While CNV and CL often occur concurrently, they are similar and different at the same time.

The existing research on angiogenesis is relatively thorough. Shi et al (13) described CNV using five models (suture induction, alkali burn, fungal infection and immunogen or tumor cell implantation) with three key steps in its formation: Sprouting, progressive and declination phases. Although the initial corneal edema and vasodilatation in these five models are consistent with the findings of Cogan (14), Shi et al (13) provided a more) detailed description of the pathogenesis of CNV. First, the basement membrane (BM) of the limbal capillary network is degraded by proteases released by ECs. Second, ECs migrate and invade the extracellular matrix (ECM), followed by proliferation. Finally, the lumen of new vessels forms, and the BM reshapes. An increasing number of studies suggests that CNV and CL are initiated by the budding of existing limbal vessels and lymphatics (3,15). Additionally, accumulating evidence indicates that corneal edema is not a prerequisite for the development of CNV (16,17). The processes involved in CNV vary depending on the trigger, but most steps are consistent. Therefore, the process of CNV can be summarized as follows:

i) When corneal injury or inflammatory irritation occurs, corneal edema, along with venules and capillary congestion, corneal edema can also stimulate cells, such as corneal epithelial cells, corneal ECs, vascular ECs and immune cells, to release angiogenic cytokines, including VEGF and basic fibroblast growth factor (bFGF). The upregulation of proangiogenic factors and the downregulation of angiogenesis inhibitory factors, such as soluble fms-like tyrosine kinase 1 (sFlt-1), metalloprotease 3, proteinase inhibitors, angiostatin, endostatin, platelet response protein, interleukin-1 receptor antagonist, pigment epithelial derived factor, vasoactive intestinal peptide and α-melanocyte-stimulating hormone, can occur under normal physiological conditions.

ii) Then, the combination of angiogenic factors and receptors activates the ECs of corneal perilimbal blood vessels. Corneal limbal ECs secrete matrix metalloproteinases to breakdown the ECM and vascular BM. Simultaneously, the combination of VEGF-A and VEGFR2 triggers vascular migration of ECs through the RAS-RAF-MAPK-ERK signaling cascade, the PLCγ-ERK 1/2 pathway, Ca²⁺ signaling, the PI3K-AKT pathway, small G protein, the SRC pathway, stress kinases and STATs. This enables ECs to migrate towards the cornea through chemotaxis (12,18).

iii) Some of these vascular ECs germinate and form vascular buds on the original blood vessels. VEGF-induced ECs exhibit a promigratory phenotype and undergo chemotaxis toward inflammatory sites. At those sites they form a structure that looks like a dendritic cell with a number of pseudopod-like structures to invade the corneal stroma. These cells are called tip cells and stalk cells. Tip cells are motile and invasive and can extend numerous filamentous pseudopodia that can respond to growth factors, the ECM, and other attractive or repulsive signals.

iv) A portion of ECs will undergo a transformation into a tip cell phenotype and initiate sprouting. Subsequently, the ECs (stalk cells) proliferate and contribute to the maintenance of the structural and functional integrity of neovascularization. The tips of the aforementioned pseudopod-like structures formed by ECs continue to secrete MMP and promote further growth of the blood vessels. However, new blood vessels are vulnerable due to the absence of pericyte support (19).

v) As neovascularization progresses, platelets release platelet-derived growth factor, which binds to receptors on pericytes, ultimately leading to their proliferation and migration. The chemotactic movement of pericytes to new blood vessels combines with them to stabilize new blood vessels (20).

vi) Other stimuli that can induce CNV include inflammation, hypoxia, limbal stem cell deficiency and denervation. Different stimuli stimulate different cells and cytokines to generate CNV. Inflammation triggers the release of cytokines such as VEGF, bFGF and BMP from corneal epithelial cells, corneal ECs, vascular ECs and immune cells, which promote CNV. Hypoxia leads to the expression of VEGF in the corneal epithelium, ECs and endothelium of the limbal blood vessels. Limbal stem cell deficiency recruits macrophages, a significant source of VEGF, and causes secondary increases in TGF-β, both of which promote CNV. Finally, the deprivation of nerves results in a reduction in corneal angiogenesis inhibitors (21). Although the initial processes of these conditions vary, most of their downstream processes overlap with those caused by inflammation-induced vasculo-genesis. The common process is demonstrated in Fig. 1.

The formation process of new corneal lymphatic vessels differs from that of CNV. The currently recognized mechanism is that they sprout from the original blood vessels and veins. The process is as follows:

i) The original venous vascular ECs express high levels of VEGF-3 and LYVE-1, and a subset of bone marrow-derived macrophages undergo trans-differentiation. Subsequently, both cell types undergo the process of converting into lymphatic ECs.

ii) The original venous vascular ECs and a subset of bone marrow-derived macrophages start to express the transcription factor SOX18, leading to the upregulation of prospero-related homeobox 1 (Prox-1). This represents a pivotal step in the conversion of these cells into lymphatic ECs (22). Vascular ECs and bone marrow-derived macrophages initially differentiate and sprout towards lymphatic ECs (22).

iii) These two proteins express neuropilin-2 (NP-2), and the synergistic effect of NP-2 and VEGFR-3 renders cells sensitive to VEGF-3, which promotes the extension of lymphatic endothelial tip cells and the continuation of sprouting (23).

iv) