The pathophysiologic cause(s) of cognitive impairment leading to dementia remains unclear. Accumulations of amyloid-beta, neurofibrillary tangles with hyperphosphorylated tau, selective neuronal loss, as well as poorly classified ischemic brain lesions are all implicated, yet appear to incompletely explain the pathophysiology. Recent studies suggest that cerebral microinfarcts, in particular, cortical microinfarcts (CMIs), are another type of brain injury related to cognitive decline and dementia syndromes. In part because of their small size, comparatively little is known about how these lesions might contribute to cognitive decline and dementia. Cerebral microinfarcts are not visible in current neuroimaging resolution limits, only recognized on microscopic examination (1). Even with the ultra-high-resolution 7.0 T magnetic resonance imaging (7 T MRI), only lesions larger than 0.5 mm can be visualized in vivo (2), and current clinically utilized MRI protocols will miss nearly all CMIs (1). The size of a microinfarct is generally defined as “not visible with the naked eye” or “only visible upon light microscopy,” but the size of a microinfarct is generally accepted as being 50–400 μm and described as gliotic or cystic lesions with neuronal loss and tissue pallor that may occur in all brain regions, although the cerebral cortex may be a preferential location (3). Thus, detailed pathologic analysis of these lesions is critical to understanding how they contribute to cognitive decline.

Cortical microinfarcts are present in 6.3–30% of aging brains and even more commonly in those with dementia (36.5%) (4, 5). CMIs are associated with both subtle signs of cognitive decline and dementia (6). Moreover, these small lesions also increase the risk for recurrent ischemic and hemorrhagic stroke (7). Although the importance of CMIs in cognitive impairment, dementia, and overall brain health is increasingly recognized, the underlying pathophysiology of CMIs is not completely understood. Recent efforts to model these lesions in rodents suggest that in part their effect is cumulative and results from a decrease in regional neuronal network activity (8). In part, the cumulative effect of CMIs may be a consequence of expanded regional injury in the form a penumbral field where damage to axons of surviving neurons impairs neuronal connectivity. Building on prior work that demonstrated alterations of the molecular organization of axons in regions adjacent to human white matter lacunar and microinfarcts (9), in this study, we hypothesized that similar changes were present in nodal, paranodal, and axon initial segments (AISs) adjacent to human CMIs. The findings suggest that CMIs are indeed associated with an expanded region of impaired axonal microdomain organization with a significant loss of AISs in surviving neurons and disruptions of adjacent nodal and paranodal segments in cortical regions adjacent to the microinfarct. This supports the concept of a penumbral region adjacent to human CMIs in which progressive neuronal and axonal injury could be blunted to mitigate the effect of CMIs on cognitive impairment.

To determine the molecular organization of axons adjacent to CMIs, we selected 5 clinical cases with pathologically identifiable microinfarcts from a cohort of 18 cases in which the neuropathology report included the term “microinfarct.” This reflects an incidence rate of CMIs of 28.1% (18/64) in the larger patient cohort available for review. Age-matched control subjects were identified from the neuropathology database available at our institution. The age, sex, infarct location, post-mortem interval, infarct size, and medical comorbidities of each case are presented in Table 1. The patients selected ranged in age from 72 to 93 years. All of the subjects carried a pre-mortem diagnosis of dementia.

