The data for this study were extracted from the Science Citation Index Expanded (SCI-EXPANDED) within the Clarivate Analytics Web of Science Core Collection (WoSCC). The WoSCC is a highly esteemed global database that provides comprehensive information for use with bibliometric software and is a popular choice for bibliometric analysis. TS = (axon recovery) or TS = (axon repair) or TS = (axon regrowth) or TS = (axon regeneration) was used to search for relevant articles. A systematic search was performed within the WoSCC for literature published between January 1900 and December 2023. Document retrieval and recording were concluded on December 31, 2023, to prevent possible bias resulting from subsequent database updates. Only complete original articles and reviews were considered for bibliometric analysis, regardless of the language. For each article, two researchers separately screened the title, abstract, and document type. If necessary, the researchers perused the full article for a more comprehensive evaluation of whether to include it in the analysis. The records of the 100 most influential publications were obtained from WoSCC in the “Full Record and Cited References” format and saved as.txt files (Figure 1). The impact factor of each journal was derived from the most recent (2022) Journal Citation Reports.

Descriptive statistical analysis and diagram generation were performed using Microsoft Excel 2019. An online analytic platform¹ was used to identify collaboration networks among authors and institutions. VOSviewer, a robust bibliometric tool developed by Van Eck and Waltman (van Eck and Waltman, 2010), was employed for the construction of networks involving keywords, authors, and more. Scimago Graphica software (Hassan-Montero et al., 2022) was used to create a world map illustrating the number of countries cited. Additionally, the downloaded data were integrated into CiteSpace (Chen, 2004) to construct author networks, identify keywords with the highest burst strength, and illustrate the relationship between time and keyword bursts. CiteSpace was also used to generate an institutional cooperation network, an author co-authorship network, and a co-occurrence network. In the network map rendered by CiteSpace, different elements, such as institutions, authors, and keywords, are represented by various nodes. The thickness of the lines connecting the nodes signifies the strength of the co-authorship and co-occurrence links. In the co-occurrence analysis, the correlation of items was determined based on the number of documents in which they co-occur, with different clusters denoted by different colors. The “bibliometrix” R package was employed for bibliometric analysis of the relevant literature. Bibliometrix allows gaining rapid insights into the seminal works, prominent figures, and future trends in the field and aids in organizing the pertinent literature and presenting the results in a visual format. One notable feature of bibliometrix is its ability to generate Sankey diagrams, also known as Sankey energy balance diagrams. These diagrams depict data flows, with the width of the branches proportional to the magnitude of the flow. They function as effective tools for visually analyzing data.

All the 100 most cited studies were published between 2003 and 2019 (Figure 2). Table 1 provides detailed information on the 100 most influential publications. The top 100 articles garnered between 215 and 1,604 citations, with a median of 326 and an average of 395.48 citations per article. Focusing on the top three studies, the most cited article (Kigerl et al., 2009) (“Identification of Two Distinct Macrophage Subsets with Divergent Effects Causing either Neurotoxicity or Regeneration in the Injured Mouse Spinal Cord”) was published in the Journal of Neuroscience in 2009 and was cited 1,604 times. The second most cited paper (Park et al., 2008) (“Promoting Axon Regeneration in the Adult CNS by Modulation of the PTEN/mTOR Pathway”) was published in Science in 2008 and received 1,187 citations. The third most cited paper (Anderson et al., 2016), titled “Astrocyte scar formation aids central nervous system axon regeneration,” was published in 2016 in Nature and received 1,139 citations. Earlier publications had an advantage in terms of the total citation count. However, when ranked according to average citations per year, some later publications were found to have a greater impact. For instance, although the above-mentioned article (Anderson et al., 2016) “Astrocyte scar formation aids central nervous system axon regeneration,” published in 2016, ranked third for total citations, it ranked first for average number of citations per year (142.38).

As shown in Figures 3A,B, a total of 16 countries/regions contributed to the 100 most cited papers, with 6 of these countries/regions contributing more than 3 articles. Scimago Graphica software was employed to generate a world map highlighting the countries associated with the top 100 cited publications (Figure 3C). The field of axon regeneration was dominated by the USA, with 72 out of the 100 most cited papers, 27,072 citations, and an average citation count of 423, which was the highest among all the countries (Figures 3, 4). The United Kingdom was ranked second, contributing 11 papers and receiving 3,145 citations, with an average citation count of 393.1. Germany was the third-ranked country, contributing 8 articles and receiving 2,499 citations, with an average citation count of 357.

