The search was performed on the Science Citation Index-Expanded (SCI-E) of the Web of Science Core Collection online database (Thomson Reuters) hosted by Clarivate Analytics. All electronic searches were conducted on a single day, July 3, 2019, to avoid changes in citation rate as much as possible. The string “TOPIC:(glucocerebrosid^* OR glucosylceramid^*) AND Language:(English)” was used to retrieve any documents published from 1956 to 2019 mentioning for example “glucocerebrosides,” “glucosylceramides,” “glucosylceramide synthase,” or “glucocerebrosidase” in their titles, abstracts, author keywords, or KeyWord Plus (keywords unique to Web of Science that consist of words and phrases harvested from the titles of the cited articles). Retracted publications were excluded. Bibliometric data were extracted including titles, author information, abstracts, keywords, source, publication year, Web of Science category of the publication and cited references. The data were downloaded in plain text and imported in CiteSpace 5.5.R2 (64-bit) (Synnestvedt et al., 2005) for analysis.

Every record in Web of Science Core Collection is assigned to at least one subject category of its source publication. CiteSpace was used to generate and visualize networks of the journal-based Web of Science Subject Categories assigned to the publications in our dataset. Burst detection was applied to detect subject categories or keywords that had a surge of their appearance/citation for a specific period of time.

A document co-citation network represents a network of references that have been co-cited by a set of publications. Briefly, if two articles are both cited as references in another article, then these two papers have a co-citation relationship (Small, 1973). Citespace was used to generate document co-citation networks derived from the 50 most cited articles published in every single year y as Document co-citation analysis (DCA[y]). The time series of these individual DCAs were then integrated for all the period to produce a synthesized network. The synthetized network was divided into a number of clusters of co-cited references so that tightly coupled references were within the same clusters and loosely connected references were in different clusters. Log-likelihood ratio test method was used to label these clusters with terms extracted from the titles of the most representative articles for each cluster. Betweenness centrality of publications were calculated in an attempt to identify those that bridge two or more clusters. Burst detection was applied to publications to detect those cited at an increasingly faster rate during a period of time. Sigma, introduced by Chen and coworkers in 2009 and defined as (centrality+1)^burstness, was used as a measure of scientific novelty such that the brokerage mechanism plays more prominent role than the rate of recognition by peers (Chen et al., 2009).

A total of 5,324 publications met the search criteria. These include 3,999 research articles (75.11%), 563 reviews (10.57%), and 511 meeting abstracts (9.59%). According to the analysis of the Web of Science database, the global h-index was 153 with an average citation per item of 31.19.

Each publication indexed in the Web of Science is associated to one or more journal-based subject categories. In our dataset, a total of 68 unique subject categories were found. The most represented categories were: (1) Biochemistry & Molecular Biology (1,534 publications); (2) Genetics & Heredity (742 publications); (3) Medicine, Research & Experimental (519 publications); (4) Clinical neurology (436 publications); (5) Cell Biology (410 publications); (6) Hematology (380 publications); (7) Neuroscience (362 publications); (8) Pharmacology and Pharmacy (358 publications); (9) Biophysics (317 publications); (10) Endocrinology and metabolism (315 publications). Figure 2 shows a network of these Web of Science categories largely assigned to the publications in our dataset. Burstness analysis revealed that the categories Clinical Neurology (25.45) and Neuroscience (37.46) showed very strong bursts starting in 2012 and 2016, respectively (Table 1).

Publications in the dataset are assigned keywords. The 300 keywords that occurred the most in the 5324 publications or our dataset were extracted and were analyzed for their burstness. The keywords “glucocerebrosidase mutation” (2012, 2019), “Parkinson's disease” (2013, 2019), and “alpha-synuclein” (2014, 2019) were among those with the highest burstness and their abrupt increase in citation all occur in the last decade (Table 2).

