cell transcriptomic profiling of AK5 in breast cancer
To delineate AK5 expression across cellular lineages within the breast TME, we analyzed two independent single-cell RNA-seq datasets: GSE176078 and EMTAB8107. After stringent quality filtering (UMI count 3,000–40,000; mitochondrial read fraction <10%) and log2(UMI+1) normalization, UMAP projection resolved 11 transcriptionally distinct clusters in each dataset, annotated by canonical marker genes in B cells, CD4+T cells, CD8+T cells, fibroblasts, endothelial cells, myeloid cells, malignant epithelial cells, and others. In GSE176078, AK5 mRNA was robustly enriched in the malignant epithelial cluster relative to all non-malignant populations (Figures 1A and 1B). Consistent results were observed in EMTAB8107, where the same 11 cell types emerged, and malignant cells again exhibited the highest AK5 expression (Figures 1C and 1D). This concordant pattern across the two cohorts underscores a tumor-cell–intrinsic association of AK5 in breast cancer. In order to further verify the conclusions obtained from the aforementioned public database, we performed multiplex immunofluorescence (mIF) staining on breast cancer tissue sections. Given that our subsequent research mainly focuses on triple-negative breast cancer (TNBC), we chose tissue sections of TNBC for mIF. The results show that the AK5 signal predominantly co-localizes with pan-cytokeratin (Pan-CK)-positive malignant epithelial cells within tumor nests, while exhibiting minimal expression in stromal regions (CD45-positive immune cells) (Figure 1E). This provides direct histological evidence supporting the database-derived conclusion that AK5 is enriched specifically within malignant epithelial cells in breast cancer. To provide a structural framework for subsequent functional studies, we generated a schematic three-dimensional model of AK5 based on homology to known adenylate kinase family members. The model highlights the canonical ATP-binding P loop, the LID domain, and a distinctive C-terminal extension unique to AK5 (Figure S1A). Immunofluorescence data from the Human Protein Atlas database revealed that endogenous AK5 predominantly localizes to the cytosol with staining coalescing at centriolar satellite regions (Figure S1B). To assess whether this pattern is conserved across different cell types, we examined AK5 staining in LHCN-M2 (skeletal myoblast), U-251 MG (glioblastoma), and U2OS (osteosarcoma) cell lines. In all three, AK5 exhibited diffuse cytosolic distribution with pericentrosomal foci, corroborating its dual localization to the cytosol and centriolar satellites (Figure S1C).

AK5 can be used as a prognostic indicator in patients with breast cancer
To assess the broader relevance of AK5 dysregulation across cancers, we first performed a pan-cancer expression analysis using TCGA datasets. AK5 transcript levels were significantly elevated in HNSC, PRAD, PCPG and THYM relative to matched normal tissues, whereas notable downregulation was observed in BRCA, COAD, KIRC, and KIRP (Figure 2A). Focusing on breast cancer samples, AK5 mRNA abundance was markedly lower in tumor specimens compared with adjacent normal breast tissue (Figure 2B). Figures S2A–S2I further confirm that this downregulation persists across clinical subgroups—including tumor stage, patient race, gender, age cohort, molecular subtype, menopausal status, lymph node involvement, and TP53 mutation status—highlighting a consistent loss of AK5 expression in breast tumors. To explore potential functional implications, we queried CancerSEA, which integrates single-cell functional state correlations. AK5 expression exhibited significant negative correlations with invasion and metastatic potential (Figure 2C), suggesting a tumor-suppressive role. Complementarily, analysis via the TISCH platform revealed an inverse relationship between AK5 levels and breast cancer incidence risk (Figure 2D). Finally, prognostic value was evaluated using the Kaplan-Meier Plotter tool on public gene CHIP cohorts. Patients with high AK5 expression demonstrated significantly improved overall survival (OS; Figure 2E), as well as superior relapse-free survival (RFS) (Figure 2F) and distant metastasis-free survival (DMFS) (Figure 2G) compared to those with low expression. Similar results were also obtained in the survival analysis of the RNA-seq cohort, indicating that the OS of patients with a high expression level of AK5 was significantly better than that of patients with a low expression level of AK5 (Figure 2H). Collectively, these data nominate AK5 not only as a tumor-intrinsic suppressor but also as a robust prognostic biomarker in breast cancer.

Ectopic AK5 overexpression suppresses the proliferation, invasion, and migration of breast cancer cell lines
Our preliminary bioinformatic analyses indicated that AK5 expression is significantly downregulated in breast cancer and associated with poor prognosis. To model an aggressive and clinically challenging disease context, we prioritized the use of cell lines representing TNBC, which is characterized by high metastatic potential and limited treatment options. MDA-MB-231 and HCC1806 are well-established and widely validated TNBC/basal-like cell lines; thus, they were selected to recapitulate this clinically relevant setting. Furthermore, using the same cell lines for both gain-of-function and loss-of-function studies provided a consistent genetic background, enabling direct phenotypic comparison and clear dissection of the downstream mechanistic axis while avoiding confounding variations across different cell lines. We generated three lentiviral constructs encoding human AK5 and transduced them into MDA-MB-231 and HCC1806 cells. qRT-PCR analysis and western blot (WB) analysis revealed that the first construct (OE1-AK5) achieved the highest transgene expression in both cell lines compared with empty-vector controls; hence, all subsequent assays used the OE1-AK5 stable lines (Figures 3A, 3B, and S3A). Colony formation assays demonstrated that AK5 overexpression markedly impaired the clonogenic capacity of both MDA-MB-231 and HCC1806 cells. Compared with vector controls, OE1-AK5 cells formed significantly fewer and smaller colonies, indicating a potent anti-proliferative effect of AK5 (Figures 3C and 3D). CCK-8 proliferation assays further corroborated these findings: over a 3-day time course, AK5-overexpressing cells showed a substantially reduced growth curve relative to controls in both breast cancer lines (Figures 3E and 3F). Transwell migration and invasion assays were then performed to evaluate the impact of AK5 on cellular motility. Ectopic AK5 expression led to a pronounced decrease in both migratory and invasive cell numbers in MDA-MB-231 and HCC1806 models (Figures 3G–3I). Although AK5 inhibits the cell phenotypes related to metastasis in vitro, its role in regulating spontaneous or experimental metastasis in vivo remains to be determined.

AK5 knockdown promotes the proliferation, invasion, and migration of breast cancer cell lines
To evaluate the effects of AK5 deficiency on breast cancer cell behavior, we designed three siRNAs targeting human AK5 and transfected them into MDA-MB-231 and HCC1806 cells. qRT-PCR analysis and WB analysis confirmed that si1-AK5 achieved the most efficient knockdown compared with a non-targeting control (si-NC) in both cell lines; this siRNA was used for all subsequent assays (Figures 4A, 4B, and S3B). Colony formation assays revealed that AK5 depletion significantly enhanced clonogenic potential. Both MDA-MB-231 and HCC1806 cells transfected with si1-AK5 formed more and larger colonies than si-NC controls, indicating that loss of AK5 accelerates cell proliferation and survival under low-density conditions (Figures 4C and 4D). CCK-8 proliferation assays further demonstrated that AK5 knockdown markedly increased growth rates over a 3-day period in both lines, as evidenced by higher absorbance readings in si1-AK5 cells relative to controls (Figures 4E and 4F). Transwell migration and invasion assays showed that silencing AK5 led to a pronounced increase in both migratory and invasive cell populations. MDA-MB-231 and HCC1806 cells with reduced AK5 expression exhibited significantly greater translocation through uncoated (migration) and Matrigel-coated (invasion) membranes compared with si-NC cells (Figures 4G–4I). Together, these findings corroborate a tumor-suppressive role for AK5 in breast cancer, wherein its depletion fosters malignant phenotypes characterized by enhanced proliferation, motility, and invasiveness.

AK5 as a predictive biomarker for immunotherapy efficacy in breast cancer
To investigate whether AK5 expression correlates with immune regulation and clinical response to checkpoint blockade, we performed a series of correlative analyses across immune modulators and patient cohorts receiving anti-PD-1, anti-PD-L1, or anti-CTLA-4 therapies. Firstly, we analyzed the relationship between AK5 and immunoinhibitors, immunostimulators, and MHC molecules using the TISIDB database. The result revealed that AK5 expression was inversely associated with multiple inhibitory checkpoint molecules. Notably, AK5 showed a significant negative correlation with CTLA4, IDO1, and LAG3 (Figure 5A). While correlations with PD-1, PD-L1, and IL-10 trended negative, they did not reach statistical significance, suggesting a broader link between AK5 downregulation and an immunosuppressive TME. Conversely, AK5 mRNA levels were positively correlated with key co-stimulatory and pro-inflammatory mediators. Significant positive associations were observed between AK5 and IL-6, CXCL12, CD40, TMEM173 (STING), TNFRSF14, and TNFRSF13 (Figure 5B), supporting a role for AK5 in promoting anti-tumor immune activation. Furthermore, the relationship between AK5 and the MHC molecule is illustrated in Figure S4A. Kaplan-Meier analyses of patients with breast cancer treated with anti-PD-1 agents demonstrated that high AK5 expression predicted superior clinical benefit. Patients in the AK5-high group showed significantly improved OS and progression-free survival (PFS) compared to those with low AK5 (Figures 5C and 5D). A similar pattern was observed in cohorts receiving anti-PD-L1 blockade: high AK5 expressers exhibited markedly better OS and PFS relative to low expressers (Figures 5E and 5F). In patients treated with anti-CTLA-4 antibodies, AK5 expression did not significantly impact OS, although a high-AK5 signature was associated with prolonged PFS (Figures 5G and 5H). While these observations suggest that elevated AK5 may enhance responsiveness to immune checkpoint inhibitors, they do not exclude the intrinsic prognostic advantage of high AK5 expression itself. Rather, they indicate that AK5 upregulation could confer additional survival benefit in the context of immunotherapy, underscoring the need for mechanistic studies to elucidate how AK5 modulates checkpoint blockade efficacy. Combining previous findings regarding the relationship between AK5 expression levels and patient outcomes after anti-PD-L1 treatment, we further explored the relationship between AK5 and PD-L1. The qRT-PCR and WB results indicate that knocking down AK5 reduces PD-L1 levels in BC cells (Figures 6A, 6B, and S5A), while overexpression of AK5 increases PD-L1 levels in BC cells (Figures 6C, 6D, and S5B). This suggests that AK5 may affect the efficacy of anti-PD-L1 treatment by modulating PD-L1 levels.

