Metabolic Adaptation of Regenerative Hematopoiesis Depends on Docking-Independent Mitochondrial Connexin-43

Abstract

Hematopoietic stem cells (HSC) exhibit a distinctive antioxidant profile during steady-state and stress hematopoiesis. HSC and multipotential progenitors (HSC/MPP) are metabolically coupled to bone marrow (BM) mesenchymal stromal cells through mitochondrial transfer, a process dependent on hematopoietic connexin 43 (Cx43) and low AMP-activated protein kinase (AMPK) activity. However, the mechanism by which Cx43 preserves mitochondrial functionality in HSC remains elusive. Here, through integrated transcriptomic, proteomic, metabolomic, phenotypic, and functional analyses of HSC and their isolated mitochondria, we identified that Cx43 is present on inner and outer mitochondrial membranes of HSC/MPP, where it primarily regulates mitochondrial metabolism and ATP synthesis by preserving the mitochondrial cristae, activation of mitochondrial AMPK and 2-oxoglutarate dehydrogenase (OGDH)-a rate liming enzyme in TCA cycle and electron transfer chain. During replicative stress, Cx43 deficient HSC/MPP fail to adapt metabolically, accumulate mitochondrial Ca²⁺, increase mitochondrial AMPK activity, mitochondrial fission, mitophagy, and production of reactive oxygen species, thereby limiting HSC/MPP regeneration potential. Disruption of hyper mitochondrial fragmentation and mitophagy by Drp1 dominant negative mutant (Drp1^K38A) or restoration of mitochondrial function through ex vivo heteroplasmy prevent the harmful effects of Cx43 deficiency on mitochondrial metabolism and restore HSC activity in serial transplantation experiments. Re-expression analysis of Cx43 structure function mutants indicate that Cx43 hemichannels are sufficient to reset HSC mitochondrial metabolism, dynamics, Ca²⁺ levels, and regeneration capacity. This report defines the cell-autonomous mechanism of action behind the role of Cx43 in HSC activity and opens a venue to translational applications in transplantation.

Graphical Abstract

Introduction

Bone marrow (BM) hematopoietic stem cell (HSC) transplantation (HSCT) is widely used in the clinic to treat pathological mutations in the lympho-hematopoietic system¹. Lifelong blood cell production after transplantation depends on the ability of HSC and multipotential progenitors (MPP) to balance their ability to self-renew and their capacity to generate differentiated progenies. HSC exhaustion after transplantation is a cause of late graft failure following HSC transplantation², an event relatively frequent in the context of gene-corrected stem cell therapy³. Understanding the mechanisms that ensure HSC/MPP survival upon replicative stress on a molecular level is crucial to enable stem cell therapies to succeed.

Cx43 containing channels mediate transfer of gradient-dependent small (<1 kDa) molecules^4–6, cell-to-cell adhesion⁷ and can also be expressed and function as a vehicle of communication in extracellular vesicles and tunnelling nanotubes⁸, as a nuclear transcriptional repressor⁹, and in mitochondria where it may play distinct tissue- and Cx43 isoform-dependent roles¹⁰. Unchallenged HSC and MPP have lower reactive oxygen species (ROS) content than their replication stressed counteparts^11–13. During replicative stress, like in the context of transplantation of myeloablated hosts, rapidly cycling HSC/MPP accumulate ROS, which may lead to hematopoietic failure^11,14. Rapidly dividing HSC have been shown to reduce intracellular ROS via mitochondrial transfer to their surrounding BM microenvironment (ME)¹⁵, a process that requires the hematopoietic expression of Connexin 43 (Cx43) and reduced adenosine monophosphate dependent protein kinase (AMPK) activity¹⁵. To facilitate this process, replication stressed HSC and their MPP progeny upregulate Cx43 expression^16,17 and hematopoietic Cx43 acts as a positive regulator of HSC survival, required to prevent HSC senescence and exhaustion^16,18–20.

However, direct evidence delineating the mechanism by which Cx43 mediates mitochondrial activity and preserves HSC/MPP regenerative activity has been elusive. Here, we uncovered that mitochondrial Cx43 activity is directly responsible for the engraftment failure of Cx43-deficient hematopoiesis since serial transplantation dependent HSC/MPP exhaustion can be restored by wild-type mitochondrial pool replacement. Mechanistically, rapidly dividing HSC/MPP require mitochondrial, docking independent Cx43 to prevent excessive mitochondrial calcium accumulation, regulate tricarboxylic acid cycle and electron transfer chain activities, control the balance between mitochondrial fusion and fission dynamics, restrain hyperactivated mitophagy, and preserve their ability to transfer mitochondria and mitochondrial associated ROS.

Materials and Methods

Mice

All animal experiments were approved by the Cincinnati Children’s Hospital and Dana-Farber Cancer Institute Institutional Animal Committee. Generation and crossing of experimental mice are provided in Supplemental Material and Methods.

BM cell isolation, Flow cytometry and sorting

BM mononuclear cells were isolated from femurs, tibiae, and pelvis. Flow cytometry measurements and cell sorting procedures are described in supplemental Materials and Methods.

Mitochondrial isolation, ex vivo HSC loading and Serial competitive repopulation assays

Functional mitochondria were isolated from Dendra2-mito Lin⁻/cKit⁺/Sca1⁺ (LSK) cells and co-culture and serial transplant experiments were performed as described in Supplemental Material and Methods.

Viral transduction and hematopoietic stem cell transplant

HSC were transduced with gamma-retroviral constructs of Cx43 (FL, cys-less or ΔCT₂₅₇) or EF1α-expressing lentiviral vectors containing dynamin-related protein (Drp1) mutant K38A (Drp1^K38A) constructs and transplanted into sub-lethally irradiated (2.5Gy) NSG or lethally irradiated CD45.1⁺ congenic mice. Experimental details can be found under Supplemental Material and Methods.

Immunofluorescence studies, proximity ligation assay (PLA) and transmission electron microscopy (TEM)

Mitochondrial network remodeling (fission, fusion, mitophagy), pAMPKα1,2 (Thr172) and 2-oxoglutarate dehydrogenase (OGDH) expression, HSC fate determination by nuclear c-Myc allocation, and PLA were performed by confocal microscopy or TEM. Procedures are detailed in supplemental Materials and Methods.

Transcriptomics, proteomics and metabolomics and analysis

Bulk RNAseq in HSC, proteomics, and metabolomics in isolated mitochondria were performed and the procedures and analytical methods are described under Supplemental Material and Methods.

Immunoblotting, OGDH activity, ATP/ADP measurements, and Seahorse metabolic analysis

OGDH activity, Seahorse Metabolic Flux and intracellular ADP and ATP content analyses in HSC/MPP were performed for metabolic analyses. Western blot analysis was performed in LSK cells or isolated mitochondria. For additional details see Supplemental Material and Methods.

Statistical analysis

Data are shown as mean ± standard error of the mean (SEM) from 2–3 independent experiments. Statistical analysis used α error less than 0.05. Detailed information on specific statistical tests is provided in supplemental Materials and Methods.

Results

Cx43 is required for AMPK repression and HSC metabolic homeostasis.