Figure 3 shows the synthesized co-citation network for glucocerebrosides research publications. In this representation, nodes represent cited references. The radius of a node increases with its frequency of citations within the network (log transformed). Edges represent instances of co-citation and their thicknesses are proportional to the number of times that two documents are jointly cited. Nodes with a betweenness centrality >0.1 (e.g., nodes linked with over 10% of the nodes in the whole network) are represented with a purple ring. We identified most cited publications within the network (showing the largest radii) as focusing on the causal association between glucocerebrosidase mutations and Parkinson's disease (Neumann et al., 2009; Sidransky et al., 2009), glucocerebrosidase activity deficiency in Parkinson's disease (Gegg et al., 2012), the interaction between glucocerebrosidase and alpha-synuclein (e.g., the protein that accumulates in neurons in Parkinson's disease) (Mazzulli et al., 2011), enzyme replacement therapy in Gaucher disease (Barton et al., 1991), Gaucher disease physiopathology (Beutler and Grabowski, 2001), and glucocerebrosidase mutations in Gaucher disease (Hruska et al., 2008). Publications with the highest betweenness centrality focus on the pathophysiology of Gaucher disease (Brady and Barranger, 1983; Wong et al., 2004), report the heterogeneity of mutations in the GBA gene of Gaucher disease patients (Latham et al., 1991) or characterize the links between glucocerebroside metabolism and epidermis (Holleran et al., 1994; Marsh et al., 1995). These latest publications likely bridge two or more underlying lines of research and may be important from different disciplinary perspectives.

The co-citation network for glucocerebrosides research was divided into clusters, so that references that are tightly connected are within the same clusters, but those loosely connected are in different clusters. The modularity Q score was higher than 0.5 (0.8423) indicating that the network was reasonably divided into loosely coupled clusters. Each cluster was labeled by words in the titles of citing publications of the cluster. The automatically chosen cluster label, size and average year of publication of the ten major clusters by their size are summarized in Table 3. The visualization of the network divided into distinct co-citation clusters is seen in Figure 4.

The cited publications of a cluster define its intellectual base. Table 4 shows the title and main conclusion of the most cited publication for each of the ten largest clusters. In addition, we provide in Supplementary Table 1 the list of the 10 most-cited members of these clusters as well as their structural, temporal, and saliency metrics such as citation count, betweenness centrality, citation burstness and sigma that can be considered as a measure of scientific novelty (Chen et al., 2009).

Taking into account the 10 most cited publications of each cluster, we propose a brief description of the intellectual content of the ten major clusters ranked by their size as follow:

Cluster #0 (Parkinson's disease): This cluster includes publications exploring the links between glucocerebrosidase, glucocerebrosides, and synucleinopathies such as Parkinson's disease and Dementia with Lewy bodies. These focus for example on GBA mutations, their impact on Parkinson's disease clinical phenotype, glucocerebrosidase enzymatic activity in Parkinson's disease and possible mechanistic links between glucocerebrosidase deficiency and alpha-synuclein accumulation.

Cluster #1 (Gaucher disease): Publications of this cluster outline the interest of glucocerebrosides as biomarkers for multidrug resistance in cancer cells and explore how the modulation of glucocerebroside metabolism (i.e., the modulation of glucocerebrosidase or glucosylceramide synthase enzymatic activities) can confer or reverse drug resistance in cancer cells.

Cluster #2 (Replacement therapy): This cluster relates to the therapeutic goals and strategies for Gaucher disease. It includes publications evaluating the clinical response to the treatments approved in the treatment of Gaucher disease and relying on enzyme replacement therapy (i.e., use of imiglucerase and previously aglucerase) or substrate reduction therapy (e.g., use of miglustat and eliglustat).

Cluster #3 (Human spleen): This cluster contains publications on the isolation, characterization and comparison of wild-type and mutant glucocerebrosidase from human tissues and cells (i.e., spleen, placenta, fibroblasts). Several publications also concern anionic phospholipids and the activator protein saposin C that promote glucocerebrosidase enzymatic function and contribute to Gaucher disease heterogeneity.

Cluster #4 (Molecular analysis): This cluster englobes publications on the GBA gene encoding the lysosomal glucocerebrosidase and its highly homologous pseudogene. Publications can focus on their structure, evolution, heterogeneity and to their mutation and their association with the diagnosis and clinical severity in Gaucher disease.

Cluster #5 (Kupffer cell): This cluster contains publications on the cloning and nucleotidic sequence of the GBA gene and on the polypeptide sequence of the lysosomal glucocerebrosidase. It contains articles describing the purification of human glucocerebrosidase or its synthesis from cDNA clones for enzyme replacement therapy in Gaucher disease.