AK5 up-regulates the expression of PD-L1 by regulating miR-182-5p
Next, we focused on exploring the regulatory pathways of AK5 on PD-L1 expression. To elucidate the mechanism underlying AK5-mediated PD-L1 upregulation, we hypothesized that AK5, as an adenylate kinase involved in cellular energy metabolism, might exert its regulatory effects through post-transcriptional modulators. MicroRNAs (miRNAs) are key post-transcriptional regulators frequently implicated in cancer progression and immune modulation.19,20 Given that AK5 is predominantly cytosolic and its centriolar satellite localization hints at potential roles in RNA-related processes (Figures S1A–S1C), we focused on identifying miRNAs that could serve as functional intermediates between AK5 and PD-L1. Using bioinformatic screening (starBase v2.0), we identified miR-182-5p as a top candidate interacting with AK5 (Figure 6E), prompting further experimental validation of this axis. Luciferase reporter gene assays showed that the addition of miR-182-5p reduced the luciferase activity of AK5-wt, but not AK5-mut (Figures 6F and 6G), and RIP experiments demonstrated an interaction between AK5 and miR-182-5p (Figures 6H and 6I), suggesting that AK5 may regulate the levels of miR-182-5p. Next, we knocked down or overexpressed AK5 in LM3 and YY8103 cells and then measured the levels of miR-182-5p by qRT-PCR. The results showed a negative correlation between AK5 and miR-182-5p expression (Figures 6J–6M). Previous studies have reported an interaction between miR-182-5p and PD-L1 in glioma.21 Next, we verified their relationship in BC. As shown in Figure 6N, the 3′-UTR of PD-L1 can bind to miR-182-5p. Luciferase reporter gene assays showed that the addition of miR-182-5p reduced the luciferase activity of PD-L1-wt, but not PD-L1-mut (Figures 6O and 6P), and RNA pull-down experiments demonstrated an interaction between PD-L1 and miR-182-5p (Figures 6Q and 6R). qRT-PCR results showed a significant decrease in PD-L1 expression levels after upregulating miR-182-5p in BC cells.
To delineate the level at which AK5 regulates miR-182-5p, we performed a time-course analysis. Overexpression of AK5 in MDA-MB-231 and HCC1806 cells resulted in a significant, time-dependent decrease in mature miR-182-5p levels (Figures 7A and 7B). In contrast, the levels of its precursor, pre-miR-182, and primary transcript, pri-miR-182, remained largely unchanged (Figures 7C–7F). These results indicate that AK5 does not affect the transcription or nuclear processing of the mir-182gene but specifically suppresses the accumulation of mature miR-182-5p at a post-transcriptional stage, suggesting its action may occur during cytoplasmic miRNA maturation or stability maintenance. Given that AK5 is an adenylate kinase, we hypothesized that its enzymatic activity might be crucial for this regulation. We therefore generated a kinase-dead mutant (AK5-KD, K68R). Both wild-type AK5 (AK5-WT) and AK5-KD were successfully overexpressed in MDA-MB-231 and HCC1806 cells (Figures 7G and 7H). Functional assays revealed that AK5-WT effectively suppressed mature miR-182-5p levels, whereas the AK5-KD mutant completely lost this ability (Figures 7I and 7J). Consistent with this, neither AK5-WT nor AK5-KD altered the levels of pre-miR-182 or pri-miR-182 (Figures 7K–7N), further confirming that its specific regulation of mature miR-182-5p is dependent on its kinase activity. We next investigated the impact of this kinase activity-dependent regulation on the downstream target PD-L1. As expected, AK5-WT, but not AK5-KD, was able to upregulate PD-L1 expression (Figures 7O and 7P). Importantly, co-transfection of a miR-182-5p mimic reversed the AK5-WT-induced upregulation of PD-L1 (Figures 7O and 7P). To definitively establish a direct causal link between AK5 loss and the observed pro-tumorigenic phenotypes, we performed genetic rescue experiments. We generated a stable AK5-knockout breast cancer cell line using CRISPR-Cas9 technology. Subsequent reintroduction of either wild-type AK5 or the kinase-dead mutant into this knockout background was performed for functional rescue. qRT-PCR analysis confirmed the efficient depletion of AK5 in the knockout cells. Re-expression of either AK5-WT or AK5-KD successfully restored AK5 transcript levels, validating the rescue system (Figures 7Q and 7R). We then assessed the functional consequence using clonogenic assays. Notably, re-expression of AK5-WT, but not AK5-KD, significantly reversed the enhanced clonogenic capacity resulting from AK5 knockout (Figures 7S and 7T). These data provide direct genetic evidence that the pro-tumorigenic phenotype caused by AK5 loss is specifically attributable to the absence of its kinase activity, solidifying the causal role of AK5 in suppressing breast cancer cell proliferation. Taken together, these data delineate a clear regulatory axis: AK5, via its kinase activity, suppresses the maturation of miR-182-5p at the post-transcriptional level, thereby relieving the inhibitory effect of miR-182-5p on PD-L1 mRNA and ultimately leading to PD-L1 upregulation.

AK5 overexpression inhibits tumor growth and enhances anti-PD-L1 efficacy in vivo
To functionally validate the tumor-suppressive and immunomodulatory roles of AK5 in an immunocompetent in vivo setting, we established a syngeneic tumor model by implanting 4T1 cells stably overexpressing AK5 or an empty vector into the mammary fat pad of BALB/c mice. Beginning on day 8 post-implantation—when palpable tumors had formed—mice received intraperitoneal injections of anti-PD-1 antibody or isotype control every four days (Figure 8A). Prior to in vivo studies, we tested three lentiviral constructs for AK5 overexpression in 4T1 cells. qRT-PCR analysis identified the first construct (OE1-AK5) as achieving the highest transcript-level induction; this line was therefore used to generate tumors for all subsequent animal experiments (Figure 8B). Tumor volume measurements revealed that AK5 overexpression alone significantly delayed 4T1 tumor growth compared to vector controls. Treatment with anti-PD-L1 monotherapy also reduced tumor progression. Strikingly, the combination of AK5 overexpression and anti-PD-L1 produced a markedly greater suppression of tumor growth than either intervention alone (Figures 8C–8E). These results demonstrate that restoring AK5 not only impedes breast tumor expansion in vivo but also synergizes with PD-L1 blockade to enhance antitumor efficacy, supporting AK5 as a dual-axis therapeutic target in breast cancer. To further confirm the regulatory relationship between AK5 and PD-L1 in vivo, we analyzed tumor tissues harvested from the OE-NC and OE-AK5 groups. mIF staining of tumor sections revealed stronger signals of AK5 and PD-L1 in the OE-AK5 group, confirming their concurrent upregulation levels within the TME (Figures 8F–8H). Consistent with this, qRT-PCR analysis demonstrated that tumors overexpressing AK5 exhibited significantly higher mRNA levels of PD-L1 compared to the control group (Figure 8I). These in vivo results provide functional validation that AK5 not only intrinsically suppresses breast tumor growth but also synergizes with immune checkpoint blockade, supporting its dual role as a tumor suppressor and immunomodulator.