Cortical microinfarcts were defined pathologically by the presence of a focal necrotic cystic cavity by hematoxylin and eosin staining. In selected cases, clinically available MRI scans were reviewed, and the lesion was visible and therefore co-registered with the pathologic specimen (Figure 1). Hematoxylin and eosin staining of each of the CMIs studied is presented in Figure S1 in Supplementary Material, and each demonstrate the classic hallmarks of infarction with a necrotic core, tissue pallor, and an elliptical or wedge-shaped pattern consistent with infarction secondary to occlusion of a penetrating arteriole from the cortical surface. The area (in square millimeter) of each defined microinfarct was measured and is presented in Table 1. All of the infarcts were less than 1 mm² in area. To examine the peri-infarct tissue in a systematic way that would account for potential changes at different cortical depths, we identified six peri-infarct regions of interest adjacent to each microinfarct measuring 375 µm³ (Figure S2 in Supplementary Material). Axonal microdomains were immunolabeled for Ank-G to identify nodes of Ranvier and AISs as well as caspr to label paranodes. In age-matched control cortex, normal labeling of nodes and paranodes (Figure 2A) and robust labeling of AISs within corresponding regions of cortex was observed (Figure 2B). In cortical tissue adjacent to microinfarcts, AISs could be identified (Figures 2C,D; upper insets) but were decreased in density by 87% compared to controls (p = 0.008 using Student’s t-test, Welch’s correction; Table 3). The number of node/paranode complexes identifiable in cortical tissue was similarly decreased by 86.2% compared to controls (Figures 2C,D; Table 2). Closer examination of the peri-infarct regions adjacent to CMIs demonstrated profound disorganization of paranodal segments in tissue both immediately adjacent to the infarct and in peri-infarct tissue 100–200 µm away from the infarct core (Figure 3). Very few recognizable node/paranode complexes could be identified in peri-infarct tissue. Beyond this pattern of disorganization, a frequent finding included paranodal segments with “empty” nodes that were not stained for Ank-G (inset, Figures 3B,C). Occasional AISs were identifiable in peri-infarct tissue but were rare (Figures 3C,D). In regions of cortical tissue away from the microinfarct, we observed typical patterns of nodal, paranodal, and AIS staining, indicating that these effects were specific to the peri-infarct regions adjacent to CMIs.

A key determinant of the functionality of the AISs is segment length, which correlates with firing potential (10, 11). Therefore, we measured AIS length in remaining AISs adjacent to CMIs. There was a modest increase in AIS length by 21.7% although this was not significantly differently from control. Similarly, paranodal length has been shown to be an indicator of axoglial injury (9, 12–14). The length of measured paranodes within the cortical tissue adjacent to CMIs demonstrated at 30.2% increase in length compared to controls (p = 0.077 using Student’s t-test, Welch’s correction; Table 3). The disorganized pattern of caspr staining within the peri-infarct regions with frequent “empty” nodes complicates this analysis, as we limited our paranodal length measurements to node–paranode complexes in which both staining elements were positive, so as to limit the effect of measuring paranodal segments affected by tissue sectioning.

Molecular disorganization of specific proteins within the axon, such as ankryin G or caspr, may reflect loss of myelin or other structural disruption of the axons (15–19). To examine the structural integrity of myelin and axons adjacent to CMIs, we labeled serial sections from each case for MBP and NF200 (Figure 4). While there is some evidence of denuded axons immediately adjacent to the infarct core (Figure 4B), high magnification images at increasing distances from the infarct core (Figures 4C,D), where we had examined axonal microdomain organization, revealed relative preservation of myelin and axons as evidenced by the close co-localization of MBP and NF200 staining.

The relationship between cerebrovascular disease (CVD) and Alzheimer’s pathology has become increasingly complex. Careful pathologic inspection of sporadic cases of Alzheimer’s disease brains reveals microinfarcts including the presence of CMIs as a frequent finding (20). The extent to which these tiny lesions contribute to the cognitive dysfunction of dementia remains unknown. Here, we show that within the surviving cortical tissue immediately adjacent to CMIs, the functional molecular organization of axons is disrupted. Penumbral tissue adjacent to CMIs is marked by a profound loss of AISs, a loss of node/paranode complexes, and “empty” nodal domains. These changes occur in the absence of changes in the basic structure of myelinated axons within the penumbral region. These findings provide support for the concept of a microinfarct penumbra that can extend beyond the focal loss of neurons associated with the primary infarct to jeopardize cortical connectivity. Moreover, they provide further evidence that axoglial interactions are exquisitely sensitive to ischemic injury and fertile ground for the development of neural repair strategies targeting axonal regeneration (21).

Advances in high-resolution neuroimaging studies have resulted in increasing recognition of the many manifestations of CVD that go well beyond classic multi-infarct dementia (22). These include cortical, white matter, and deep nuclei microinfarcts detected on high-resolution structural imaging as well as disruptions of white matter connectivity evidenced by diffusion tensor imaging (6, 23). This more complete spectrum of CVD and its pathologic correlates demand greater attention to the neurobiologic substrates that result from these injuries. Recent studies have shown that unlike cortical microhemorrhages, which could be partially detected using routine clinical MRI (1.5 T), CMIs were uniformly not detected and only identifiable on pathologic examination of brain specimens (1). Such pathologic specimens therefore provide a unique opportunity to improve our understanding of the effect of these lesions. Furthermore, the identification of therapeutics targeting repair of these lesions necessitates knowledge of the molecular events occurring around them. Notably, the 200-µm peri-infarct region examined in this study correspond exactly to the peri-infarct region most active in axonal sprouting and neural repair after focal cortical stroke in rodents (24–27). Animal modeling of CMIs has demonstrated support for a micropenumbral region adjacent to CMIs with the occlusion of multiple neighboring penetrating arterioles leading to microinfarct enlargement (8), indicating that the immediately adjacent cortical tissue is vulnerable to additional injuries with the axon being particularly vulnerable to partial injury.

