The review organizes diabetic retinopathy around the retinal neurovascular unit, the tightly coupled assembly of neurons, glial Müller cells, endothelial cells, pericytes and immune cells that keeps the retina supplied and its blood-retinal barrier intact. In the earliest phase, chronic hyperglycemia inflicts damage through oxidative stress, disturbed lipid metabolism and mitochondrial dysfunction. The mitochondria of retinal cells are particularly vulnerable: damaged mitochondrial DNA persists even after glucose control improves, a phenomenon the field calls metabolic memory. Studies cited in the review show that impaired removal of damaged mitochondria and weakened Nrf2-mediated antioxidant defenses allow this early injury to accumulate silently, weakening the neurovascular unit before any clinically apparent vascular change appears on an ophthalmologist’s chart.

As damage accumulates, the disease shifts into an inflammatory phase driven by a cast of signaling molecules that the review treats as a coordinated network rather than isolated culprits. Nuclear factor kappa B, high-mobility group box 1, Toll-like receptor 4, the receptor for advanced glycation end products, the cyclic GMP–AMP synthase–stimulator of interferon genes pathway, and the NOD-like receptor protein 3 inflammasome together amplify inflammation, promote leukostasis—the sticking of white blood cells to retinal vessel walls—and erode the blood-retinal barrier. Recent work highlighted in the review shows that cGAS/STING activation drives endothelial cell senescence in early disease and that STING-mediated immune activation in microglia worsens neurovascular unit damage. The NLRP3 inflammasome, meanwhile, triggers pyroptosis, an inflammatory form of cell death that further dismantles retinal endothelium.

Once enough capillaries have been lost, the retina becomes hypoxic, and the disease pivots again. Hypoxia-inducible factor 1 alpha and vascular endothelial growth factor signaling surge, driving the neovascularization that defines proliferative diabetic retinopathy. But the review emphasizes that VEGF is only one node in a much larger web. Angiopoietin-2, Wnt/beta-catenin, phosphoinositide 3-kinase/protein kinase B, semaphorin 4D, and endothelial metabolic reprogramming all sustain vessel leakage and abnormal growth. Hypoxic Müller cells secrete VEGF that induces matrix metalloproteinase activity in endothelial cells, while the HIF1α-PFKFB3 pathway rewires endothelial metabolism toward glycolysis to support proliferation. A 2024 study in Nature Communications even links protein O-GlcNAcylation to Hippo signaling as a driver of vascular dysfunction, illustrating how metabolic and mechanical pathways converge.

In advanced disease, the review identifies a final molecular pivot: the angio-fibrotic switch. Transforming Growth Factor Beta, Yes-associated protein and transcriptional coactivator with PDZ-binding motif pathways, and related signaling drive extracellular matrix deposition, fibrovascular membrane formation and traction on the retina. Connective tissue growth factor rises as VEGF falls, and endothelial cells undergo endothelial-mesenchymal transition, morphing into matrix-producing cells. Single-cell transcriptomic studies of fibrovascular membranes reveal unexpected participants, including microglia and the matrix protein AEBP1, suggesting that the scarring phase recruits immune and stromal cells in ways that earlier bulk analyses missed. This stage is what ultimately detaches the retina and causes irreversible vision loss, yet it remains the least pharmacologically addressed.

Against this mechanistic backdrop, the review evaluates the therapeutic landscape with unusual candor. Systemic metabolic control remains the single most powerful early intervention: long-term follow-up from the DCCT/EDIC trial showed durable reductions in retinopathy progression with intensive diabetes therapy. Fenofibrate, a lipid-lowering agent, has shown the most promise among drugs tested for early-stage disease, reducing the need for laser treatment in the FIELD study and slowing progression in the ACCORD trial. Anti-vascular endothelial growth factor therapy, including aflibercept, ranibizumab, bevacizumab and the bispecific antibody faricimab, remains the standard for severe vascular complications, with newer delivery systems such as the ranibizumab port delivery system extending dosing intervals. Corticosteroid implants offer alternatives for persistent macular edema.

Yet the review is blunt about the limits of current approaches. Anti-inflammatory, neuroprotective, anti-fibrotic, gene-based and multi-target strategies are promising in preclinical models but face significant barriers in delivery, safety, model-to-human translation and weak biomarker validation. Sustained intraocular VEGF neutralization, for instance, has been associated with retinal neurodegeneration in diabetic mouse models, raising the possibility that complete VEGF suppression carries its own risks. Gene therapy approaches such as subretinal delivery of RGX-314, tested in neovascular age-related macular degeneration, are being explored for diabetic retinopathy but remain early. The authors call for improved classification of molecules and clinically significant early endpoints to guide targeted therapy, noting that many candidate biomarkers have not been validated in prospective studies.

What makes the review notable is its insistence on pathway crosstalk and feedback loops rather than linear causation. NF-κB, for example, restricts inflammasome activation by promoting the elimination of damaged mitochondria, meaning that blocking NF-κB indiscriminately could paradoxically worsen inflammation. STAT3 activation in endothelial cells increases vascular leakage by downregulating tight junction proteins, but STAT3 activation in microglia increases pericyte apoptosis through a different signaling route, and STAT3 activation in circulating myeloid cells contributes to microvascular dysfunction through yet another mechanism. The same transcription factor, in other words, plays different pathogenic roles depending on which retinal cell type it inhabits—a level of specificity that the review argues must inform any serious therapeutic strategy.

The authors also distinguish carefully between clinical evidence and non-clinical studies, a distinction that matters enormously in a field where rodent models have repeatedly produced therapies that failed in human trials. The PKC beta inhibitor ruboxistaurin, for example, showed promise in phase 3 studies for vision loss but never achieved widespread adoption. The review calls for studies at all stages of the disease, arguing that early intervention targeting mitochondrial dysfunction or oxidative stress may prevent the downstream cascades that currently require invasive intraocular injections. Whether the field can translate this stage-by-stage signaling map into clinically deployable biomarkers and stage-appropriate treatments remains the defining question, but the review offers a framework that could finally match the right drug to the right phase of a disease that has stubbornly resisted one-size-fits-all solutions.

Subject of Research: Stage-dependent signaling networks and therapeutic strategies in diabetic retinopathy: from neurovascular dysfunction to angiogenesis and fibrosis

Article Title: Stage-dependent signaling networks and therapeutic strategies in diabetic retinopathy: from neurovascular dysfunction to angiogenesis and fibrosis

Article References: Liu, W., Wang, D., Wu, P., Peng, Z., Liu, Z., & Chen, X. (2026). Stage-dependent signaling networks and therapeutic strategies in diabetic retinopathy: from neurovascular dysfunction to angiogenesis and fibrosis. Molecular Biology Reports, 53(1), Article 1741. https://doi.org/10.1007/s11033-026-12925-1

Image Credits: AI Generated

DOI: 10.1007/s11033-026-12925-1

Keywords: Stage-dependent, signaling, networks, therapeutic, strategies, diabetic, retinopathy, neurovascular, dysfunction, angiogenesis, fibrosis, scientific research

The study, published in the open-access journal iScience, combined deep transcranial magnetic stimulation, or DTMS, with intermittent theta burst stimulation, an excitatory patterned protocol known as iTBS. Unlike conventional TMS, which uses a figure-of-eight coil to focus stimulation on a small patch of superficial cortex, DTMS employs specialized H-coils whose magnetic fields spread across broader cortical and subcortical territory. The H7 coil used in the trial engages the bilateral prefrontal cortex and the anterior cingulate cortex simultaneously, two regions that anchor executive control and conflict monitoring. Pairing this wide-field engagement with iTBS, which mimics endogenous theta-gamma coupling and efficiently induces synaptic plasticity in under four minutes, offered the researchers a way to modulate an entire distributed network rather than an isolated hotspot.

Nineteen right-handed healthy adults, seven men and twelve women with a mean age of about twenty-five, completed the randomized crossover trial. Each participant received both active stimulation and a sham condition in separate sessions separated by a two-week washout, with the order counterbalanced and assignments concealed in sealed envelopes. Before and after each intervention, the researchers recorded cortical hemodynamics with a portable functional near-infrared spectroscopy, or fNIRS, device. The sensor packed twenty-four light sources and sixteen detectors into fifty-two channels covering fourteen regions of interest, from the dorsolateral prefrontal cortex through primary and secondary motor areas to somatosensory and parietal language regions. Resting-state signals were captured over eight minutes, followed by a three-minute dual-task test in which participants walked at a comfortable pace, turned to avoid obstacles, and answered oral serial subtraction questions posed by the experimenters.