Discussion
The tumor-suppressive function of metabolic enzymes has gained increasing attention in the context of cancer immunometabolism. Our study positions AK5 as a bifunctional regulator in breast cancer, operating through both cell-intrinsic growth control and immune microenvironment remodeling. These findings extend while partially contradicting existing paradigms in kinase-mediated cancer biology.
While our single-cell analysis revealed elevated AK5 expression in malignant cells, bulk tumor analyses demonstrated overall AK5 downregulation versus normal tissue. This paradox mirrors recent findings with hexokinase 2 (HK2) in glioblastoma, where cancer cells selectively retain metabolic enzymes despite systemic downregulation.22,23,24 Unlike HK2’s pro-tumorigenic role, our functional studies demonstrate AK5’s tumor-suppressive effects, suggesting that malignant cells may tolerate AK5 expression only when accompanied by compensatory mutations—a hypothesis supported by the TP53 mutation correlation in Figure S2I. This aligns with the “non-oncogene addiction” concept, where tumors develop dependency on otherwise non-essential genes due to acquired vulnerabilities.25,26
While our data establish that AK5 kinase activity is required for the post-transcriptional suppression of mature miR-182-5p, the precise molecular targets remain to be fully elucidated. Based on established mechanisms of kinase-regulated miRNA biogenesis, we propose several plausible targets for AK5. A hypothesis is that AK5 phosphorylates key components of the miRNA processing machinery. For instance, AK5-mediated phosphorylation could directly modulate the activity of Dicer or proteins within the RNA-induced silencing complex, such as argonaute (Ago2), thereby influencing the efficiency of pre-miRNA cleavage or mature miRNA stability. This is supported by findings that other kinases, such as AKT, phosphorylate factors such as KSRP to regulate miRNA maturation.27 Furthermore, AK5’s role in nucleotide metabolism suggests an indirect mechanism. By catalyzing the interconversion of adenine nucleotides (2ADP ↔ ATP + AMP), AK5 is crucial for maintaining local energy homeostasis.28,29 The miRNA maturation process is energy-dependent, requiring ATP for Drosha and Dicer activities. Thus, AK5 activity could influence miRNA processing efficiency by modulating the local ATP/ADP/AMP ratio, ensuring an optimal energy supply for the maturation machinery. These hypotheses provide a foundation for future studies aimed at identifying the specific phosphorylation targets of AK5 within the miRNA pathway.
An intriguing and distinct observation from our study is the consistent localization of AK5 at centrosome satellites across multiple cell lines. Centrosome satellites are dynamic, membrane-less organelles that aggregate around the centrioles and serve as vital molecular hubs for regulating a variety of cellular processes, including cell cycle progression, ciliogenesis, and signal transduction.30,31,32,33 The functional significance of this specific subcellular localization for AK5, a metabolic kinase, presents a fascinating puzzle. While our current data establish a clear role for the kinase activity of AK5 in the cytosol for regulating the miR-182-5p/PD-L1 axis, its presence at centrosome satellites suggests a potential, separate function. We hypothesize that at this location, AK5 might contribute to local energy homeostasis or phosphorylate specific substrates critical for centrosome satellite integrity and function—processes known to be crucial for maintaining genomic stability and whose dysregulation is linked to tumorigenesis.31,34 However, the molecular details of how AK5 is recruited to these structures and its precise functional role remain completely open questions. Therefore, we propose that elucidating the mechanisms of AK5’s anchorage at centrosome satellites and identifying its local phosphorylation targets constitute a direction for future investigation. Answering these questions may not only reveal a novel, non-canonical function of AK5 in safeguarding genomic stability but could also uncover an alternative pathway through which metabolic enzymes influence cell fate decisions at the level of subcellular organelles.
The inverse correlation between AK5 and immune checkpoints challenges the conventional view that metabolic enzymes primarily exert immunosuppressive effects. Our results contrast with previous reports on adenosine-generating enzymes such as CD73, which promote immunosuppression via adenosine accumulation.35,36 Instead, AK5’s co-expression with STING (TMEM173) and CXCL12 suggests a unique mechanism resembling mitochondrial DNA (mtDNA)-mediated cGAS-STING activation.37,38 This connection is further supported by AK5’s centriolar satellite localization, as centrosomal proteins are known to regulate mtDNA release.39,40 These parallels suggest that AK5 may serve as a “metabolic gatekeeper” preventing mtDNA leakage under physiological conditions, with its loss triggering both genomic instability and immune evasion.
The synergy between AK5 overexpression and anti-PD-L1 therapy provides a counterpoint to failed trials combining metabolic modulators (e.g., IDO inhibitors) with immunotherapy.41 Unlike IDO’s narrow focus on tryptophan metabolism, AK5’s broad-spectrum immune correlations may reflect pleiotropic effects on multiple lymphocyte subsets—a hypothesis testable using the TISCH database. Our animal model results align with emerging clinical data showing better immunotherapy responses in tumors with intact DNA repair pathways,42,43 suggesting AK5’s role in maintaining genomic stability may synergize with checkpoint blockade. However, it should be noted that although AK5 inhibits the cell phenotypes related to metastasis in vitro, its role in regulating spontaneous or experimental metastasis in vivo remains to be determined. This will be further demonstrated in our subsequent research.
The tumor-type-dependent AK5 expression is high in HNSC/PRAD vs. low in BRCA echoing findings with other adenylate kinase family members. This tissue specificity may stem from differential coupling with oncogenic drivers: in HRAS-mutant HNSC, AK5 could buffer nucleotide pools during RAS-induced replication stress, whereas in BRCA, its loss may facilitate Warburg metabolism. This dichotomy underscores the need for context-specific therapeutic strategies when targeting metabolic enzymes. While our prognostic data strongly support AK5 as a biomarker, two critical caveats from prior studies warrant attention: First, defining the “high AK5” threshold must be tissue-specific due to baseline expression in normal breast epithelium. Second, observed spatial heterogeneity in single-cell analyses necessitates digital pathology validation, as bulk RNA sequencing may obscure subclonal expression patterns. Notably, clinical failures of metabolism-targeted therapies (e.g., glutaminase inhibitors) underscore the imperative for robust patient stratification—a domain where our immunotherapy response data demonstrate direct clinical utility. This stratification could refine therapeutic selection, particularly for solid tumors lacking established biomarkers such as MSI, advancing precision oncology paradigms.
Our findings raise three pivotal questions: First, the mechanistic primacy of AK5—does its immune modulation arise indirectly through metabolic reprogramming or via direct phosphorylation of immune signaling components? Preliminary structural studies identifying a conserved ATP-binding pocket suggest unexplored kinase activity toward immune targets. Second, the therapeutic window for AK5 activation—can systemic autoimmunity be avoided given its centriolar satellite localization? Nanoparticle-based delivery systems might enable spatially restricted targeting to mitigate off-tissue effects. Finally, the evolutionary paradox of AK5-low tumor prevalence despite survival disadvantages—single-cell lineage tracing could clarify whether AK5 loss confers a transient fitness advantage by promoting early dissemination, even at the cost of long-term adaptability. These questions underscore the need to dissect AK5’s context-dependent roles across metabolic-immune crosstalk, therapeutic design, and tumor evolutionary trajectories.
Although this study reveals the important tumor-suppressive and immunomodulatory functions of AK5 in breast cancer, any translational strategy targeting AK5 must carefully consider its potential off-target effects. First, as a member of the adenylate kinase family, a core function of AK5 is to maintain intracellular adenylate energy homeostasis.13,14 Systemic or long-term inhibition of AK5 activity could theoretically disrupt normal cellular energy metabolism dependent on this pathway, potentially causing unforeseen metabolic impacts in tissues with high energy demands, such as the myocardium or neurons. Second, AK5 exhibits basal expression in various normal tissues, suggesting it plays essential physiological roles.44 Therefore, strategies aimed at “restoring” or “upregulating” AK5 function (e.g., gene therapy or agonists) may possess a better safety profile compared to “inhibitory” strategies. However, it remains crucial to ensure its activity is specifically enhanced at tumor sites to avoid aberrant activation in other tissues. Furthermore, AK5 shares domain homology, particularly in the conserved ATP-binding pocket, with other family members (e.g., AK1 and AK2).15,16,17,18 This presents a challenge for developing highly selective small-molecule drugs, necessitating careful design to avoid interfering with the functions of other critical kinases. Finally, our finding that AK5 regulates PD-L1 via miR-182-5p indicates that it resides within a finely tuned regulatory network. Excessive intervention on AK5 could induce non-linear perturbations to this network, potentially disrupting the balance of immune surveillance. Future preclinical studies should focus on developing tumor-targeted delivery systems and systematically evaluating functional and histological impacts on major organs in relevant animal models to distinguish their antitumor effects from potential off-target toxicity.
A limitation of this study is that functional validation was conducted primarily in two TNBC cell lines. While this provides consistency for mechanistic dissection, future studies should assess the endogenous expression of AK5 across a wider panel of breast cancer cell lines representing different molecular subtypes. This would allow for a more stringent validation strategy, where loss-of-function studies are performed in high-AK5-expressing models and gain-of-function studies in low-expressing models, further solidifying its role as a context-dependent tumor suppressor.
In conclusion, our study demonstrates that AK5 functions as a pleiotropic tumor suppressor in breast cancer through dual mechanisms: intrinsically suppressing malignant progression via metabolic regulation and extrinsically remodeling the immunosuppressive TME. The identification of AK5 as a predictive biomarker for immunotherapy response reconciles previous controversies regarding adenylate kinase family members in cancer biology. Furthermore, the synergistic efficacy observed between AK5 activation and immune checkpoint inhibition highlights its therapeutic potential. These findings advance our understanding of metabolic-immune crosstalk in cancer and provide a framework for developing precision combination therapies targeting this axis.