In total, 155 institutions contributed to the top 100 cited articles relating to axon regeneration. Figure 5A shows the publication count for the most pertinent organizations. Among these, Harvard University emerged as the leading institution, generating the highest number of top-cited articles (n = 19). CiteSpace software was used to visualize the interconnections between institutions (Figure 5B). Several prominent research institutions, such as Harvard University, Boston Children’s Hospital, the University of California, and the University of Miami, among others, can be identified.

The most relevant authors were identified by employing the “bibliometrix” R package. As demonstrated in Figure 6A, He, Zhigang had the highest number of publications (n = 11). Figure 6B depicts the timeline of publications for the 10 most influential authors. Additionally, our analysis indicated that author collaboration was relatively frequent, and partnerships between authors were consistent (Figure 6C). The relationships among countries, institutions, and authors are shown in Figure 6D, with authors from the USA having a substantial impact in this respect.

Table 2 displays the journals in which the 100 most cited articles were published. The Journal of Neuroscience published the highest number of articles (n = 21), with 8,510 citations. This was followed by Neuron, with 12 publications and 3,824 citations; Experimental Neurology, which published 10 of the 100 papers and had 3,623 citations; and Science, which contributed 9 articles and received 4,345 citations, with an average citation count of 482.78. Furthermore, 9 journals published more than 3 articles—Nature Neuroscience, Nature Medicine, Biomaterials, Nature, and Proceedings of the National Academy of Sciences of the United States of America, which accumulated 3,465, 2,252, 2,201, 2,546, and 1,825 citations, respectively.

Keywords play a crucial role in delineating the focus of an article, thereby providing researchers with a comprehensive understanding of the published subject matter. The co-occurrence of two keywords within a given paper signifies that there is an inherent relationship between them, with the frequency of their appearance denoting the intensity of this association. By undertaking keyword co-occurrence and emergent item analysis, we can discern hot topics within a specific field across varying periods, and aggregate the keywords provided by authors in the dataset. Keyword clustering characterizes the inherent knowledge structure within a specific research field and categorizes its domain. In this study, cluster analysis revealed that keywords within the field of axon regeneration could be partitioned into 16 categories (Figure 7) —“chondroitin sulfate proteoglycan,” “alternative activation,” “neurite growth,” “exosome,” “spinal cord,” “axonal regeneration,” “Schwann cells,” “axon growth,” “axonal protein synthesis,” “neurite remodeling,” “electrical stimulation,” “injury,” “adult rats,” “nerve regeneration,” “therapeutic factors,” and “remyelination.” Figure 8 illustrates the trend in keyword changes over time. The dynamic relationship between keywords and time is portrayed in Figure 8A. The size of each block relates to the popularity of the respective keyword, the larger the block the higher the frequency of keyword occurrence. Moreover, keywords exhibiting a recent increasing trend may represent popular future research topics.

We conducted a burst analysis of the keywords associated with axon regeneration and identified several prominent keywords including growth myelin-associated glycoprotein spinal cord injury neurite outgrowth peripheral nerve regeneration outgrowth neurite growth inhibitors and cells. Keywords with the highest burst strength may represent the forefront of research in a given field. Keywords with earlier bursts are likely to have been the focus of research in the early years whereas those with more recent bursts represent topics exhibiting a sudden surge in attention

As depicted in Figure 8B, the five keywords with the greatest burst strengths were “growth,” “myelin-associated glycoprotein,” “spinal cord injury,” “neurite outgrowth,” and “peripheral nerve regeneration,” showing burst strengths of 4.84, 4.41, 2.56, 2.45, and 2.26, respectively. The earliest bursting keywords were “myelin-associated glycoprotein,” “spinal cord injury,” “outgrowth,” “neurite growth inhibitors,” “axon growth,” and “adult rat,” while the most recent bursting keywords comprised “local delivery.”

Chondroitin sulfate proteoglycans (CSPGs) comprise a class of large molecules ubiquitously present in the CNS that play pivotal roles in neural development, synaptic plasticity, and neural injury repair. However, CSPGs tend to accumulate in substantial quantities at neural injury sites, thereby establishing an inhibitory milieu that impedes axon regeneration (Silver and Miller, 2004), and surmounting this inhibitory activity of CSPGs to achieve axon regeneration represents a significant challenge. CSPGs hinder axonal growth through their interaction with a range of receptors, including Nogo receptor (NgR) and protein tyrosine phosphatase sigma (PTPσ). Following their activation on neurons, these receptors trigger a cascade of signal transduction pathways, culminating in cytoskeletal reorganization and the consequent inhibition of axonal growth (Shen et al., 2009). Nevertheless, recent research has uncovered multiple strategies to counteract the inhibitory impact of CSPGs and foster axon regeneration. These include employing the enzyme chondroitinase ABC (ChABC) to degrade CSPGs, which nullifies their inhibitory effect on axonal growth (Bradbury et al., 2002), and obstructing the interactions between CSPGs and their receptors or intervening in the signal transduction pathways activated by CSPGs (Monnier et al., 2003). In summary, the accumulation of CSPGs post-neural injury engenders an environment that curbs axon regeneration. Nevertheless, this inhibitory effect of CSPGs can be surmounted, thereby facilitating axon regeneration, through their degradation, by obstructing their interaction with receptors, or by intervening in their signaling pathways. These findings provide novel therapeutic strategies for the repair of neurons after injury.