The behavioral result was striking. In a mixed-effects model controlling for baseline performance, the increase in the number of completed calculations was significantly greater after active DTMS than after sham stimulation, with a treatment effect of F(1,17) = 12.546 and p = 0.003. Participants who averaged roughly forty-eight calculations before stimulation completed about fifty-nine afterward, a jump of nearly eleven problems in three minutes. No significant period or sequence effects emerged, ruling out simple practice or ordering artifacts. Accuracy, already above ninety percent at baseline, did not differ significantly between conditions, though a trend favored active stimulation. The authors argue that a ceiling effect on correctness limited sensitivity, and that the gain in throughput therefore reflects improved processing efficiency rather than a speed-accuracy trade-off.

The neurovascular data told a complementary story. During the dual-task condition, mean oxygenated hemoglobin, or HbO, concentration rose significantly more after DTMS than after sham in the right secondary motor cortex, which encompasses the premotor and supplementary motor areas, and in the left somatosensory association cortex, with p values of 0.022 and 0.033 respectively. Deoxygenated hemoglobin fell significantly in the right secondary motor cortex, consistent with elevated neural activation. Whole-brain waveforms showed overall greater cortical activation following active stimulation, while resting-state activation did not differ between conditions. The authors interpret this task-specific pattern through the lens of state-dependent plasticity, the principle articulated by Silvanto and Pascual-Leone that neuromodulatory effects require specific brain states, such as task activation, to be expressed. The synaptic changes induced by iTBS may lie dormant until the network is recruited by actual cognitive-motor demand.

Correlation analyses added mechanistic texture. Changes in calculation quantity were positively correlated with HbO changes in the left secondary motor cortex and negatively correlated with activation in the right Wernicke area, the parietal region corresponding to the right supramarginal gyrus. Accuracy improvements rose with bilateral secondary motor cortex activation but declined with greater activation in the right primary motor cortex and right Wernicke area. Under the neural efficiency hypothesis, efficient networks activate critical regions precisely while suppressing interfering ones. The serial subtraction task is language-dependent and left-lateralized, so fluent performance plausibly demands focused allocation to the left language-motor circuit while suppressing right-hemisphere vigilance and error monitoring signals that would otherwise divert attention.

Perhaps the most intriguing finding concerns the secondary motor cortex as a relay station. After DTMS, interhemispheric activation correlations between the left and right dorsolateral prefrontal cortex, secondary motor cortex, and somatosensory association cortex were particularly strong, suggesting that excitatory stimulation may reduce interhemispheric inhibition and shift homologous regions from alternating suppression toward synchronous excitation. Left prefrontal and posterior parietal activation rose in parallel, and left secondary motor cortex hemodynamics increased alongside both. The authors propose that DTMS under the iTBS protocol strengthens the frontoparietal network, with the secondary motor cortex functioning as a transitional hub that transforms abstract cognitive commands from the prefrontal cortex and concrete sensory information from the parietal cortex into coordinated motor output. This bridging role is especially prominent in the left hemisphere, where the language and motor systems converge.

Exploratory functional connectivity analyses revealed several DTMS-induced changes, including enhanced connectivity between the left Broca area and right prefrontal, primary motor, and secondary motor regions during the dual task, but none survived false discovery rate correction, so the authors treat these as preliminary observations rather than definitive evidence of network reorganization. Baseline comparisons did show, after correction, that dual-task functional connectivity was significantly lower than resting-state connectivity, a pattern consistent with the neural efficiency hypothesis but also with alternative explanations such as task-dependent network segregation. Additional hemodynamic features indicated that cortical HbO response speed accelerated significantly in the bilateral prefrontal cortex and right Broca area after stimulation, and HbR response speed increased in several regions, with integral values mirroring mean concentration changes and confirming task-related activation levels.

The study has clear limits that the authors acknowledge candidly. There were no separate single-task walking or single-task calculation conditions and no quantitative gait measures, so the hemodynamic responses reflect combined cognitive and motor processes rather than isolated mechanisms. fNIRS captures cortical signals only, so the findings cannot directly establish whether DTMS modulates subcortical functional connectivity, even though prior work shows the H7 coil reaches the anterior cingulate. Blinding was procedural rather than formally verified, and the modest sample size of nineteen demands replication. Long-term effects of repeated deep stimulation on deep brain tissue, including glial responses and chronic inflammation, remain unexplored and warrant longitudinal multimodal imaging studies.

