The fovea is a specialized retinal feature of humans and most non-human primates that enables high-acuity and color vision. This region is characterized by densely packed cone photoreceptors within a foveal avascular zone (FAZ) and an excavation of inner retinal neurons, creating a shallow pit that is visible in the ocular fundus (Provis et al., 2013; Bringmann et al., 2018). Similar to non-foveate species, human retinal development exhibits overlapping fetal epochs of retinogenesis, neurogenesis, and synaptogenesis (Cepko, 2014; Hoshino et al., 2017; Zhang et al., 2020). Foveal neurons that express key molecular markers earlier than peripheral neurons are still immature at birth (Hoshino et al., 2017). Crowding of cone inner segments into the foveal center, together with migration of inner retinal neurons out of the pit, is a remarkable process starting late in fetal life and extending beyond birth (Hendrickson, 2005; Hendrickson et al., 2012). Thus, foveal structure and function are susceptible to perturbation by pre-term birth.

Foveal maturity can be assessed in clinic populations with optical coherence tomography (OCT), a depth-resolving imaging method that reveals layers of retina and choroid (Hendrickson et al., 2012; Thomas et al., 2020). Due to extended foveal development, a neonate born at term may exhibit inner retinal layers deployed across the pit and fewer than normal outer retinal reflective bands representing photoreceptor and retinal pigment epithelium (RPE) (Maldonado et al., 2011; Dubis et al., 2012). Imaging hallmarks of a premature birth include small FAZ, small pit volume, and at the foveal center, thickened inner retinal layers relative to the outer nuclear layer (Mintz-Hittner et al., 1999; Bowl et al., 2016, 2018b; Sjöstrand and Popovic, 2020). This constellation has been called macular (Bowl et al., 2016) or foveal (Sjöstrand et al., 2017) developmental arrest and occurs independently of a more severe condition, called retinopathy of prematurity (Mangalesh et al., 2022). Arrest is accompanied by reductions in visual acuity, photopic sensitivity, and electroretinographic response (Fulton et al., 2005; Akerblom et al., 2016; Bowl et al., 2016; Altschwager et al., 2017).

While neurons and associated glial cells migrate away from the pit, a small population of glia, potentially significant, remains at the pit center. Classic Müller glia have somata in the inner nuclear layer, spanning from internal to external limiting membrane, and intercalating among neurons to provide diverse services such as structural support, water homeostasis, retinoid recycling, and sequestration of a characteristic yellow pigment (xanthophyll carotenoid) (Wang and Kefalov, 2009; Daruich et al., 2017; Bringmann et al., 2018; Pollreisz et al., 2020). An inverted cone-like plug of glial tissue on the foveal floor was observed in early electron microscopy studies (Yamada, 1969). Named a “Müller cell cone,” this feature was considered by ophthalmologists as a repository for xanthophyll pigments and a participant in tractional disorders of the inner retinal surface (Gass, 1999). Accordingly, in a single-section transmission electron microscopy of one human and one macaque fovea, Syrbe et al. described a population of 25–35 large glial cell bodies in the foveal center that seemed to have little contact with adjacent neurons (Syrbe et al., 2017; Bringmann et al., 2018). How such specialized glia might participate in foveal development has not been studied.

With the long-term goal of characterizing the human fovea at the nanoscale, we acquired retinal tissue from an adult who was born pre-term, exhibiting features of arrested development. To document Müller glial morphology, abundance, and neural contacts in this fovea, we 3-dimensionally reconstructed glia using volume electron microscopy. This technique permits an unprecedented view of these elaborate, large, and numerous cells and builds on seminal two-decade-old studies of macaque retina that reconstructed glia manually (Burris et al., 2002; Ahmad et al., 2003). Volume electron microscopy has been recently used to detail neural circuitry and development of the fovea-abundant midget system for color and acute vision (Wool et al., 2019; Zhang et al., 2020). Herein, we identify a non-canonical glial phenotype, localized entirely within the inner retina, with cell bodies in the inner nuclear layer. These cells may be a heretofore unrecognized component of the incomplete inner retinal migration of prematurity.