To determine how hematopoietic-specific Cx43 controls the metabolic fate of HSC, we first analyzed the transcriptome of BM HSC isolated from WT and Cx43^Δ/Δ mice. Gene ontology (GO) cellular component analysis showed significant enrichment of differentially expressed genes associated with mitochondrial function, cytoskeletal dynamics, and microtubule organization (Figure 1A, Supplemental Figure 1A, Extended Data 1). Comparative GO analysis of the RNA-seq dataset against the mouse mitochondrial gene expression database MitoCarta 3.0 further indicated that Cx43 deficiency in hematopoietic cells alters the expression of genes involved in mitochondrial biosynthesis and maintenance, calcium ion transport, and metabolic pathways (Supplemental Figures 1B–D). Notably, Cx43-deficient quiescent HSC exhibited elevated expression of overall and mitochondrial localized Ca²⁺/calmodulin (CaM)-dependent protein kinase 2 (Camkk2), an upstream kinase that phosphorylates AMPK at Thr172²¹, and its mitochondrial recruitment (Supplemental Figures 1E–H).

Figure 1: Cx43 deficiency impairs mitochondrial bioenergetics in HSC.

(A) Gene Ontology (GO) analysis for significantly altered genes-cellular component in WT and Cx43^Δ/Δ HSC. Numbers in each column denote the number of differentially regulated genes associated with a GO term. (B) Representative immunofluorescent images of activated AMP- activated protein kinase phosphorylated at residue Thr172 (AMPKα1, 2-Thr172, green) and mitochondria (Tomm20, red) in WT and Cx43^Δ/Δ BM HSC sorted from primary murine bone marrow. Nuclei were counterstained with DAPI (blue). Scale bars, 2 µm and 0.5 µm. (C) Quantification of AMPKα1, 2-Thr172 in WT and Cx43^Δ/Δ BM HSC. Dots depict individual cells. (D) Quantification of percentage of AMPKα1, 2-Thr172 (green) co-localization with mitochondria (Tomm20, red) in WT and Cx43^Δ/Δ BM HSC. Dots show individual cells. (E) Proximity ligation assay (PLA; left) and the quantification analysis (right) demonstrating interactions between Tomm20 and AMPKα1, 2-Thr172 (red) in WT and Cx43^Δ/Δ HSC. Nuclei were counterstained with DAPI (blue). Scale bar, 2 µm. Dots depict individual cells. (F) HSC-enriched LSK BM cells were sorted from WT and hematopoietic Cx43^Δ/Δ chimeric mice (2 months post transplantation) and mitochondrial OCR was measured by Seahorse flux analyzer using sequential injections of oligomycin (O), FCCP (F), and Rotenone/ Antimycin A (R/AA). (G) Quantification summary of mitochondrial OCR in WT and Cx43^Δ/Δ HSC/MPP. All data points show a pool of 2 mice HSC/MPP and three experimental replicates were performed. (H, I) Extracellular acidification rates (ECAR) measured by Seahorse glycolytic rate assay in WT and Cx43^Δ/Δ BM HSC/MPP following sequential injections of Glucose, oligomycin (O), and 2-deoxy-glucose (2DG). All data points show a pool of 2 mice HSC/MPP. (J) ADP/ATP ratio in WT and Cx43^Δ/Δ HSC. Data point depict individual experiments. Data represented as mean ± SEM. All immunofluorescence quantification data are pool of 2–3 independent experiments. p values were generated by unpaired, two-tailed t test. *p<0.05, **p<0.01, ***p<0.001.

AMPK is the master regulator of metabolic activity²² integrating cellular energy status by monitoring AMP and/or ADP levels to regulate mitochondrial homeostasis^23–25. In stress hematopoiesis, AMPK acts as a negative regulator of hematopoietic regeneration¹⁵. We examined the activation levels of the catalytic unit (α) of AMPK, as assessed by phosphorylation at Thr172, in steady-state WT and Cx43^Δ/Δ HSC. Under homeostatic conditions, HSC AMPKα activation depends on Cx43 (Figures 1B–C). Co-localization and proximity ligation assay (PLA) experiments identified an increased recruitment of activated AMPKα (pAMPKα1/2-Thr172) to the outer mitochondria membrane in Cx43-deficient HSC (Figures 1D–E). Metabolic flux analyses of unstressed Cx43-deficient HSC/MPP further revealed that cell-autonomous Cx43 deficiency significantly attenuates basal and maximal oxygen consumption rates (OCR), spare respiratory capacity, ATP turnover, and their bioenergetic health index (Figures 1F–G). Conversely, these cells exhibited increased glycolysis and glycolytic capacity (Figures 1H–I). Consistent with these metabolic shifts, Cx43^Δ/Δ HSC showed a higher ADP:ATP ratio (Figure 1J), a known stimulus for AMPK activation and recruitment to mitochondria²⁶, despite no change in total mitochondria content (Supplemental Figure 1I) .

Deficiency of Cx43 in quiescent HSC leads to loss of OGDH activity and mitochondrial accumulation of α-ketoglutarate.

HSC fitness and maintenance rely on energy production driven by mitochondria^14,27–34. To understand the consequences of Cx43 expression on HSC/MPP mitochondrial metabolism and ATP generation, we first investigated its distribution within HSC/MPP mitochondria. Isolated WT HSC/MPP mitochondria express Cx43, where it co-localizes with outer membrane translocase, Tomm20 and with the inner membrane cytochrome c oxidase subunit CoxVb, as assessed by PLA (Figures 2A–D). Proteomic analysis of the mitochondrial fraction isolated from WT and Cx43^Δ/Δ HSC/MPP identified that the expression of 26% of mitochondrial proteins was downregulated in Cx43^Δ/Δ HSC/MPP, while only one, the cytosolic/outer membrane translocase Tomm34³⁵ was upregulated (Supplemental Figures 2A–D, Extended Data 2).

Figure 2: HSC mitochondrial Cx43 deficiency impairs TCA cycle and OXPHOS protein stability.