Even so, the trial offers a compelling proof of concept that a single session of wide-field excitatory neuromodulation can expand the computational capacity available for walking-and-thinking behavior in healthy brains, without degrading correctness. Because dual-task impairment tracks fall risk, reduced mobility, and cognitive decline in aging and neurological disorders, establishing a baseline framework in healthy adults is a prerequisite for translational work. If the state-dependent plasticity principle holds in older or clinical populations, future protocols might time stimulation to coincide with task engagement, individualize coil positioning, and combine larger cohorts with multimodal neuroimaging to determine whether the frontoparietal-secondary motor relay can be reliably strengthened. For now, the image of a brain that computes faster on its feet after a few minutes of patterned magnetic pulses is a vivid demonstration of how far noninvasive network modulation has traveled from the single-hotspot era.

Subject of Research: Neuroscience

Article Title: Deep TMS modulates neurovascular responses and enhances dual-task computation in healthy adults

Article References: Zeng, B., Wang, W., Ding, Z., Wang, C., Li, C., Yuan, R., Zhou, B., & Zhu, Y. (2026). Deep TMS modulates neurovascular responses and enhances dual-task computation in healthy adults. iScience, 29(11), Article 117586. https://doi.org/10.1016/j.isci.2026.117586

Image Credits: AI Generated

DOI: 10.1016/j.isci.2026.117586

Keywords: deep TMS, dual-task, fNIRS, neuromodulation, frontoparietal network, motor cortex, cognitive-motor integration, Deep, modulates, neurovascular, responses, enhances

The research tackled a long-standing puzzle in vascular biology: why women fare disproportionately badly as they age. Although both sexes experience arterial stiffening and neurological decline with age, women show a higher incidence of cardiovascular disease and many neurological conditions after menopause compared with age-matched men. The link between arterial stiffness and mortality is nearly twice as strong in women as in men, and cognitive decline progresses more steeply in elderly females than in elderly males. Because aging-related vascular stiffening can alter the blood–brain barrier and brain function, the investigators reasoned that the earliest changes in the vessels of the neck—the carotid arteries and jugular veins—might be tightly coupled to biochemical remodeling inside the brain.

To capture that coupling, the team combined three complementary technologies in a single cohort of healthy female C57BL/6 mice. Ultrasound imaging, performed with a high-frequency transducer on a Vevo 3100 system, allowed them to measure vessel diameters and blood velocities in living animals, from which they calculated wall shear stress, circumferential cyclic strain, pulsatility index, and volumetric flow in the carotid artery. Histology then quantified the balance of elastin and collagen—the two structural proteins that determine whether an artery snaps back like a rubber band or behaves like a rigid pipe. Finally, two mass spectrometry approaches, liquid chromatography mass spectrometry and matrix-assisted laser desorption ionization mass spectrometry imaging, charted the lipid landscape of the carotid artery, jugular vein, and brain, mapping exactly where age-sensitive lipids reside in tissue.

The ultrasound results were unambiguous. Middle-aged females, at 52 weeks of age, showed a roughly 23.5 percent drop in average carotid velocity during systole compared with 12-week-old young females, and a 27.8 percent reduction in wall shear stress during systole, a difference that was highly statistically significant. Average volumetric flow, normalized to body weight, plummeted by about 55.2 percent with age. Because the reduction in shear stress was driven primarily by the decline in systolic velocity, the authors attribute it to age-related changes in vascular compliance and cardiac output. Circumferential cyclic strain and pulsatility index trended downward but did not reach statistical significance, while the jugular vein diameter grew about 11.6 percent larger in the older animals—evidence that even veins remodel with age.

Under the microscope, the structural counterpart of those functional losses came into focus. The elastin-to-collagen ratio in the carotid artery fell in middle-aged mice relative to young controls, a shift that reflects the fundamental biomechanics of aging arteries. Elastin fibers, which allow vessels to stretch and recoil, replenish at an extremely low rate throughout life; as they dwindle, stiffer collagen fibers accumulate in relative terms. A lower elastin-to-collagen ratio means diminished arterial compliance, limiting the vessel’s ability to dampen the pulsatile pressure generated by each heartbeat. When large arteries lose that damping capacity, the delicate microvasculature surrounding the brain is exposed to higher pulsatile energies than it can safely withstand, a mechanism that previous work has linked to damage of small vessels and the blood–brain barrier. Notably, the jugular vein showed the opposite trend, with a modest increase in its elastin-to-collagen ratio, suggesting that arteries and veins age along distinct structural trajectories.