In this adult human retina, the presence of inner retinal layers in the foveal floor and an undersized avascular zone (Figure 2) supports a family-reported clinical history of a pre-term birth at 28 weeks. Our major finding is a novel class of cells with many characteristics of Müller glia yet do not extend toward the ELM like the classically defined cells. These cells looked exactly like other reconstructed cells, except for lacking an outer process. Features consistent with long-standing descriptions of Müller glia included abundant microfilaments, space-filling between neurons, presence of cell bodies in the INL, side branches, and pericellular baskets. In the human retina, previous visualizations of the foveal center Müller glia have shown only parts of cells (see next paragraph). In various species, individual Müller glia in their entirety have been visualized in Golgi-stains, peroxidase-injection, physical isolation, and artistic conceptions (Polyak, 1941; Robinson and Dreher, 1990; Franze et al., 2007; Bringmann et al., 2018). Using volume electron microscopy to fill this gap, we could distinguish inner cells clearly from those that could be traced through the volume to near the ELM, between cone nuclei, and in parallel with cone outer fibers. One limitation of our study is the lack of the ELM itself in the tissue block. Thus, we could not calculate thickness ratios of the inner and outer retina for direct comparison to clinical studies of pre-term birth retinas. These document persistent inner retinal layers and small FAZ and pit volume (Mintz-Hittner et al., 1999; Bowl et al., 2016, 2018b; Sjöstrand and Popovic, 2020) in childhood up to early adulthood ages (Bowl et al., 2016; Sjöstrand et al., 2017; Sjöstrand and Popovic, 2020) comparable with our donor.

One straightforward interpretation of these data is that inner Müller glia represent a form of arrested development in a subpopulation of cells that were retained in the foveal pit with neurons that did not migrate laterally. All retinal cells develop from progenitor cells in an evolutionarily conserved birth order, in which ganglion cells, cones, and bipolar cells are born in the first half of development, Müller glia and rods in the second half, and amacrine cells in both halves (La Vail et al., 1991; Cepko, 2014). In addition to this temporal sequence, in human and foveate non-human primates, neurons and glia are generated in a spatial sequence starting near the fovea first and then spreading across the periphery (La Vail et al., 1991, Cepko, 2014). Thus, we can surmise that a small number of foveal Müller glia are still being born or being positioned into the INL when post-receptoral neurons have begun lateral migration to form a pit. However, in this specimen, peak cone density was within the normal range (Curcio et al., 1990; Zhang et al., 2015), and Müller glial trunks intertwined with cone cell bodies and outer fibers did not raise suspicions. Thus, we cannot exclude the possibility that inner Müller cells also exist in normal retinas, and their abnormality in this pre-term birth is a lack of migration and not a lack of connection to the ELM.

What is the relation of inner Müller glia to a tissue mass (“Müller cell cone”) in the foveal center in adults of normal term birth? This plug of tissue, shaped like an inverted cone, was revealed by the first electron microscopic examination of human fovea in 1969 and later given its current name (Yamada, 1969; Gass, 1999). Within the foveal floor of presumably normal adults, within a distinct cytoplasmic mass, others have observed nuclei (Hendrickson et al., 2012; Delaunay et al., 2020) or cells negative for the glio-neuronal transmitter recycling enzyme (glutamine synthase) and positive for GFAP (Nishikawa and Tamai, 2001; Delaunay et al., 2020; Reichenbach and Bringmann, 2020). The latter was suggested to represent either astrocytes or Müller glia. Based on our discovery of inner Müller cells among the unmigrated neurons of this preterm birth retina, we hypothesize that the Müller cell cone in a normal term retina represents a small population of glia at the foveal center that has not migrated. In a normal retina, where all foveal neurons have migrated laterally, no neurons or plexiform layers remain for interactions with these Müller cells, which remain relatively undifferentiated. Outside the foveal floor glial mass, both inner and outer Müller cells contact with the ILM (Supplementary Video S2). This tissue layer, comprised of the vitreal aspect of Müller end feet and basal lamina, was very thin in this specimen as it is in a normal fovea (Syrbe et al., 2017).