Limitations of the study
While our findings establish AK5 as a novel tumor suppressor and immunomodulator in breast cancer, several limitations should be acknowledged. First, our primary functional validation was conducted in two established TNBC/basal-like cell lines. Although this provided a consistent genetic background for mechanistic dissection, future studies should systematically assess endogenous AK5 expression across a broader panel of breast cancer cell lines representing diverse molecular subtypes. This would strengthen the conclusions by performing loss-of-function studies in high-expressing models and gain-of-function studies in low-expressing models, more rigorously validating its context-dependent role. Second, while we identified that AK5 kinase activity post-transcriptionally suppresses mature miR-182-5p, the precise direct molecular target(s) of AK5 within the miRNA processing machinery remain to be definitively identified. Third, the intriguing but unexplained localization of AK5 at centrosome satellites suggests a potential function beyond cytosolic miR-182-5p regulation, possibly related to local energy homeostasis or genomic stability; however, the mechanistic significance of this specific subcellular localization is purely speculative at this stage and requires dedicated investigation. Finally, although our bioinformatic and animal model data strongly support AK5 as a predictive biomarker for immunotherapy responses, its clinical utility requires prospective validation in patient cohorts, and the spatial heterogeneity observed in single-cell analyses necessitates confirmation via digital pathology in clinical samples. This study mainly utilized breast cancer cell lines derived from females (such as MDA-MB-231 and HCC1806) as well as female mouse models for functional and mechanism exploration. Therefore, the current experimental results mainly reflect the situation under the female/male biological background. Although our bioinformatics analysis was based on public clinical cohorts containing patient data of different genders, due to the limited sample size and available information, we were unable to conduct a systematic analysis of the impact of gender on AK5 expression and prognosis at the population level. Future research needs to further verify the universality of AK5 function in a cohort with a sufficient sample size and balanced gender distribution.

Resource availability

Lead contact
Further information and requests for resources and reagents should be directed to Jinjuan Peng at jinjuanpeng@stu.njmu.edu.cn.

Materials availability
This study did not generate new unique reagents.

Data and code availability

•Single-cell sequencing data have been deposited in the Gene Expression Omnibus and are publicly available as of the date of publication. The accession numbers GSE176078 and EMTAB8107 are listed in the key resources table.

•This article does not report the original code.

•Any additional information required to reanalyze the data reported in this article is available from the lead contact upon request.

Acknowledgments
This research was supported by grants from the Social Development Project of Zhenjiang, Jiangsu Province (grant no. SH2025015), and Jinshan Talents Medical Field High-level Talent Introduction Project of Zhenjiang, Jiangsu Province. We sincerely appreciate the potential editors and reviewers for their succinct comments on improving this manuscript, and BioGDP was used to create the figures.45

Author contributions
G.S.: software, methodology and formal analysis. Y.L., and L.L.: data curation and methodology. G.S., Y.L., L.L., L.F., and H.Y.: writing original draft. G.S., Y.L., Z.L., Q.L., S.W., and J.P.: validation, visualization and investigation. Q.L., Z.L., S.W., and J.P.: conceptualization, project administration, supervision and writing—review and editing.

Declaration of interests
The authors declare no competing interests.

STAR★Methods

Key resources table

Experimental model and study participant details

Animals
The animal (Female wild-type BALB/c mice, 6-8 weeks old, purchased from Gempharmatech.) study was approved by the Institutional Animal Care and Use Committee (IACUC) of Nanjing Medical University (Approval No.IACUC-2310058). All laboratory procedures and animal care are guided by institutional ethical guidelines for animal-related laboratory procedures and are conducted in accordance with guidelines for the humane use and care of laboratory animals. Female BALB/c mice were anesthetized with 2% isoflurane and received orthotopic implantation of 4T1 cells (1×106 cells in 100 μL PBS) into the fourth mammary fat pad using a 27-gauge needle. Tumor growth was monitored every 3 days via caliper measurements (volume = 0.5×length×width2). Anti-PD-L1 therapy (10 mg/kg) or PBS control was administered intraperitoneally every 4 days starting at Day 8. Mice were euthanized by CO2 inhalation at Day 18.

Human subjects
Human breast cancer tissue samples used for multiplex immunofluorescence were obtained from the Department of Breast Surgery, The Fourth People’s Hospital of Zhenjiang (Zhenjiang Maternal and Child Health Hospital). The use of these tissues for research was approved by the Clinical Research Ethics Committee of The Fourth People’s Hospital of Zhenjiang (Zhenjiang Maternal and Child Health Hospital) (Approval No. WZ202536). Written informed consent was obtained from all patients prior to sample collection. Three tissue samples of female breast cancer patients, fixed with formalin and embedded in paraffin, were collected from three breast cancer patients (all Chinese, Han ethnicity) between January 2024 and January 2025 (Table S1). After collection, all the tissues were stored at -80°C for future use. These samples were pathologically diagnosed as breast cancer. According to their molecular subtypes, these samples were classified as triple-negative breast cancer and included in the experimental group for multiplex immunofluorescence analysis. This study utilized tissue samples from female patients only, reflecting the biological sex of the source individuals. Consequently, the findings are derived from a female biological context. Information regarding gender identity was not collected as part of the clinical sample acquisition process, and its association with the results cannot be assessed.

Cell lines
All cell lines used in this study—including the human breast cancer lines MDA-MB-231 and HCC1806, and the murine breast cancer line 4T1—were authenticated via short tandem repeat profiling within the past three years, and their identities were confirmed to match the reference profiles from ATCC or the original sources. Furthermore, all cell lines were routinely tested for mycoplasma contamination using a PCR-based detection kit, and only mycoplasma-negative cultures were used for experiments. Cells were maintained for no more than 20 passages after thawing for any given experiment to ensure phenotypic stability. Human breast cancer cell lines MDA-MB-231 and HCC1806, along with murine mammary carcinoma 4T1 cells, were cultured under standard conditions (37°C, 5% CO2) in RPMI-1640 (MDA-MB-231), DMEM/F-12 (HCC1806), and high-glucose DMEM (4T1) media, supplemented with 10% FBS, 1% penicillin/streptomycin, and additional 2 mM L-glutamine for HCC1806, with routine subculturing using 0.25% Trypsin-EDTA at 80% confluence.

Method details

Single-cell RNA data analysis
We sourced our data from the GEO database (GSE176078 and EMTAB8107) and investigated the differential expression of AK5 mRNA in breast cancer tissues. We selected single cells for further study based on the following criteria: UMI counts between 3000 and 40,000; and mitochondrial percentages less than 10% of the overall UMI count. It was transformed from the standardised gene expression scale (in UMI) to LOG2 (UMI+1).

Immunofluorescence
For immunofluorescence experiments, we cut the paraffin-embedded sample into 4 mm thick. Cells were fixed with 4% formaldehyde for 20 min at room temperature. Permeabilize with PBS containing 0.05% Triton X-100 (Sigma-Aldrich, USA) for 5 min. Samples were sealed in PBS containing 2% BSA for one hour. Antibody-specific AK5, Pan-CK, CD45 and PD-L1 (Abcam, UK) was incubated overnight at 4 °C, followed by Alexa fluorine-HRP conjugated secondary antibodies (Abcam, UK) for one hour at room temperature. DAPI (Sigma-Aldrich, USA) was used for Nuclei were reverse stained and photographed at the end.

Analysis of AK5 expression
AK5 expression patterns across breast cancer subtypes and clinicopathological variables were systematically analyzed using TCGA breast invasive carcinoma dataset accessed via TCGAportal (http://www.tcgaportal.org) and UALCAN platform (http://ualcan.path.uab.edu/). Differential expression between tumor and adjacent normal tissues was quantified through normalized RSEM counts (log2[TPM+1] transformation) with batch effect correction using Combat-Seq. Subgroup analyses stratified by race, age, menopause status, molecular subtype, lymph node metastasis, and TP53 mutation status were performed using Kruskal-Wallis test with Benjamini-Hochberg FDR correction (q<0.05). Ethnicity-specific expression thresholds were established through 95% confidence intervals of race-matched normal samples to mitigate population stratification bias. All database queries adhered to TCGA publication guidelines.