The lipidomic data added a molecular dimension to this picture. In the brain, ten lipid headgroups—including hexosylceramides, phosphatidic acid, phosphatidylcholine, phosphatidylethanolamine, phosphatidylglycerol, phosphatidylinositol, phosphatidylserine, and sulfated hexosylceramides—decreased significantly with age, while N-acylethanolamines increased, possibly as a compensatory response to age-related stress or inflammation given their known neuroprotective roles. These reductions matter because the brain is among the most lipid-rich organs in the body. Phospholipids protect cells from damage, sphingolipids stabilize the myelin that insulates nerves, and low levels of these molecules can signal neuronal vulnerability. In the carotid artery, cardiolipins and lysophosphatidylethanolamines rose with age, changes the authors link to mitochondrial dysfunction and membrane remodeling within the vessel wall. A triglyceride species, TG 45:9, accumulated in the aged carotid—consistent with epidemiological evidence that triglycerides are the blood lipid most strongly associated with arterial stiffness—while the signaling lipid DG 55:10 declined, hinting at impaired endothelial responsiveness.

Perhaps the study’s most consequential technical insight came from comparing whole-brain and region-specific analyses. When the researchers ran an unsupervised principal component analysis on mass spectrometry imaging data from the entire brain, young and middle-aged animals overlapped almost completely, with only a faint suggestion of age-related variation. But when the same analysis was restricted to the hippocampus—the memory center known to be exquisitely sensitive to metabolic and mitochondrial stress—the two age groups separated cleanly, with minimal overlap. The dominant contributor to that separation was a cardiolipin-associated ion at mass-to-charge ratio 1489.73, which diminished in the middle-aged hippocampus. Cardiolipins are essential components of mitochondrial membranes, and their loss fits established models of mitochondrial dysfunction in brain aging. The lesson is methodological as much as biological: bulk analysis of the whole brain masks regional lipid remodeling that may drive cognitive decline, and imaging-based approaches can expose changes that homogenized tissue cannot.

Tying everything together, Pearson correlation analysis revealed moderate to strong associations—correlation coefficients exceeding 0.6 in magnitude—between vascular biomechanical measures such as wall shear stress and circumferential cyclic strain and specific lipid classes across tissues, including glycerolipids, phospholipids, and sphingolipids. The authors are careful to stress that these correlations do not establish causality, and because neuronal and glial endpoints were not directly measured, any link between vascular changes and downstream neurological consequences remains speculative. They also acknowledge limitations: the effect sizes were relatively small, the 52-week-old mice represent a middle-aged rather than truly aged cohort, and even careful use of anatomical landmarks could not eliminate all positional variability in imaging and biochemical sampling. More advanced aging or emerging pathology may be required to produce larger, more disruptive shifts in vascular metrics.

Even with those caveats, the study delivers a compelling integrated portrait of neurovascular aging in females. Declining blood flow, falling wall shear stress, stiffening arterial walls, and wholesale lipid remodeling in both vessel and brain appear together in otherwise healthy animals, well before overt disease. That timing matters: if neck vascular dysfunction genuinely precedes brain biochemical alteration, as related work from the same group suggests, the vasculature may offer an early warning system—and a potential intervention target—for the cognitive diseases that disproportionately afflict aging women. By establishing this baseline in disease-free mice, the team has laid groundwork for future mechanistic studies in models of neurodegeneration, and has underscored a point increasingly hard to ignore: brain aging cannot be understood without also understanding the blood vessels that feed it.

Subject of Research: The effects of aging on neck vascular function, vascular and brain lipid composition, and brain biochemistry in female mice.

Article Title: Impact of aging on neck vasculature and brain biochemistry in female mice

Article References: Jones, A. R., Jarrahi, A., Karpowich, K., Moody, E., Renner, L., Brown, L. P., Tressler, C. M., & Crouch, A. C. (2026). Impact of aging on neck vasculature and brain biochemistry in female mice. Physiological Reports, 14(17), Article e71059. https://doi.org/10.14814/phy2.71059

Image Credits: AI Generated

DOI: 10.14814/phy2.71059

Keywords: aging, carotid artery, wall shear stress, arterial stiffness, lipidomics, mass spectrometry imaging, hippocampus, cardiolipin, female mice, neurovascular, sphingolipids, brain biochemistry