Cell bodies in the Müller cell cone (Syrbe et al., 2017; Bringmann et al., 2018) were called a second Müller glial type and were not obviously related to our inner Müller glia due to lack of information about connection with ELM. This small group of cells (<35) was considered glia due to a paucity of mitochondria, evenly distributed heterochromatin, and abundant smooth endoplasmic reticulum, compared to the adjacent neurons, and they were inserted into the internal limiting membrane. These foveal floor cells were said to project to the ELM, but data were not provided to support this conclusion due to the use of single section transmission electron microscopy. Thus, it was not possible to determine whether a glial population lacking connections to the ELM was present or absent. Relative to the specimens of Bringmann et al., our foveal specimen differed in several aspects, in addition to visualization techniques and the absence of a pit. Ours also had a thick ONL throughout the tissue volume, scattered pedicles and continuous INL across the foveal floor, and a small umbo in the ganglion cell layer. In contrast, in the Bringmann specimens, all layers internal to the ONL were absent and the ONL itself was thinned in the center [Figure 1 of (Syrbe et al., 2017); Figure 10 of (Bringmann et al., 2018)). Thick and thin ONL configurations have been illustrated previously (respectively (Rochon-Duvigneaud, 1943, Kadomoto et al., 2021) vs. Figure 42 of (Polyak, 1957)] and Figure 8 of (Govetto et al., 2017). These configurations may represent a continuum of developmental lateral migration which also affects cone cell bodies in the ONL or differences in how a tiny foveal center is sampled with single vertical sections. Because lateral migration of neurons creates space for the large pedicles of cone photoreceptors as the smaller inner segments squeeze together toward the foveal center (Schein, 1988; Packer et al., 1990), it has been suggested that delayed inner retinal migration may yield miniature (2.1–4.5) pedicles; these, in turn, may degrade signal transmission (Sjöstrand et al., 2017; Sjöstrand and Popovic, 2020). We did find non-migrated pedicles in the fovea, however, these were typically large (Haverkamp et al., 2000).

In envisioning the foveal center Müller glia of preterm birth in three dimensions, we also estimated their abundance. At the level of the INL, inner cells were 1.7 times more numerous than outer cells (41,872 cells/ mm² vs. 24,631 cells/ mm²). Furthermore, outer Müller glia were found to outnumber cones at the level of the cone outer fibers in the ONL by 1.8-fold (221,448 cells/ mm² vs. 123,026 cells/ mm²). Because neurons and glia were vertically aligned in this specimen (Supplementary Video S1), a correction for lateral displacement of Henle fibers and post-receptoral cells (Schein, 1988; Drasdo et al., 2007) was unneeded. However, the non-availability of tissue outside the foveal center of our specimen prevented us from defining an eccentricity relationship and comparing it directly with the previous literature. These studies collectively suggest a unity ratio of cones and all Müller glia outside the foveal center in presumably normal human retinas. In macaque monkeys, monoclonal antibodies specific to Müller glia labeled 30,000–37,000 cells/mm² in the parafoveal region compared to 6,000 cells/mm² in the far periphery (Distler and Dreher, 1996). Using serial section electron microscopy and manual reconstruction, Ahmad et al. (2003) found at 575 μm from the foveal center (corresponding to 1°) 26,661 cells/mm² of Müller glia outer trunks and cone pedicles 26,512 cells/mm² for a 1:1 ratio. Using antibodies against glutamine synthetase and differential interference contrast optics to visualize unstained cells, Masri et al. (2021) found in four human specimens that the average peak density of Müller glia at 0.8 mm from the foveal center was 24,000 cells/mm² with high variability. These authors could not assess the foveal floor. They did find that the ratio of Müller glia to cone inner segments in the same retinas within 2 mm eccentricity varied between 1 and 2 glia per cone. Bringmann et al. (2018) counted nuclei in histological cross-sections of an adult human retina and reported 0.73 ± 0.27 cone nuclei per Müller cell nuclei in the foveal center. Of note, foveal Müller cells were described later than the neurons (Polyak, 1941), and their presence, abundance, and distinctive features were debated (Yamada, 1969; Miller and Bernard, 1983; Krebs and Krebs, 1989; Hoang et al., 2002). Some investigators (Delaunay et al., 2020) considered a population of GFAP+ cells in the foveal floor to be astrocytes that may have lost this marker in development (Provis et al., 2000) or perhaps were not detected in specimens lacking the vitreous. The high abundance of foveal Müller glia is novel and should be investigated in other eyes of known developmental history to determine its functional significance.

