Beyond homogenates: New tool available for estimating glycogen’s numerical subcellular distribution

The glycogen content of a skeletal muscle is typically determined by a chemical method after homogenizing a muscle specimen of 20–50 mg. This measure, therefore, represents an average of hundreds of muscle fibers and their subcellular compartments. The glycogen content is the product of the numerical density and the size of the glycogen granules, and given the uneven distribution of granules within the muscle fibers, local compartments form where the numerical density can be 20-fold higher than in other compartments (Marchand et al., 2002). Those distinct local compartments of glycogen are used by different energy-requiring processes during muscle work (Nielsen et al., 2022) and show differential numerical regulation during recovery achieving above-normal glycogen content (Jensen et al., 2021), suggesting that the glycogen content of a homogenate may overlook important spatial information and regulation at the granule level. However, the role and regulation of this subcellular distribution of glycogen are far from understood. One reason why only a few studies have been undertaken may reside in the time-consuming process of image analyses. A precise estimate may be obtained after around 1–2 h of work per muscle fiber. In a typical study with 50–100 muscle biopsies, including 300–600 muscle fibers, this accumulates to a substantial amount of resources spent on image analyses. The study by Ríos et al. (2024) in this issue shows how artificial intelligence (AI) can be used to estimate the subcellular distribution of glycogen granules using images from transmission electron microscopy, significantly reducing the image analysis load.

The imaging of glycogen granules, and hence, the subcellular distribution, dates back to at least 1918. Using light microscopy and Best’s carmine to stain glycogen, Takeuchi described the subcellular distribution of glycogen in different muscles from humans several hours after death (Takeuchi, 1918). The introduction of the electron microscope to the biomedical science (Palade, 1952) and later a refined preparation protocol for optimal glycogen staining (De Bruijn, 1973) ensured that a detailed subcellular distribution of glycogen could be envisioned. Different research groups, using different preparation protocols, qualitatively described how glycogen granules disappeared or became smaller in specific subcellular locations after various types of exercise (e.g., Oberholzer et al., 1976; Sjöström et al., 1982). Later, a quantitative method was conducted by Marchand et al. (2002), describing granule size and numerical density in three defined locations: intermyofibrillar, intramyofibrillar, and subsarcolemmal. In human skeletal muscle and expressed as a volumetric density, the intermyofibrillar location contains ∼80% of the whole cellular amount, and the intramyofibrillar and subsarcolemmal locations each contain 10% (Marchand et al., 2002). The method by Marchand et al. (2002) was semi-automated, with manual outlining of the locations combined with auto-detection of glycogen granules. The volumetric density of glycogen in distinct subcellular compartments can also be estimated by a stereological point-counting technique combined with granule diameter measurements (e.g., Jensen et al., 2021). Although the point-counting technique is a manual technique, it is optimized to be time efficient (Gundersen et al., 1988).

Most of the studies define only the abovementioned three subcellular locations, although some with notes on differences between I- and A-bands of the sarcomeres and between peri–sarcoplasmic reticulum (SR) and peri-mitochondria (e.g., Fridén et al., 1985). Thus, when prolonged exercise mediates a preferential depletion of intramyofibrillar glycogen (Schytz et al., 2024), it is unknown if it differs between I- and A-bands. And when short-term high-intensity exercise mediates a high depletion rate of intermyofibrillar glycogen (Schytz et al., 2024), it is unknown if it differs between peri-SR and peri-mitochondria sub-compartments. The AI model provided by Ríos et al. (2024) can be trained to make these discernments and others in a time-efficient process.

A time-efficient and objective tool to quantify the subcellular distribution of glycogen can provide new insights into several other research questions. Two examples relate to the interplay between a muscle’s glycogen content and its uptake of glucose. Despite a metabolic mix of blood-borne and intramuscular substrate stores (plasma glucose and fatty acids versus intramuscular glycogen and triglycerides) during muscle work, no studies have shown considerable effect sizes of carbohydrate intake during exercise on muscle glycogen utilization, i.e., a glycogen-sparing effect (e.g., Coyle et al., 1986; Fell et al., 2021). It could be speculated that such an effect resides within a restricted subcellular compartment and has therefore been overlooked in homogenates. Similarly, the rate of glucose uptake after exercise appears to be independent from glycogen content or utilization during the preceding exercise (Hingst et al., 2022). However, above-normal levels of glycogen after exercise are not uniformly distributed (Jensen et al., 2021), suggesting that a potential glycogen dependency of glucose uptake rate could be overlooked in a homogenate.

The model by Ríos et al. (2024) is more detailed than previous quantitative approaches, as it also quantifies the presence of glycogen close to the SR and the distribution within sarcomeres. A glycogen–SR association may form a functional compartment, where energy from glycogen breakdown facilitates active transport by calcium pumping along the SR membrane (Cuenda et al., 1993). It is also suggested that a calcium-leaky SR disturbs glycogen metabolism in an SR-membrane sub-compartment (Tammineni et al., 2020). Thus, the major deposition of glycogen in the intermyofibrillar compartment may comprise several sub-compartments, which could be considered in an AI model as suggested by Ríos et al. (2024).

The distribution of glycogen particles within the sarcomeres (intramyofibrillar glycogen), defined as either I- or A-band particles, was three- to fivefold higher in the I-band than in the A-band (Ríos et al., 2024), which aligns with the non-quantitative previous studies (Sjöström et al., 1982; Fridén et al., 1985). In a fine-tuned model, it would also be interesting to develop a measure of strings of glycogen granules versus isolated (non-strings) granules. Wanson and Drochmans (1968) noted that intramyofibrillar glycogen granules were often aligned as “beads in a string,” the functional relevance and regulation of which remain to be investigated. Examples of high and low numerical density of granules in different compartments are shown in Fig. 1.

In summary, information about the uneven distribution of glycogen has been available at least since 1918 and was more extensively documented from the 1960s to the 1980s. However, only a few studies have taken this information into account. It could be argued that the hundreds to thousands of research papers, where glycogen content is determined chemically from mixed muscle homogenates, more accurately measure intermyofibrillar glycogen (the major pool) while the influence of the other minor pools remains an unknown confounding factor.

Based on a discrepancy between the measured numerical density of glycogen granules and the expected glycogen content according to the literature, Ríos et al. (2024) suggest that glycogen may exist in a non-granular form. The current understanding is that glycogen exists in three forms: β-granules, α-granules formed by several β-granules, and γ-particles forming parts of the β-granules (Drochmans 1962). It remains to be investigated whether γ-particles (described as rod-like structures 3 nm wide and up to 20 nm long) can be found beyond β-granules and thus contribute to the discrepancy. Ríos et al. (2024) propose the intriguing idea that the non-granular form of glycogen (not necessarily as γ-particles) is present in the SR and T-tubular membranes, where glycogenolytic and glycolytic enzymes are known to have binding sites (Wanson and Drochmans, 1972). This could form the basis for a very time-efficient energy delivery for membrane transport processes. It is important to note that the numerical density of granules has been reported as higher in other studies (Marchand et al., 2002; Jensen et al., 2021), and that the numerical density may be subject to underestimation due to undetected small granules and overlapping of the 20–25-nm-sized granules in the thicker 60-nm section projected as a 2-D image using transmission electron microscopy. Ríos et al. (2024) addressed this issue by simulating the overlap of granules at given volume densities and granule sizes. Together, a robust measure of granule numerical density and volumetric density, as provided by Ríos et al. (2024), combined with chemical determination of glycogen content, will form the basis for understanding the various forms of glycogen under different conditions.

A better understanding of glycogen in skeletal muscle requires investigations of its subcellular distribution and regulation at the granule level. This can be accomplished using a transmission electron microscopy protocol optimized for glycogen visualization, followed by objective and time-efficient quantitative image analysis using the AI model provided by Ríos et al. (2024).

Olaf S. Andersen served as editor.

The author thanks Daiki Watanabe (Osaka University of Health and Sport Sciences, Japan) for kindly providing the paper by Takeuchi (1918).