The AIS is a complex and unique axonal microdomain that essentially functions as an on–off switch for the neuron (28). At the AIS, dendritic and cell body input signals are integrated and modulated by axo-axonic synapses to control the firing rate of an individual neuron. Diminishing the length of the AIS in vivo results in fewer voltage-gated sodium channels at the AIS and therefore a higher firing threshold (29). Ank-G has been shown to be the critical regulator of AIS formation (30), and acting together with beta-IV spectrin, it functions to anchor voltage-gated sodium channels (Na_V) to the AIS membrane (31). In rodents, the AIS has shown exquisite sensitivity to direct ischemic injury that in part is driven by sensitivity to the activation of calpain (32). In this context, AIS loss in the penumbral regions of cortical tissue adjacent to human CMIs is not surprising. This study is the first to report such sensitivity of the AIS to partial ischemia in humans. In severe brain ischemia, calpain activation results in degradation of the structural elements of the axon including neurofilaments (33); however, we did not observe evidence of significant axonal degeneration in micropenumbral regions adjacent to CMIs nor did we observe complete absence of AIS in these regions. This suggests a more subtle disruption of the AIS in penumbral neurons adjacent to microinfarcts and further highlights the selective vulnerability of some cortical neurons.

Myelinated axons are predominantly associated with white matter, although in lesser density are readily identifiable throughout the cortex with some regional variation and increases in density in lower cortical layers. Here, we show the presence of mostly myelinated axons in the cortex, although smaller unmyelinated axons may be lost during post-mortem intervals and during tissue processing. They may also not stain robustly with NF200 and therefore be below the level of detection of this staining and imaging approach. Regardless, the function of myelinated axons is to speed neural transmission by clustering Na_V at nodes of Ranvier facilitating saltatory conduction. Disruption of the normal molecular organization of axons at nodal and paranodal regions is associated with axoglial pathologies including ischemia (34, 35). This disruption primarily results in the loss of nodal components and lengthening of paranodes due to lost cell–cell adhesion between oligodendrocytes and axons as we observed adjacent to CMIs. The observed decrease in nodal/paranodal complexes in the absence of significant structural changes in myelinated axons indicates that reliable, high-fidelity axonal conduction in cortical tissue adjacent to CMIs is not possible. Similarly, paranodal lengthening within penumbral regions indicates that axoglial interactions at the paranode are jeopardized in these regions, implying their sensitivity to partial ischemia. Modifications of the extracellular matrix after cortical ischemia may play an important role in stabilizing axoglial interactions (36). A limitation of this post-mortem human study is that because this peri-infarct region is known to be an active participant in axonal sprouting and neural repair, altered axoglial interactions could reflect incomplete myelination of newly sprouted axons. The phenomena of peri-infarct myelination during ischemic recovery deserves further study with animal models; however, these authors favor that the observed changes in axonal microdomain organization at nodes and paranodes adjacent to CMIs reflect an injury signal rather than evidence of a partial reparative phenomena.

The emergence of CMIs as a contributor to cognitive dysfunction in dementia and a previously underrecognized manifestation of cerebral microvascular disease highlights the need to understand the pathology of these lesions in the context of known principles of neural repair after stroke. A penumbral region adjacent to CMIs marked by neurons with diminished firing potential and disrupted axonal communication may exacerbate the cognitive effect of these tiny cerebrovascular lesions. Further identification of the molecular events active within this penumbral region may help to generate novel therapeutic strategies to repair the brain after cortical microinfarction.

HC: had conducted this project and drafted the manuscript under the supervision of JH. ST: assisted HC in doing immunohistochemical and immunofluorescence stainings. HV: contributed to this project with his expertise in neuropathology particularly identifying lesions and case selection and he also edited the manuscript. BY: reviewed the MRI images of the subjects. JH: supervised the project from beginning to drafting and editing the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.