Our data on layer-specific glial occupancy (Figure 4) are relevant to layer-specific concentrations of macular pigments (lutein and zeaxanthin). These xanthophyll carotenoids are lipophilic nutrients of dietary origin that intercalate into plasma membranes, enhance cone-mediated vision, confer antioxidant protection, and account for the characteristic yellow spot (“macula lutea”) of the human retina (Bone et al., 1988; Grudzinski et al., 2017; Mohn et al., 2017; Kar et al., 2020). Müller glia are major xanthophyll reservoirs as learned through research on glio-degenerative disease (Helb et al., 2008; Powner et al., 2010, 2013; Theelen et al., 2014; Pang et al., 2016; Obana et al., 2017) and persistence of xanthophyll signal after the death of foveal cones (Li et al., 2018; Schultz et al., 2022). An overall high concentration in the foveal center that extends into plexiform and nerve fiber layers (Snodderly et al., 1984; Trieschmann et al., 2008; Li et al., 2020) is parsimoniously accounted for Müller glia, without excluding other cells in those locations. Xanthophylls can be detected chromatographically in prenatal retinal tissue, becoming sufficiently concentrated to permit visibility in the fundus at 6 months of age or later (Bone et al., 1988), presumably as the fovea, and Müller glia within it, mature.

In conclusion, in an adult retina born pre-term, we used volume electron microscopy to discover and characterize a novel class of Müller glia with abundant cytoplasm in the inner retina and lacking an extension toward the ELM-like classically defined cells. Many features of these cells support a hypothesis that Müller glia are impacted by the lateral migration of inner retinal neurons in development, a phenomenon well-known from histology and clinical imaging of children and adults born pre-term (Bowl et al., 2016, 2018a; Sjöstrand et al., 2017). There are insufficient data in this case and in the literature to assess whether a class of foveal Müller glia may lack a connection to the ELM in a normal human retina. More human eyes of different developmental histories are needed to make that determination. Because much cellular-level anatomy is now accessible through OCT-based clinical imaging (Kadomoto et al., 2021; Ramtohul et al., 2022; Gupta et al., 2023), we expect that in vivo studies, including longitudinal follow-up (Seely et al., 2022), will allow testing of what aspects of inner Müller glia are unique to pre-term birth.

The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding author/s.

The studies involving humans were approved by Institutional Review Boards at University of Washington and University of Alabama at Birmingham. The studies were conducted in accordance with the local legislation and institutional requirements. The human samples used in this study were acquired from by an IRB-approved protocol, and in collaboration with organ and eye procurement organizations; eye tissue was recovered from a brain-dead donor at the time of organ recovery for transplant. Written informed consent for participation was not required from the participants or the participants’ legal guardians/next of kin in accordance with the national legislation and institutional requirements.

DK: Visualization, Validation, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization, Writing – review & editing, Writing – original draft. RaS: Investigation, Writing – original draft. YK: Visualization, Resources, Funding acquisition, Writing – review & editing. OP: Visualization, Resources, Investigation, Data curation, Writing – original draft. RiS: Visualization, Methodology, Writing – original draft. DC: Writing – review & editing, Visualization, Investigation, Data curation, Writing – original draft. KS: Writing – review & editing, Software, Writing – original draft. AP: Visualization, Funding acquisition, Writing – review & editing. DD: Visualization, Supervision, Resources, Project administration, Methodology, Investigation, Funding acquisition, Data curation, Conceptualization, Writing – review & editing, Writing – original draft. CC: Visualization, Supervision, Resources, Project administration, Funding acquisition, Data curation, Conceptualization, Writing – review & editing, Writing – original draft.