(A) Schema representing isolation of functional mitochondria from WT and Cx43^Δ/Δ HSC/MPPs and WB and proteomics analysis were performed. (B) Representative Western Blot depicting expression of Cx43 in mitochondria isolated from WT and Cx43^Δ/Δ HSC/MPPs. Tomm20 was used as a mitochondrion loading control. (C) Left: Representative PLA of Tomm20 and Cx43 (red) in WT and Cx43^Δ/Δ HSC. Nuclei were counterstained with DAPI (blue). Scale bar, 5 µm. Right: Quantification of number of Tomm20/Cx43 PLA signals. Dots show individual cells. (D) Left: Representative image of WT and Cx43^Δ/Δ HSC depicting PLA between CoxVb and Cx43 (red). Nuclei were counterstained with DAPI (blue). Scale bar, 5 µm. Right: Quantification of number of CoxVb and Cx43 PLA signals. Dots show individual cells. (E) Functional mitochondria were isolated from WT and Cx43^Δ/Δ mito-Dendra2⁺ BM HSC/MPP and differential protein expression analysis were performed. GO pathways enrichment analysis exhibiting top differentially regulated mitochondrial pathways. Numbers in each column denote the number of differentially expressed mitochondrial proteins within each GO process. List of mitochondrial protein obtained from mouse MitoCarta3.0 database (Broad Institute mouse list). n = 3 ind. biol. rep. (F) Left: Representative western blots depicting expression of NADH dehydrogenase beta sub-complex subunit 8 of Complex I (NDUFB8), succinate dehydrogenase subunit B of Complex II (SDHB), cytochrome b-c1 complex subunit 2 of Complex III (UQCRC2), cytochrome c oxidase subunit 1 of Complex IV (MTCO1), and ATP synthase subunit alpha of Complex V (ATP5A) in WT and Cx43^Δ/Δ HSC/MPPs. GAPDH was used as an internal loading control. The blots are shown in two different exposures to highlight the difference in respective complex proteins. Right: Relative expression of mitochondria complex proteins (CI-CV) vs GAPDH in WT and Cx43^Δ/Δ HSC. Dots depict individual experiments. (G) Differential levels of TCA cycle metabolites following pairwise comparison of mitochondria isolated from WT and Cx43^Δ/Δ HSC/MPP. n = 3 ind. biol. rep. (H) Relative levels of TCA cycle metabolites (α-Ketoglutarate; αKG, succinate, and fumarate) in WT and Cx43^Δ/Δ BM HSC/MPP derived mitochondria. n = 3 ind. biol. rep. (I) OGDH activity and (J), ratio of succinate and fumarate to αKG in HSC-enriched LSK BM cells sorted from WT and hematopoietic Cx43^Δ/Δ mice. n = 3–4 ind. biol. rep. (K) Left: Representative image depicting PLA between OGDH and Cx43 (red) in WT and Cx43^Δ/Δ HSC. Nuclei were counterstained with DAPI (blue). Scale bar, 3 µm. Right: Quantification of number of OGDH/Cx43 (red) PLA signals. Dots show individual cells. (L) Left: Representative immunofluorescent image depicting staining of OGDH (red) and Dendra2 (green) in WT and Cx43^Δ/Δ BM HSC. Nuclei were counterstained with DAPI (blue). Scale bars, 2 µm. Right: Quantification of OGDH intensity. Dots show individual cells. Data represented as mean ± SEM. All immunofluorescence quantification data are pool of 2–3 independent experiments. p values were generated by unpaired, two-tailed t test. *p<0.05, **p<0.01, ***p<0.001.

GO analysis of cellular components and pathway enrichment showed that the downregulated proteins were significantly associated with critical mitochondrial functions, including mitochondrial architecture (biogenesis and maintenance), ROS detoxification, calcium ion transport, the tricarboxylic acid (TCA) cycle, respiratory electron transport chain (ETC) complexes, and ATP synthesis (Figure 2E). Notably, the expression of most proteins in ETC complex I and to a lesser extent other ETC complexes (II, III, IV), components of the respirasome supercomplex (reviewed in ³⁶) and the succinate dehydrogenase complex, crucial to funnel electrons from the oxidation of NADH and FADH2, respectively, were downregulated in Cx43^Δ/Δ HSC/MPP mitochondria (Figure 2F, Supplemental Figure 2E), suggesting that Cx43 activity crucially controls the composition of the key functional complexes controlling oxidative phosphorylation (OXPHOS) activity.

Protein expression changes were accompanied by functional metabolic alterations, and we observed significant changes in TCA cycle metabolites, with an accumulation of mitochondrial α-ketoglutarate (αKG), and decreased levels of succinate and fumarate. Additionally, the ratios of fumarate and succinate to αKG, and the catalytic activity of OGDH were reduced in Cx43^Δ/Δ HSC/MPP mitochondria (Figures 2G–J, Supplemental Figure 2F). PLA analysis further revealed that Cx43 and OGDH are spatially proximal (<50 nm), and HSC/MPP Cx43 deficiency results in ~50% decreased expression of mitochondrial OGDH protein (Figures 2K–L). Deficiency of Cx43 did not affect the expression and the activity of another NADH generating enzyme of the TCA cycle, isocitrate dehydrogenase 1 (IDH1) (Supplemental Figures 2G–H), supporting the notion that the effect of Cx43 deficiency on OGDH is specific. Together, these findings suggest that mitochondrial Cx43 is essential for the steady-state maintenance of HSC/MPP ATP levels through its role in regulating OGDH activity and expression of ETC complex proteins.

Mitochondrial transplant prevents rapid exhaustion of Cx43-deficient HSC

Elevated ROS levels¹¹ and Myc abundance are the critical indicators of HSC activation and commitment^37,38. Hematopoietic Cx43 deficiency results in increased mitochondrial ROS and membrane depolarization in dividing HSC (Supplemental Figures 3A–B). Ex vivo analysis of Myc expression and its nuclear translocation in paired daughter cells demonstrated a 1.5-fold increase in the frequency of differentiated/committed daughter cells, accompanied by a reduction in symmetric self-renewal divisions (Figures 3A–B, Supplemental Figures 3C–E). To determine whether mitochondrial Cx43 mediated metabolic activity is crucial for HSC function, we used Dendra2-mitochondria (Dendra2-mito) transgenic mice in which the Dendra2 fluorescent protein is fused with a mitochondrial transmembrane protein cytochrome oxidase subunit 8a (Cox8a)³⁹. We performed ex vivo mitochondrial loading experiments, supplementing WT and Cx43^Δ/Δ BM HSC/MPP with the mitochondria isolated from WT or Cx43^Δ/Δ BM HSC/MPP expressing Dendra2-mito reporter. Flow cytometry analysis revealed that both WT and Cx43^Δ/Δ HSC/MPP incorporated isolated mitochondria at comparable levels (30–40%), independent of Cx43 expression (Supplemental Figures 3F–H). Notably, mitochondrial loading led to an approximate 30–50% increase in heteroplasmy at the single-cell level, without inducing HSC apoptosis (Supplemental Figures 3I–K).

Figure 3: Mitochondrial Cx43 prevents rapid exhaustion of regenerating HSC.

(A) Representative images showing paired daughter cell assay of nuclear Myc distribution in WT and Cx43^Δ/Δ HSC. (i) Low Myc expression in paired daughter cells represents symmetric self-renewal (SD), (ii) high and low Myc expression between the 2 daughters represents an asymmetric division (AD), whereas (iii) high Myc expression in both cells is indicative of symmetric commitment (SC). (B) Pie charts depicting relative HSC fate distribution in WT and Cx43^Δ/Δ paired daughter assay. 98–101 cell pairs analyzed from three independent experiments. (C-H) WT and Cx43^Δ/Δ HSC were co-cultured overnight with Dendra2⁺ mitochondria isolated from WT and Cx43^Δ/Δ mito-Dendra2⁺ LSK cells and analyzed. (C) Schematic representing Dendra2⁺ mitochondria loading in WT and Cx43^Δ/Δ HSC. (D-E) Representative immunofluorescence staining of Myc and the corresponding HSC fate determination (D), and its quantification (E) in WT and Cx43^Δ/Δ paired daughter HSC loaded with WT or Cx43^Δ/Δ mito- Dendra2⁺ HSC/MPP isolated mitochondria. 49–51 paired cells were analyzed from three independent experiments. *p<0.05 WT HSC^{WT mito} vs Cx43^Δ/Δ HSC^{Cx43Δ/Δ mito}; ^$p<0.05 Cx43^Δ/Δ HSC^{Cx43Δ/Δ mito} vs Cx43^Δ/Δ HSC^{WT mito}. (F-G) Representative immunofluorescence depicting OGDH (red) staining (F) and quantification (G) in WT and Cx43^Δ/Δ BM HSC loaded with WT or Cx43^Δ/Δ Dendra2⁺ mitochondria. Nuclei were counterstained with DAPI (blue). Scale bars, 2 µm. Dots show individual cells. (H) ATP levels in WT and Cx43^Δ/Δ BM HSC supplemented with WT or Cx43^Δ/Δ mito-Dendra2⁺ mitochondria. Data point depict individual experiments. (I) Schematic illustration of serial competitive repopulation assay. WT and Cx43^Δ/Δ BM HSC loaded with mitochondria isolated from WT or Cx43^Δ/Δ mito-Dendra2⁺ HSC/MPP were transplanted into lethally irradiated congenic CD45.1⁺ B6.SJL- Ptprc^a Pepc^b /BoyJ mice. On day 17 post-transplant BM cells were isolated from primary transplant mice (CD45.2⁺, pool of 5 mice), mixed with congenic CD45.1⁺ B6.SJL- Ptprc^a Pepc^b /BoyJ competitor cells in 1:1 ratio and transplanted into lethally irradiated B6.SJL- Ptprc^a Pepc^b /BoyJ recipient mice for secondary transplantation followed by tertiary transplant and analyzed as indicated. (J-K) Peripheral blood chimerism for donor derived CD45.2⁺ cells (J) and count of peripheral blood CD45.2⁺ leukocyte and CD11b⁺ cells (K) in primary, secondary, and tertiary recipients (4–5 mice/group, per time point, and two independent experiments were performed). (L) BM counts of LT-HSC and LSK in primary, secondary, and tertiary transplant mice (4–5 mice/group, per time point, and two independent experiments were performed). Data represented as mean ± SEM. Statistical significance was assessed using unpaired, two-tailed t test except in panel B and E where chi-square test and panel G where One-way ANOVA with Tukey’s multiple comparisons test was used. *p<0.05, **p<0.01, ***p<0.001.