Cluster #6 (Human glucocerebrosidase): This cluster is on the correction of glucocerebrosidase deficiency and its therapeutic response. Studies can rely on various strategies including intravenous infusion of glucocerebrosidase (e.g., replacement therapy), bone marrow transplantation in Humans as well as evaluation of gene therapy (i.e., retroviral-mediated gene transfer into bone marrow or hematopoietic cells) in experimental models or patients-derived cells.

Cluster #7 (Sphingolipid precursors): This cluster is on the different steps involved in the synthesis of glucocerebrosides and glycosphingolipids, their cellular localization as well as their regulatory function in cellular development and differentiation.

Cluster #8 (Pharmacological chaperone): Most publications of this cluster explore the ability of pharmacological chaperone to increase the enzymatic activity of the glucocerebrosidase possibly by decreasing its retention in the endoplasmic reticulum and restoring its trafficking toward the Golgi apparatus and the lysosome.

Cluster #9 (Glucosylceramide beta-glucosidase): This cluster is on the glucocerebrosidase and its mechanisms of activation. Many publications focus on the four saposins (A to D) that are activators of the glucocerebrosidase and especially on saposin C that mutation results in a Gaucher disease-like phenotype in Human.

Clusters were next represented in a timeline view so that nodes of a cluster share a horizontal line. Nodes with a betweenness centrality >0.1 are still represented with a purple ring. This representation helps to understand the temporal characteristics of each cluster and thus the changes in glucocerebrosides research trends over time. As seen in Figure 5, the development of cluster #3 (human spleen; that several publications contributed to the isolation, characterization and comparison of wild-type and mutant glucocerebrosidase from Human tissues and cells) occurred first. Its development was paralleled with that of the smaller cluster #15 (embryonic chicken brain) that largely investigated the composition of the glycolipid fraction of the embryonic chicken brain. Cluster #5 (kupffer cells) developed in the 1970s, reflecting the advances in the sequencing, purification and synthesis of the glucocerebrosidase. Subsequently, the development of the clusters #9 (glucosylceramide beta-glucosidase), #7 (sphingolipid precursor), and then #6 (human glucocerebrosidase) reveals the progress in our understanding on the synthesis and degradation of glucocerebrosides and glycosphingolipids and on the therapeutic response following the correction of glucocerebrosidase deficiency. Cluster #4 (molecular analysis) developed from the 1990, a period of intensive development of tools for genomic analysis 5 years after the Human Genome Project was conceived in 1985. Its publications largely contributed to our knowledge on the structure and variants of the GBA gene and its pseudogene and well as their association with phenotypes of Gaucher disease. Cluster #1 (Gaucher disease; that publications evaluate the interest of glucocerebrosides as biomarkers for multidrug resistance in cancer cells) developed in the 1990s. Cluster #2 (replacement therapy) and later cluster #8 (pharmacological chaperone) developed subsequently and helped to define the therapeutic goals and strategies for Gaucher disease. The period of development of these Clusters #2 #8 preceded the U.S. Food and Drug Administration approval of the first effective treatments for Gaucher disease (enzyme replacement therapies; alglucerase in 1991, imiglucerase in 1994) and of substrate reduction therapies (miglustat in 2003 and eliglustat in 2014). Finally, the timeline view evidences that cluster #0 on Parkinson's disease is the most recently formed cluster (Figure 5).

In an attempt to characterize the most recent trends in glucocerebrosides research, we searched for the ten publications with the strongest citation bursts starting after 2016. The number of citations of these articles had a surge during the last four years, suggesting a recent and rapid acceptance and dissemination. All of these top ten publications belong to Cluster #0 on Parkinson's disease, focusing for example on the relationship between GBA mutations or polymorphism and the onset, progression and clinical features of Parkinson's disease (Seto-Salvia et al., 2012; Beavan et al., 2015; Brockmann et al., 2015; Gan-Or et al., 2015; Cilia et al., 2016) and on the decrease of glucocerebrosidase activity in blood samples from patients with Parkinson's disease (Alcalay et al., 2015). The title, cluster, burstness score and main conclusion of each of these publications are summarized in Table 5.