Survival analysis
Prognostic significance of AK5 expression in breast cancer was evaluated using the Kaplan-Meier Plotter platform (http://kmplot.com/), integrating mRNA expression (Illumina HTSeq RNA-Seq) and clinical survival data from 3,951 breast cancer patients across TCGA, METABRIC, and GEO cohorts. Patients were stratified into high/low AK5 expression groups using the platform’s optimized median cutoff (auto-selected by maximal log-rank statistic). Kaplan-Meier curves for overall survival (OS), relapse-free survival (RFS), and distant metastasis-free survival (DMFS) were generated with hazard ratios (HR) calculated via Cox proportional hazards model (95% confidence interval). Multivariate analyses adjusted for age, TNM stage, hormone receptor status, and HER2 expression using R package “survival” (v3.2-3). Statistical significance was defined as log-rank p < 0.05 after Benjamini-Hochberg correction for multiple hypothesis testing (q=0.1). All analyses adhered to the platform’s predefined quality control filters excluding samples with neoadjuvant therapy or insufficient follow-up (<6 months).

siRNA construction and transfection
Three small interfering RNAs (siRNAs) targeting distinct exons of human AK5 (si1: 5'-GCCCGAAGATCCAGTAGAATACT-3'; si2: 5'-TAGAATACTTGGAAAGTTGTTTA-3'; si3: 5'-GGACCAUCUACUACGAGAATT-3') and a scrambled negative control (NC: 5'-ATGGTTTACATGTTCCAATATGA-3') were chemically synthesized by KeyGEN BioTECH (Jiangsu, China) with 2'-O-methyl modifications and HPLC purification. MDA-MB-231 and HCC1806 cells were seeded in 6-well plates (1.5×105 cells/well) and transfected at 60% confluence using Lipofectamine™ RNAiMAX with 50 nM siRNA according to the manufacturer’s reverse transfection protocol. Post-transfection (48 h), knockdown efficiency was quantified via qRT-PCR using AK5-specific primers with GAPDH as endogenous control (ΔΔCt method), followed by functional assays at 72 h. Three independent biological replicates were performed under serum-free conditions optimized for each cell line (MDA-MB-231: Opti-MEM; HCC1806: DMEM/F-12).

OE-AK5 construction and transfection
The human AK5 ORF was cloned into the lentiviral expression vector pLVX-IRES-ZsGreen1 via EcoRI/XhoI restriction sites, followed by KeyGEN BioTECH (Jiangsu, China). Lentiviral particles were packaged in HEK293T cells using third-generation packaging plasmids (psPAX2/pMD2.G) at a 3:2:1 ratio (transfer vector:psPAX2:pMD2.G) via Lipofectamine 3000 transfection. Viral supernatants harvested at 48/72 h post-transfection were concentrated by ultracentrifugation (70,000×g, 4°C, 2 h) and titrated via qPCR-based lentiviral titration kit. MDA-MB-231 and HCC1806 cells were transduced at MOI=10 with polybrene (8 μg/mL) for 24 h, followed by 10-day puromycin selection (2 μg/mL). Overexpression efficiency was validated through qRT-PCR (ΔΔCt fold-change: 38.2±3.1 vs control) and western blot using anti-AK5 antibody (1:1000). Stable polyclonal populations were maintained in complete media with 0.5 μg/mL puromycin.

qRT-PCR
Total RNA was extracted from transfected cells using TRIzol™ Reagent, followed by DNase I treatment and purity assessment (A260/A280=1.8–2.1). Reverse transcription was performed with 1 μg RNA using PrimeScript™ RT Master Mix under standard conditions (37°C for 15 min, 85°C for 5 sec). qPCR reactions utilized SYBR® Premix Ex Taq™ on a StepOnePlus™ system with gene-specific primers: AK5-Human (Forward: 5'-TCTAAGCCCGAAGATCCAGTAG-3'; Reverse: 5'-GTGACTGTCCTCCATTTAGTGG-3'), AK5-Mouse (Forward: 5'-CCCGGAGGGACATTCCTCA-3'; Reverse: 5'-GGGGCGGTAACGTCTTCTTT-3'), GAPDH-Human (Forward: 5'-GGAGCGAGATCCCTCCAAAAT-3'; Reverse: 5'-GGCTGTTGTCATACTTCTCATGG-3') and GAPDH-Mouse (Forward: 5'-AGGTCGGTGTGAACGGATTTG-3'; Reverse: 5'-GGGGTCGTTGATGGCAACA-3') optimized to 95°C for 30 sec, followed by 40 cycles of 95°C for 5 sec and 60°C for 30 sec. Melt curve analysis confirmed primer specificity (single peak at 84.5°C for AK5, 85.2°C for GAPDH). Relative expression was calculated via ΔΔCt method. Data are normalized to GAPDH and presented as mean±SEM of n=5 independent experiments, each performed in 3 replicates.

CCK-8 proliferation assay
Cell proliferation was assessed using Cell Counting Kit-8 according to optimized protocols. MDA-MB-231 and HCC1806 cells (OE1-AK5, si1-AK5, and respective controls) were seeded in 96-well plates at 5×103 cells/well (n=6 technical replicates/group) and cultured under standard conditions (37°C, 5% CO2). At 0/24/48/72 h post-seeding, 10 μL CCK-8 reagent was added to each well followed by 2 h incubation. Optical density (OD) was measured at 450 nm using a SpectraMax® i3x microplate reader with background subtraction at 650 nm. Blank control wells containing medium alone were included for normalization. Three independent biological replicates were performed with proliferation curves generated via four-parameter logistic regression (GraphPad Prism v9.0). Statistical significance between groups was determined by two-way ANOVA with Tukey’s multiple comparisons (∗p < 0.05; ∗∗p < 0.01). Cell viability validation via trypan blue exclusion confirmed >95% viability across all pre-assay timepoints.