We then performed a pair-daughter first HSC division assay in WT and Cx43^Δ/Δ HSC supplemented with WT or Cx43^Δ/Δ Dendra2⁺ mitochondria. Supplementation with functional mitochondria isolated from WT, but not from Cx43^Δ/Δ HSC/MPP, restored symmetric division in Cx43-deficient HSC to WT levels, with a concomitant decrease in differentiated/committed daughter cells frequency (Figures 3C–E). In contrast, supplementation of Cx43^Δ/Δ mitochondria to WT HSC did not affect HSC commitment (Supplemental Figure 3L). Although supplemented WT mitochondria rapidly diluted with successive HSC divisions (Supplemental Figures 3M–N), they induce mitochondrial metabolic reprogramming of Cx43^Δ/Δ HSC, as observed by OGDH expression and ATP levels (Figures 3F–H). Of note, no significant changes in OGDH expression and ATP levels were observed in WT HSC/MPP loaded with Cx43^Δ/Δ Dendra2⁺-mitochondria (Figures 3F–H).

Next, to evaluate the role of mitochondrial Cx43 on HSC fitness, we studied the long-term reconstitution capacity of HSC by serially transferring WT and Cx43^Δ/Δ HSC/MPP containing WT or Cx43^Δ/Δ Dendra2⁺-mitochondria, alone or with a WT competitors through primary, secondary, and tertiary recipients (Figures 3I). Analysis of peripheral blood (PB) and BM revealed that supplementation of Cx43^Δ/Δ HSC/MPP with WT, but not Cx43^Δ/Δ mitochondria, significantly mitigated the progressive loss of regenerative capacity of Cx43-deficient HSC/MPP (Figures 3J–L, Supplemental Figures 3O–P). Collectively, these findings highlight a mitochondria-intrinsic role of Cx43 in regulating HSC fitness, self-renewal, and fate determination.

Lack of Cx43 results in the accumulation of calcium-rich, autophagic mitochondria in dividing HSC.

During replicative stress, the HSC cell cycle is preceded by the activation of ATP-related intracellular pathways. AMPK integrates cellular energetics to regulate mitochondrial quality control, including dynamics and fate, in order to maintain energetic homeostasis^23,40. Our confocal immunofluorescence microscopy data show that Cx43-deficient, dividing HSC exhibit sustained activation of AMPK (pAMPKα1/2-Thr172) and its colocalization with the outer mitochondrial membrane, with no effect on cell viability (Supplemental Figures 4A–C). We hypothesized that impaired hematopoietic regeneration in Cx43-deficient HSC might be associated with AMPK activation-mediated changes in mitochondrial dynamics and fate during cell division. To investigate this, we first examined the mitochondrial morphology in freshly sorted quiescent HSC and after 48h of culture induced proliferation. Transmission electron microscopy (TEM) analysis revealed that Cx43 deficiency in quiescent HSC associated with increased average mitochondrial circularity and area, along with abnormal mitochondrial ultrastructure, including vacuole formation via extension of the mitochondrial outer membrane, expansion of the intermembrane space, and the presence of concentric (onion-shaped) and disorganized cristae (Figures 4A upper panels and B, Supplemental Figure 4D–E). Upon forced division, Cx43^Δ/Δ HSC exhibited exacerbated mitochondrial defects, including diminished average mitochondrial length, increased circularity, and accumulation of severely damaged mitochondria with mito-phagolysosomal features (Figures 4A lower panels and B, Supplemental Figure 4D–E). Next, to define the molecular basis of these alterations, we performed transcriptomic profiling of WT and Cx43^Δ/Δ dividing HSC (Extended Data 1). GO cellular component analysis and mitochondrial Reactome pathway enrichment revealed significant differential expression of genes associated with mitochondrial function, lysosomal activity, the TCA cycle, ATP production, and electron transport chain (Supplemental Figures 4F–H).

Figure 4: Cx43 deficiency induces mitochondrial calcium and defective mitochondria accumulation in dividing HSC.