The term “alternative activation” generally refers to the activation state of immune cells, specifically macrophages. Following neural injury, the behavior of macrophages and other immune cells substantially impacts the potential for axon regeneration. Macrophages are multifunctional immune cells that can adopt diverse activation states contingent on environmental cues. These activation states comprise two primary categories, namely, M1 and M2. M1 macrophages primarily contribute to inflammation and immune responses, whereas M2 (also termed “alternatively activated”) macrophages participate in repair and regenerative processes (Kigerl et al., 2009). Following neural injury, M1 macrophages aggregate at the injury site and secrete abundant inflammatory mediators and cytokines, which can exacerbate neuronal damage and hinder axon regeneration. Conversely, factors synthesized by M2 macrophages can foster neuronal survival and axonal expansion, thereby facilitating neuronal repair. These observations suggest that axon regeneration can be promoted by modulating the activation state of macrophages. For instance, inducing the transition of macrophages toward the M2 phenotype might assist in enhancing repair following neural injury (Gensel and Zhang, 2015). Although the viability of this strategy has been corroborated in some experimental models, its implementation in clinical settings requires further investigation.

Schwann cells are the primary glial cells in the PNS and play a crucial role in axon regeneration. Their key functions in this process include the following: (1) The provision of growth factors. Schwann cells can produce and release a variety of growth factors, such as nerve growth factor (NGF), brain-derived neurotrophic factor (BDNF), and neurotrophin-3 (NT-3), which can promote axonal growth and regeneration (Kingham et al., 2007). (2) The formation of regeneration tracks: Following peripheral nerve injury, Schwann cells can proliferate and align into bundles, forming so-called “regeneration tracks” or “Bands of Büngner” that provide physical support and guidance cues for regenerating axons (Fouad et al., 2005). (3) Myelination: Schwann cells can form myelin, a lipid-rich membrane that wraps around axons, providing electrical insulation and accelerating nerve impulse conduction. (4) Immune regulation: Schwann cells can modulate immune responses by secreting a variety of cytokines that can influence the inflammatory response and neuron repair following nerve injury. These functions of Schwann cells underscore their pivotal role in the repair process following nervous system insult.

Exosomes are diminutive vesicles produced by living cells that are subsequently released into the extracellular environment. They can carry molecules, including proteins, lipids, and RNAs, which are transferred to recipient cells following their internalization, a process that represents a form of intercellular communication. Evidence increasingly suggests that exosomes play a pivotal role in neural repair and axon regeneration through the delivery of growth-promoting molecules. For example, it has been shown that exosomes produced by Schwann cells can carry and deliver growth factors, enzymes, and other molecules that can promote axon regeneration and provide neuroprotection (Lopez-Verrilli et al., 2013). Although these preliminary research findings hold promise for achieving axon repair, our understanding of the specific mechanisms by which exosomes influence axon regeneration remains inadequate. Further comprehensive research is required to address this limitation and determine the potential of exosomes in neural repair.

Protein synthesis within axons is vital for axonal growth and regeneration. It was originally thought that most protein synthesis occurred in the cell body, after which the synthesized proteins would be transported to the axon via axoplasmic transport; however, it is now acknowledged that there exists a mechanism for local protein synthesis within the axon itself, which allows neurons to rapidly respond to environmental stimuli. The possible roles of this mechanism include the following: (1) Supporting axonal growth. Within the growth cone of axons, local protein synthesis can quickly provide the proteins required for axonal growth. For example, the local translation of β-actin mRNA was found to be crucial for axonal growth and guidance (Jung et al., 2012). (2) Responding to environmental signals. Axons can rapidly respond to environmental signals through local protein synthesis. For example, some growth factors and guidance molecules can stimulate the translation of specific mRNAs in axons, thereby affecting axonal growth and guidance. (3) Regeneration following nerve injury. Following nerve injury, local protein synthesis may be crucial for axonal regeneration. For example, local protein synthesis in axons increases after nerve injury, which may facilitate axonal regeneration (Twiss and Fainzilber, 2009).

Despite our limited understanding of local protein synthesis in axons, research within this domain presents novel possibilities for developing strategies for neural regeneration therapy.

Electrical stimulation has been demonstrated to foster axon regeneration and has been utilized in the treatment of certain nerve injuries. The potential underlying mechanisms include the following: (1) Increasing nerve growth factor expression. Electrical stimulation can enhance the expression of nerve growth factors (such as NGF and BDNF, among others), which can promote axonal growth and regeneration. (2) Modifying the electrophysiological activity of nerve cells. Electrical stimulation can alter the electrophysiological activity of nerve cells by, for example, modifying the activity of ion channels and increasing the excitability of nerve cells, thereby promoting axonal growth and regeneration. (3) Manipulating intracellular signaling pathways. Electrical stimulation can influence intracellular signaling pathways, such as increasing the level of cAMP and activating the mTOR signaling pathway, both of which play a key role in axonal growth and regeneration. (4) Enhancing angiogenesis and ameliorating the local microenvironment. Electrical stimulation can also promote angiogenesis and improve the microenvironment of the nerve injury area, providing favorable conditions for axonal growth and regeneration.

Although the facilitating effect of electrical stimulation on axon regeneration has been widely confirmed, further research is needed to elucidate in detail the specific mechanisms of action. Furthermore, the need to optimize electrical stimulation parameters, such as the intensity, frequency, and duration of stimulation, and thereby achieve the most beneficial therapeutic effect, is also a significant issue in current research.

Remyelination and axon regeneration constitute two closely related pivotal steps in the process of neural repair. An axon is a protrusion of a neuron through which information is transmitted to adjacent cells. The axons of many types of neurons are enveloped in a lipid structure called myelin, which serves both to accelerate the propagation of nerve impulses and protect the axon. In the CNS, myelin is produced by oligodendrocytes, while in the PNS, it is produced by Schwann cells. Following nerve injury, axons might be severed, and myelin may be lost or damaged. Despite their importance for nerve repair, axon regeneration and remyelination do not invariably proceed in synchrony. For example, axons may grow without sufficient myelin, or myelin may form without axons. Remyelination and axon regeneration represent intertwined processes, with each influencing the other. For instance, during axon regeneration, neurons release factors, such as NGF, that stimulate the proliferation and differentiation of Schwann cells or oligodendrocytes, thereby promoting remyelination (Höke et al., 2006). Regarding the effect of remyelination on axon regeneration, in addition to forming myelin, myelinating cells can also secrete factors, such as neurotrophic factors, that promote axon growth. Moreover, myelin-forming cells can also provide a structural framework for guiding axon growth through physical contact with growing axons (Höke et al., 2006; Kingham et al., 2007). In the treatment of nerve injury, strategies to promote both remyelination and axon regeneration are often considered simultaneously.

The development of local delivery systems has emerged as a particularly promising strategy for promoting axon regeneration, allowing for the sustained and targeted release of therapeutic agents directly at the site of injury. These systems, including hydrogels, nanoparticles, and biomaterials, have been engineered to provide a supportive environment for axonal growth as well as to overcome the inhibitory effects of the glial scar. Lu et al. (2003) and Xin et al. (2013) demonstrated the neurotrophic factor-secreting potential of neural stem cells and mesenchymal stem cells, and that these cells can be leveraged for local delivery, fostering a conducive environment for axonal regeneration. After nerve injury, especially after SCI, the local microenvironment, such as inflammation and scar tissue formation at the injury site, is one of the important factors that hinder its repair. If the local delivery of chemical drugs or bioactive molecules can effectively improve the inhibitory local microenvironment, it will be very beneficial to the treatment of SCI. Thus, with the increasing development of “local delivery” technique, this technique will be widely used in SCI repair, by reducing the dose of drugs, to achieve better efficacy, while reducing its toxic side effects.

A collaborative, multidisciplinary approach that integrates molecular biology, bioengineering, and clinical neuroscience will be crucial for developing effective treatments. Standardized protocols for preclinical studies and the establishment of sensitive biomarkers will facilitate the translation of research findings into clinical practice, ultimately improving patient outcomes following CNS insult. The continued exploration of the intricate balance between regeneration-promoting and regeneration-inhibiting factors, along with the optimization of local delivery systems, will be instrumental in realizing the full potential of axon regeneration therapies. However, although our animal experiments have yielded encouraging results, compare to human, animals are more compensatory after SCI, and walking on four legs can easily lead to misjudgments in behavioral assessments. Therefore, more primate experiments should be conducted in the future, and more sophisticated behavioral assessment methods should be developed.

This study had several limitations. Despite being the most used database for literature searches, Web of Science does not encompass all publications. To ensure the accuracy of our analysis, we employed the topic search method instead of the subject keywords search method. Although our search results demonstrate some precision, they could have been more comprehensive.