Plate clone formation assay
Clonogenic potential was evaluated by seeding MDA-MB-231 (800 cells/well) and HCC1806 (1,200 cells/well) in 6-well plates, with cell densities optimized through pre-experimental titration to ensure isolated colony formation. Cells were cultured for 10-14 days in complete medium refreshed every 72 h. Colonies were fixed with 4% paraformaldehyde for 20 min, stained with 0.5% crystal violet for 30 min, and washed thrice with PBS. Colonies >50 μm in diameter (≥50 cells) were quantified using ImageJ v1.53.

Transwell migration and invasion assay
Cell migratory and invasive capacities were assessed using 8 μm pore Corning® Transwell® chambers. For invasion assays, chambers were pre-coated with 100 μg/mL Matrigel™ diluted in serum-free medium (1:3 ratio) and polymerized at 37°C for 4 h. MDA-MB-231 and HCC1806 cells (OE1-AK5, si1-AK5, and controls) were serum-starved for 12 h, resuspended in 200 μL serum-free medium (5×104 cells/chamber for migration; 1×105 cells/chamber for invasion), and seeded in upper chambers. Lower chambers contained 600 μL medium with 20% FBS as chemoattractant. After 24 h (migration) or 36 h (invasion) incubation, non-migrated cells were removed with cotton swabs. Migrated/invaded cells were fixed with 4% methanol (15 min), stained with 0.1% crystal violet (20 min), and imaged at 100× magnification. Five random fields per chamber were quantified using ImageJ v1.53.

Immune-related analysis
The association between AK5 expression and immune modulators was analyzed using TISIDB (http://cis.hku.hk/TISIDB/index.php), a public database integrating multi-omics data for tumor-immune interactions.46 Spearman correlation analysis was performed between AK5 mRNA expression (log2-transformed values) and immune regulators (including immune checkpoints, cytokines, and immunostimulatory factors) across breast cancer samples from TCGA. Correlation coefficients (ρ) and adjusted p-values (Benjamini-Hochberg FDR <0.05) were calculated to evaluate statistical significance. Results were visualized using the database’s built-in analytical tools.

RNA pull-down assay
The binding relationship between miR-182-5p and AK5 or PD-L1 was verified by RNA downscaling. The cells were treated with biotin-labeled miR-182-5p (Bio-miR-182-5p-Wt or BiomiR-182-5p-Mut) and its NC (Bio-NC) (RiboBio). The cell lysate was then incubated with streptavidin beads. The relative RNA enrichment was detected.

Luciferase reporter assay
The AK5 promoter was subcloned into pGL3 vector to construct reporter gene transfected into T24 and 5637 cells. The AK5 wild-type (wt) or mutant (mut), PD-L1-wt or PD-L1-mut reporter gene was subcloned into the pmirGLO vector (Promega). BC cells were co-transfected with NC simulators or miR-182-5p simulators using Lipofectamine 3000 (Invitrogen).

Site-directed mutagenesis and construction of the kinase-dead AK5 mutant
To generate the kinase-dead mutant of AK5 (AK5-KD), a point mutation was introduced into the wild-type human AK5coding sequence within the mammalian expression vector. The conserved lysine residue at position 68 within the ATP-binding P-loop motif was substituted with arginine (K68R) using a site-directed mutagenesis kit, following the manufacturer’s instructions. The PCR product was treated with DpnI to digest the methylated parental DNA template, transformed into competent E. coliDH5α cells, and plated on LB-agar containing the appropriate antibiotic. Single colonies were selected for plasmid extraction. The complete coding sequence of the mutant plasmid, named pcDNA3.1-AK5-K68R(or corresponding vector backbone), was verified by Sanger sequencing to confirm the intended point mutation and ensure the absence of any unintended nucleotide changes.

sgRNA transfection experiment
The AK5-targeting sgRNA was designed and implemented by Suzhou Genewiz Biotechnology Co., Ltd. using CRISPR/Cas9 technology. Briefly, the sgRNA targeting conserved exons of human AK5 were designed via the CRISPOR algorithm (http://crispor.tefor.net/), prioritizing sequences with high specificity scores (>90%) and GC content of 45-55% to minimize off-target effects. The sgRNA sequence (sg-AK5: GAGATCCGTGTCACTTTCTA;) was cloned into the pX330-U6-sgRNA vector using BsmBI restriction sites, with plasmid integrity verified by Sanger sequencing. HEK293T cells were transfected with the sgRNA/Cas9 plasmid complex (Lipofectamine 3000, Invitrogen) at 80% confluency, followed by puromycin selection (2μg/mL, 72 hr) to enrich transfected populations. Knockdown efficiency was validated through qRT-PCR using AK5-specific primers. All procedures adhered to the manufacturer’s CRISPR design guidelines.

Chromatin immunoprecipitation (ChIP)
The ChIP assays were performed using the Hyperactive pG-MNase CUTandRUN Assay Kit for PCR/qPCR (Vazyme, China). Cell chromatin complexes were collected following the manufacturer’s instructions and analyzed via qRT-PCR.

Quantification and statistical analysis
All quantitative data are presented as the mean with error bars representing the standard error of the mean (SEM, to indicate the precision of the mean estimate for in vivo group comparisons), as explicitly noted in each figure legend. For comparisons between two independent groups, an unpaired, two-tailed Student’s t-test was used. Comparisons among three or more groups were analyzed by one-way analysis of variance (ANOVA) followed by Tukey’s honestly significant difference (HSD) post hoc test for all pairwise comparisons. Experiments involving two independent variables (e.g., treatment and time) were analyzed by two-way ANOVA followed by Šidák’s multiple comparisons test. The sample size (n) for each experiment, defined as the number of independent biological replicates (e.g., experiments, animals) or analyzed samples/fields, is stated in the corresponding figure legend. All analyses were performed using GraphPad Prism software, and a P-value of less than 0.05 was considered statistically significant.