(A and B) Representative images of mitochondrial ultrastructure from WT and Cx43^Δ/Δ quiescent (0 h) and dividing (48 h) BM HSC revealed by transmission electron microscopy (TEM). The boxed areas a, b, c, and d in panel I were magnified in panel II. The mitochondria indicated in panel II (boxed areas i, ii, iii, and iv) were further magnified in panel III and mitochondrial ultrastructure were analyzed. Green triangles, healthy mitochondria with dense matrix and preserved cristae; Cyan*, mitochondria with disorganized/concentric cristae; Red*, damaged mitochondria. Scale bars, 500 nm. (A). Quantification of mitochondria length (in µm), circularity, and mitochondrial fate in WT and Cx43^Δ/Δ quiescent (0h) and dividing (48h) HSC (B). Dots depict individual cells. (C) Left: Representative immunofluorescence image of Dendra2⁺ mitochondria (green) and Pink1 (red) in WT and Cx43^Δ/Δ dividing HSC. Nuclei were counterstained with DAPI (blue). (i) Imaris surface-building algorithm was used to reconstruct mitochondrial surface (green) and Imaris spot-building algorithm was used to define Pink1 inside (reconstructed big red spot; white arrow) and outside (reconstructed small pink spots) of mitochondria surface. Scale bars, 2 µm and 0.5 µm. Right: Quantification of Pink1 colocalization to Dendra2⁺ mitochondria (mito-Pink1) in WT and Cx43^Δ/Δ dividing HSC. Dots depict individual cells. (D) Left: Representative immunofluorescence image showing mitochondria (green) and Parkin (red) staining in WT and Cx43^Δ/Δ dividing HSC. (i) Imaris reconstructed mitochondrial surface (green) and Parkin spots was used to define inside (reconstructed big red spot; white arrow) and outside (reconstructed small pink spots) Parkin to mitochondria surface. Scale bars, 2 µm and 0.5 µm. Nuclei were counterstained with DAPI (blue). Right: Quantification analysis depicting Parkin colocalization with mitochondria (mito- Parkin). Dots show individual cells. (E) Left: Representative image of Dendra2⁺ mitochondria (green), LC3 puncta (red), and LAMP2 (purple) in WT and Cx43^Δ/Δ HSC cultured for 48h in presence of SCF (50 ng/ml) and TPO (50ng/ml). Nuclei were counterstained with DAPI (blue). (i) Reconstructed mitochondrial surface is depicted in green. Reconstructed big red and violet spots depict LC3 and lamp2 inside mitochondria surface, respectively. Reconstructed small pink and lavender spots show LC3 and lamp2 outside mitochondria surface, respectively. Scale bars, 2 µm and 0.5 µm. Right: Quantification of Dendra2⁺ mitochondria related flux of LC3 puncta (mito-LC3) and LAMP2 (mito-Lamp2). Dots depict individual cells. (F-G) Representative flow cytometry histogram overlay (left) and the quantification (right) demonstrate mitochondrial calcium levels as assessed by Rhod-2 AM staining in WT and Cx43^Δ/Δ (F) quiescent (0h) and (G) dividing (48h) HSC. Dots show independent experiment. (H-I) Flow cytometry histogram overlay (left) and the quantification (right) depict intracellular calcium levels as measured by Fluo-4 AM in WT and Cx43^Δ/Δ (H) quiescent (0h) and (I) dividing (48h) HSC. Dots show independent experiment. (J-K) Kinetic plot (J) and quantification (K) of mitochondrial Ca²⁺ (Rhod-2 AM intensity) in WT and Cx43^Δ/Δ steady-state HSC following treatment with EGTA, TG, and Ru265. n=3 independent biological replicates. (L) Mitochondrial calcium levels in WT and Cx43^Δ/Δ HSC loaded with Dendra2⁺ mitochondria isolated from WT or Cx43^Δ/Δ mito-Dendra2⁺ HSC/MPP. Data normalized with mitochondrial calcium levels in WT HSC supplemented with WT mitochondria and expressed as fold change. n=3 independent experiments. Data represented as mean ± SEM. TEM and immunofluorescence quantification data are pool of 2–3 independent experiments. p values were generated by unpaired, two-tailed t test except in figure B and L where One way ANOVA, Tukey’s multiple comparisons test was used. *p<0.05, **p<0.01, ***p<0.001.

To further dissect into the transcriptional wiring of Cx43^Δ/Δ cycling HSC, we assessed mitophagy by quantifying the recruitment of macro-autophagy proteins PTEN-induced putative kinase 1 (Pink1), the E3 ubiquitin ligase Parkin, oligomerized LC3 (puncta), and the autophagolysosome marker Lamp2 to mitochondria in WT and Cx43^Δ/Δ quiescent and dividing HSC. Subcellular colocalization analyses indicate that hematopoietic Cx43 deficiency enhances mitophagy, as evident by colocalization of Dendra2⁺ mitochondria with Pink1, Parkin, LC3 puncta, and Lamp2 in dividing, but not in quiescent HSC (Figures 4C–E, Supplemental Figures 4I). This mitophagy induction was not associated with activation of TFEB, a master regulator of lysosomal biogenesis and autophagy in HSC⁴¹ (Supplemental Figures 4J–K), suggesting the use of alternative autophagy inducers like FOXO1, a documented effect of Cx43 deficiency in HSC/MPP¹⁶. Together, these results suggest that while basal mitophagy is crucial for HSC/MPP regenerative potential and fate determination^28,42–44, Cx43 deficiency mediated hyperactivated mitophagy prevents cycling HSC from optimally engaging in metabolism, thereby limiting hematopoietic regeneration³³.

Ca²⁺ plays a crucial role in OXPHOS, ATP production, and mitochondrial activity, and is inversely correlated with HSC maintenance^30,45. Staining with Rhod-2AM, a fluorescence indicator of mitochondrial Ca²⁺ showed that the levels of mitochondrial Ca²⁺ (mCa²⁺) were elevated in both quiescent and cycling Cx43^Δ/Δ HSC (Figures 4F–G). In contrast, no significant change in intracellular Ca²⁺ (iCa²⁺) levels as assessed by Fluo-4AM staining was noted in both WT and Cx43^Δ/Δ quiescent and dividing HSC (Figures 4H–I), indicating that Cx43 selectively regulates mitochondrial, but not cytosolic, calcium homeostasis in HSC. To further characterize the role of Cx43 in Ca²⁺ homeostasis, we examined iCa²⁺ and mCa²⁺ flux in WT and Cx43^Δ/Δ HSC. Consistent with earlier findings, basal mCa²⁺ levels were elevated in Cx43^Δ/Δ HSC, while iCa²⁺ levels remain unchanged. Chelation of extracellular calcium with EGTA drastically reduced iCa²⁺ levels in both WT and Cx43^Δ/Δ HSC although residual levels of mCa²⁺ remained higher in Cx43^Δ/Δ HSC even when adding the sarcoplasmic/endoplasmic reticulum calcium/ATPase (SERCA) pump inhibitor thapsigargin (TG) and the mitochondrial calcium uniporter (MCU) Ru265, supporting the notion that Cx43 is required for complete mitochondrial calcium efflux. Restoration of normal calcium levels confirmed the accumulation of Ca²⁺ in the mitochondrial compartment, but not overall cellular compartment, of Cx43-deficient HSC (Figures 4J–K, Supplemental Figures 4L–M). Notably, exogenous WT, but not Cx43^Δ/Δ, HSC/MPP mitochondria loading restored mCa²⁺ levels of Cx43^Δ/Δ HSC/MPP, consistent with the level of heteroplasmy achieved (Figure 4L). Collectively, these results demonstrate the existence of a mitochondria-autonomous role of Cx43 in regulating mCa²⁺ homeostasis in HSC.

Cx43 regulates mitochondrial network remodeling in dividing HSC.

Dynamic modulation of mitochondrial networks is an integral component of mitochondrial quality control and plays a crucial role in preserving functional HSC pool both in homeostasis and in stress conditions^{14,27–32,46}. To investigate whether Cx43 regulates mitochondrial organization and integrity upon HSC activation, we performed live imaging of WT and Cx43^Δ/Δ HSC expressing Dendra2⁺ mitochondria. As Dendra2 can be photoconverted to a red fluorescent protein, we tracked photoconverted regions of interest. In WT HSC, photoactivated red mitochondria rapidly merged into the green-fluorescent mitochondrial network, indicating active mitochondrial fusion. By contrast, in Cx43-deficient dividing HSC, photoactivated red mitochondria fragmented into numerous smaller units, remained unfused, and were dispersed throughout the cytoplasm, indicating low mitochondrial fusion, but retained motility (Figures 5A–D, Supplemental Videos 1–2). We next examined whether Cx43 deficiency alters mitochondrial motility and Ca²⁺ levels, potentially triggering mitochondrial network remodeling. Expression and mitochondrial recruitment of mitochondria fusion mediator and Ca²⁺ homeostasis ⁴⁷, MFN2 and OPA1 was significantly reduced in dividing Cx43-deficient HSC but remained unchanged in quiescent HSC (Figures 5E–I, Supplemental Figures 5A–E, L). Additionally, the deficiency of Cx43 increases the frequency of HSC with “polarized” Opa1 following cell division, while no significant change in microtubule network formation was observed (Figure 5J, Supplemental Figure 5F). On the other hand, the activation of the mitochondrial fission mediator Drp1 (pDrp1-S616), a key regulator of mitochondrial fission⁴⁸, and its recruitment to mitochondria were significantly increased in dividing, but not in quiescent, Cx43^Δ/Δ HSC (Figures 5K–O, Supplemental Figure 5G–K). Conversely, the inhibitory form of Drp1 (pDrp1-Ser637), which promotes Drp1 detachment from mitochondria and suppresses fission, showed reduced expression and mitochondrial co-localization in dividing Cx43^Δ/Δ HSC compared with WT controls (Supplemental Figures 5M–O). The recruitment of activated Drp1 to mitochondria is regulated by the mitochondrial membrane adaptors Mff and Fis-1 ^40,49. Strikingly, no significant difference in Fis1 and MFF expression was observed among WT and Cx43-deficient cycling HSC (Supplemental Figures 5P–Q). We next investigated the role of Cx43 on mitochondrial dynamics and fate in in vivo stress hematopoiesis models following transplantation of WT and Cx43^Δ/Δ HSC and analysis at seven days post-transplant, when all the cells were cycling (Supplemental Figure 6A). Immunofluorescence imaging of proliferating Cx43^Δ/Δ HSC from regenerating BM revealed a marked decreased MFN2 intensity and mitochondrial localization, along with a concomitant increase in both intensity and mitochondrial localization of pDrp1 (Ser616) and total Drp1 (Supplemental Figures 6B–I). Notably, HSC from Cx43^Δ/Δ chimeric mice also displayed increased colocalization of mitochondria with LC3 puncta and Lamp2, indicating enhanced mitophagy (Supplemental Figures 6J–L). Together, these results demonstrate that Cx43 alter the localization of key mitochondrial dynamics protein, which disrupt mitochondrial morphology and appear to be associated with metabolic dysfunction.

Figure 5: Cx43 through regulation of mitochondrial dynamics regulates hematopoietic regeneration.

Representative images from WT and Cx43^Δ/Δ dividing HSC depicting the following:

(A-B) Time-lapse microscopy of Dendra2⁺ mitochondria before (0 min) and after (0.5–30 min) photoconversion. Subsets of Dendra2⁺ mitochondria (white boxed area) were photoconverted by 405 nm and both photoactivated (red fluorescence, rD2) and non-photoactivated (green fluorescence, gD2) mitochondria were tracked over time. (B) The intensity profile of red (rD2) and green (gD2) signals along lines from A (lower panel) are provided in B. The line analysis demonstrates mitochondrial fusion (co-localization of rD2 with adjoining gD2 mitochondria over time) in WT HSC. rD2 mitochondria does not overlap with adjoining gD2 mitochondria in Cx43^Δ/Δ HSC over time. (C) Each color-coded surface represents individual photoconverted mitochondria distribution over a time of 30 min. White arrow represents start (arrow tail; 0.5 min) vs end (arrowhead; 30 min) of photoconverted mitochondrial tracking. (D) Quantification of photoconverted mitochondria distribution over time (0.5–30 min) in WT and Cx43^Δ/Δ HSC. n=3–4 independent experiment. (E-G) Dendra2⁺ mitochondria (green) and Mfn2 (red) staining (E), quantification of Mfn2 intensity (F), and co-localization of Mfn2 with mitochondria (Mito-Mfn2) (G). Nuclei were counterstained with DAPI (blue). (i) Imaris surface-building algorithm was used to reconstruct mitochondrial surface (green) and spot-building algorithm was used to define Mfn2 inside (reconstructed big red spot; white arrow) and outside (reconstructed small pink spots) mitochondria surface. Dots depict individual cells. (H-I) Dendra2⁺ mitochondria (green) and Opa1 (red) staining (H), and quantification of total Opa1 spots (I). Nuclei were counterstained with DAPI (blue). Dots depict individual cells. (J) Quantification of the percentage of HSC with polarized Opa1. A reference point was positioned to the center of the nucleus (DAPI) and distribution of Imaris reconstructed Opa1 spots (red) were analyzed. For each data point 10–12 HSC were scored to calculate the percentage of Opa1 polarized distribution (n=3 independent experiment). (K) Dendra2⁺ mitochondria (green), S616-phosphorylated (active) dynamin-related protein 1; pDrp1 (S616) (red), and total Drp1 (purple) staining. Nuclei were counterstained with DAPI (blue). (i) Reconstructed mitochondrial surface were shown in green. Reconstructed big red and violet spots depict pDrp1 (S616) and total Drp1 inside mitochondria surface, respectively. Reconstructed small pink and yellow spots depict pDrp1 (S616) and total Drp1 outside mitochondria surface, respectively. (L-M) Quantification of pDrp1 (S616) (L), and co-localization of pDrp1 (S616) with mitochondria (Mito-pDrp1; S616) (M). Dots depict individual cells. (N-O) Quantification of total Drp1 (N), and co-localization of total Drp1 with mitochondria (Mito-total Drp1) (O). Dots depict individual cells. (P-R) WT and Cx43^Δ/Δ HSC were transduced with RFP-tagged empty or Drp1 dominant negative (Drp1^K38A) vector. (P) Left: Representative immunofluorescence image of Dendra2⁺ mitochondria (green) and Lamp2 (red) staining. (i) Imaris surface (for mitochondria) and spot (for Lamp2) building algorithm was used to define Lamp2 inside (big red spot) and outside (small pink spots) mitochondria surface (green). Nuclei were counterstained with DAPI (blue). Right: Quantification of Dendra2⁺ mitochondria colocalization with lamp2 (Mito-Lamp2). Dots depict individual cells. (Q) Representative immunofluorescence image depicting OGDH (red) staining (left) and quantification (right) in WT and Cx43^Δ/Δ HSC transduced with empty or Drp1^K38A vector. Nuclei were counterstained with DAPI (blue). Dots show individual cells. (R) ATP levels in empty or Drp1^K38A vector transduced WT and Cx43^Δ/Δ BM HSC. Data point depict individual experiments. (S) Experimental design for transplantation of Drp1^K38A vector transduced WT and Cx43^Δ/Δ mito-Dendra2⁺ HSC into lethally irradiated B6.SJL-Ptprc^a Pepc^b /BoyJ recipient mice and analysis. (T) Peripheral blood count of mito-Dendra2⁺ leukocyte and CD11b⁺ cells, and (U) BM frequency of donor derived mito-Dendra2⁺ LT-HSC, MPP2, and LSK cells in WT and Cx43^Δ/Δ chimeric mice containing empty or Drp1^K38A vector. Dots depict all individual mice, and two independent experiments were performed. Data represented as mean ± SEM. All immunofluorescence image quantification data are pool 2–3 independent experiments. Scale bars, 3 µm, 2 µm, 1 µm, and 0.5 µm. Statistical significance was assessed using unpaired, two-tailed Student t test except in panel D where two-way ANOVA and P, Q where one-way ANOVA with Tukey’s multiple comparisons test was used, respectively. *p<0.05, **p<0.01, ***p<0.001.

Cx43 regulates hematopoietic regeneration through regulation of Drp1-dependent mitochondrial fission.

Mitochondrial fusion and fission proteins directly interact with those regulating mitophagy. To investigate if disruption of Cx43 mediated mitochondrial fission machinery, through regulation of mitochondrial metabolism and mitophagy alter HSC function, we transduced WT and Cx43^Δ/Δ HSC with Drp1 GTPase dominant negative mutant Drp1^K38A. Our data shows that Drp1^K38A abolish Cx43 deficiency induced mitochondrial fragmentation and enhanced mitophagy (Figure 5P, Supplemental Figures 7A–D). Interestingly, overexpression of Drp1^K38A increased OGDH expression and ATP levels in Cx43 deficient HSC (Figures 5Q–R, Supplemental Figures 5A–B). Next, to determine if alleviating mito-fission can relieve the undesirable effects to Cx43 deficiency on hematopoiesis, we transplanted mice with Cx43^Δ/Δ HSC/MPP containing Drp1^K38A mutant (Figure 5S). Overexpression of Drp1 dominant negative Drp1^K38A mutant enhanced HSC function in vivo and improves donor derived leukocyte and CD11b count in the PB and BM content of LSK, colony-forming-unit (CFU) and HSC in Cx43^Δ/Δ chimeric mice by 40 days post transplantation (Figures 5T–V, Supplemental Figures 7E–F). Serial CFU replating assays revealed that inhibition of mitochondrial fission following Drp1^K38A overexpression improves self-renewal capacity of Cx43^Δ/Δ HSC/MPP (Supplemental Figures 7G–H). These findings indicate that Cx43 fine-tunes mitochondrial fission-mediated mitophagy in dividing HSC, and is necessary to preserve the metabolic integrity and regenerative capacity of HSC.

Cx43 activity on mitochondrial dynamics and hematopoietic regeneration is docking independent.

The activity of Cx43 depends on its contribution to hexameric inter-membrane channels or docking-independent hemichannels ^50,51. Three Cys residues located in each of its two extracellular loops control channel docking. The Cx43 carboxyl terminal domain plays a role in the trafficking, localization, and turnover of gap junction channels via numerous post-translational modifications and protein–protein interactions (reviewed in ⁵²). Structure-function analysis of Cx43 was performed by gene addition of docking deficient Cys-less (six point mutations in the extracellular loop where amino acid Cysteine replaced by Alanine) ⁵³ and gating deficient ΔCT₂₅₇ (C-terminus truncation at amino acid 257, resulting in impaired pH gating) functional mutants of Cx43 ⁵⁴ into Cx43^Δ/Δ HSC (Figure 6A). Re-expression of full length (FL) Cx43 (Cx43-FL) in Cx43^Δ/Δ HSC restored both OGDH expression and ATP production. Notably, the expression of Cx43 Cys-less mutant, but not of the ΔCT₂₅₇ mutant rescued OGDH expression and ATP levels in Cx43^Δ/Δ HSC (Figures 6B–D, Supplemental Figures 8A–B). Overexpression of Cx43-FL in WT HSC further enhanced OGDH expression, indicating a dose-dependent role for Cx43 in regulating mitochondrial metabolism (Figures 6B–C, Supplemental Figures 8A–B). PLA of Tomm20/Cx43 and OGDH/Cx43 further demonstrated that both Cys-less and ΔCT₂₅₇ mutants localized to the outer mitochondrial membrane and retained the ability to bind OGDH, albeit the proximity of Cx43 ΔCT₂₅₇ to OGDH was reduced (Supplemental Figures 8C–F), providing a mechanistic insight on the potential role of the C-terminus domain of Cx43 in mitochondrial OGDH stability. Next, we investigated the involvement of Cx43 structure and functional mutants in regulation of HSC mitochondrial dynamics and fate. Re-expression of either Cx43 FL and Cys-less mutant, but not ΔCT₂₅₇ mutant in Cx43^Δ/Δ HSC increased the expression of mitochondrial fusion regulator Mfn2 and Opa1, reduced the proportion of Opa1 polarized HSC, and decreased expression of mitochondrial fission regulator, pDrp1 (S616) (Figures 6E–H, Supplemental Figures 8G–K). Remarkably, the mitochondrial Ca²⁺ and AMPK phosphorylation (pAMPKα1, 2-Thr172) mirrored and only Cx43 functional hemichannels, but not the ΔCT₂₅₇ mutant prevented the Cx43 deficiency mediated increase (Figure 6I, Supplemental Figures 8L–M).

Figure 6: Cx43 activity on mitochondrial fission/fusion dynamics and hematopoietic regeneration is independent of docking.

(A) Graphical representation of Cx43 full length and the truncation mutation, cys-less and Δ CT₂₅₇ incorporated into EF1α-MCS-IRES-dsRED retroviral vector. (B-I) WT and Cx43^Δ/Δ HSC transduced with constructs coding for different forms of Cx43 (FL, cys-less or ΔCT₂₅₇) and analyzed. Representative immunofluorescence image (B) and quantification (C) depicting OGDH expression (red). Nuclei were counterstained with DAPI (blue). Dots show individual cells. (D) ATP levels in WT and Cx43^Δ/Δ BM HSC containing different Cx43 constructs (FL, cys-less or ΔCT₂₅₇). Data point depict individual experiments. (E) Immunofluorescence staining of mito-Dendra2⁺ (green) and Mfn2 (red) and (F) quantification depicting Mfn2 expression. Nuclei were counterstained with DAPI (blue). Dots depict individual cells. (G) Representative immunofluorescence staining of Dendra2⁺ mitochondria (green) and pDrp1 (S616) (red), and (H) quantification of pDrp1 (S616). Nuclei were counterstained with DAPI (blue). Dots depict individual cells. (I) Mitochondrial calcium levels (Rhod-2 AM) in WT and Cx43^Δ/Δ HSC containing empty vector or different Cx43 constructs (FL, cys-less or ΔCT₂₅₇). n= 4 independent experiments. (J) Schema depicting transplantation of WT and Cx43^Δ/Δ HSC containing different Cx43 constructs (FL, cys-less or ΔCT₂₅₇) into sub-lethally irradiated (2.5 Gy) congenic NOD-scid IL2Rgamma^null, NOD-scid IL2Rg^null mice and analyzed as indicated. (K) Peripheral blood count of mito-Dendra2⁺/dsRED⁺ leukocytes, neutrophils, and CD11b cells in NSG mice at indicated time post-transplant (*p<0.05, **p<0.01, ***p<0.001 vs Cx43^Δ/Δ Empty vector). Three independent experiments with 4 mice each were performed. (L) BM count for LT-HSC, MPP2, and LSK cell in NSG mice at indicated time post-transplant. Each data point depicts individual mice, and three independent experiments were performed. (M) HSC-enriched LSK BM cells from WT and Cx43^Δ/Δ chimeric mice (49 days post transplantation) containing different Cx43 constructs (FL, cys-less or ΔCT₂₅₇) were FACS sorted and mitochondrial OCR was measured by Seahorse extracellular flux analyzer. (N) Quantification summary of mitochondrial OCR (basal, maximal, ATP production, and Spare Respiratory Capacity; SRC) in WT and Cx43^Δ/Δ HSC/MPP containing different Cx43 constructs. All data points depict a pool of 4 mice HSC/MPP and three experimental replicates were performed. (O) WT and Cx43^Δ/Δ mito-Dendra2⁺ HSC transduced with constructs coding for different forms of Cx43 (FL, cys-less or Δ CT₂₅₇) were co-cultured over WT BM stroma for 16h. Flow cytometry histogram overlay (left) and quantification (right) demonstrate transfer of mitochondria from HSC/MPP transduced with Cx43 constructs to BM stromal cells. n= 4 independent experiments. Data represented as mean ± SEM. All immunofluorescence quantification data represent pools of 2–3 independent experiments. Scale bar, 2 µm. Statistical p values were generated by unpaired, two-tailed t test except in panel C, F, G where one-way ANOVA with Tukey’s multiple comparisons test was used. *p<0.05, **p<0.01, ***p<0.001.

To determine if Cx43 hemi-channel mediated changes in mitochondrial metabolism, dynamics, and fate affect HSC fitness, we assessed the repopulation capacity of WT and Cx43^Δ/Δ HSC transduced with Cx43 mutants (Figures 6J). PB analysis revealed improved leukocyte, neutrophils, and B lymphocytes reconstitution in mice receiving FL or Cys-less transduced HSC, but not ΔCT₂₅₇, across all time points. BM analysis did not reveal any detectable difference in chimerism; however, we observed a concomitant increase in BM cellularity and LT-HSC, LSK, and LK population in mice receiving FL or Cys-less transduced HSC, but not ΔCT₂₅₇ mutant (Figures 6K–L, Supplemental Figures 9A–E). Metabolic profiling of HSC/MPP from primary transplant mice confirmed that that only Cx43 functional hemichannels, but not ΔCT₂₅₇ mutant, restored the overall mitochondrial respiration (basal and maximal OCR, spare respiratory capacity, and ATP production) (Figures M-N). Consistent with this, the expression of the only Cys-less, but not the ΔCT₂₅₇ mutant, in Cx43^Δ/Δ dividing HSC the rescue the mitochondrial ROS, membrane depolarization, and transfer of mitochondria from HSC/MPP to the BM stromal cells (Figure 6O, Supplemental Figures 9F–G). Collectively, these findings reveal that Cx43 hemichannels, through their C-terminus domain, are necessary for calcium homeostasis, OGDH activity, AMPK activation and energy production, and HSC regenerative capacity, and their activity does not depend on hemichannel-to-hemichannel docking.

Discussion

We previously identified Cx43 as a regulator of hematopoietic regeneration through cell-contact dependent mitochondrial transfer ¹⁵ and ROS scavenging¹⁶. Here, we demonstrate that these effects are downstream consequences of the function of mitochondrial Cx43 hemichannels and are dependent on the role of the Cx43 C-terminus on preserving mitochondrial OGDH activity of HSC. Increased mito-fission in dividing Cx43-deficient HSC results in mitochondrial autophagy, loss of mitochondrial transfer, ROS accumulation, and eventually senescence or apoptosis¹⁶.

In hematopoiesis, AMPK acts as a negative regulator of regeneration¹⁵. Our group previously found that in vivo pharmacological inhibition of AMPK activity led to better hematopoietic fitness and recovery after transplantation¹⁵. Here, we provide a mechanistic link between AMPK activity and Cx43 hemichannel function, showing that homeostatic Cx43 activity inhibits AMPK activation during hematopoietic regeneration and prevents the loss of HSC self-renewal. The effect of Cx43 on mitochondrial activity depends on its mitochondrial expression since transplantation of heteroplasmic HSC/MPP in which WT mitochondrial heteroplasmy levels of ~30%, suffice to significantly prevent the HSC exhaustion after serial repopulation. Cx43 is required to maintain the mitochondrial bioenergetics of HSC and the function of the Krebs cycle through its role on stabilizing OGDH and ETC, including the respirasome and the SDH supercomplexes³⁶. Loss of Cx43 disrupts these complexes, leading to electron leakage and increased ROS production. Interestingly, these processes are specifically dependent on the C-terminus of Cx43, since the loss of the C-terminus of Cx43 suffices to phenocopy all the major hallmarks of Cx43 deficiency. The homeostatic effects of Cx43 on HSC activity do not depend on docking of Cx43 hemichannels.

Our analyses focus on stem cell transplantation associated hematopoietic regeneration. Alternative approaches include the analysis of response to fluorouracil (5-FU) administration^55,56. 5-FU administration induces apoptosis and senescence of Cx43-deficient HSC/MPP16. Unfortunately, 5-FU inhibits mitochondrial ferredoxin reductase⁵⁷, an enzyme essential for mitochondrial iron homeostasis and p53 dependent apoptosis⁵⁸, rendering confounding effects when trying to understand the mitochondrial activity of HSC.

Mitochondrial calcium homeostasis is crucial to maintain mito-fusion and prevent excess of fission in quiescent HSC²⁹ and increased mitochondrial calcium levels have been linked with HSC loss-of-function⁵⁹. While most of the calcium influx into mitochondria depends on the MCU, our data point out to a role of Cx43 hemichannels in facilitating mitochondrial calcium efflux under conditions of low extracellular calcium. In a cellular environment with zero extracellular calcium, Cx43 hemichannels are expected to exhibit a high open probability⁶⁰. Our calcium flux data support that Cx43 is required for complete efflux of calcium out of mitochondria.

The Cx43 carboxyl terminal domain plays a role in the trafficking, localization, and turnover of gap junction channels via numerous protein–protein interactions (reviewed in ⁵²). Several protein-protein interactions may play a major role in the mitochondrial phenotypes identified. First, Drp1 is a highly conserved GTPase that controls mitochondrial fission⁶¹. Upon mitochondrial division, Drp1 is recruited to the outer mitochondrial membrane to perform GTP hydrolysis-driven scission⁶¹. In cardiac cells, the C-terminus of Cx43 is internally translated to generate a small Cx43 isoform (Cx43–20 KDa), which is required for the trafficking of Cx43-FL to membranes⁶². Importantly, Cx43–20kDa can abandon its canonical Cx43-FL trafficking roles, localize to the mitochondrial outer membrane and favor mitochondrial elongation, preventing fission, preserving ATP production and generating less ROS⁶³. It is quite possible that the loss of the C-terminus in Cx43-deficient HSC/MPP favors an unbalance towards Drp1-dependent mitochondrial fission. Inhibition of Drp1 restores mitochondrial OGDH activity and ATP production, strongly supporting the notion that Cx43 C-terminus is a negative regulator of Drp1 activity in HSC/MPP and as such, its expression is essential for preserving their mitochondrial integrity. Second, Cx43 can constitutively downregulate macroscopic autophagy through a direct interaction of the Cx43 cytoplasmic loop (and C-terminus) with Atg16 and components of the autophagy initiation complex (Vps34, Beclin-1 and Vps15)⁶⁴ providing a direct line between Cx43 hemichannels and inhibition of mitophagy by repressing the autophagy initiation complex.

Finally, our data imply that ex vivo manipulation of HSC mitochondrial calcium may associate with prevention of HSC repopulation capacity and have clinical implications which include the administration of an AMPK inhibitor before transplantation or the ex vivo culture with inflammatory-signal deprived, cell-targeted mitochondria.

Key points:

1. Antioxidant activity of Cx43 depends on its autonomous activity in mitochondria of hematopoietic stem cells and multipotential progenitors.

2. Cx43 activity depends on hemichannels that control mitochondrial Ca²⁺, α-ketoglutarate dehydrogenase and electron transfer chain activity.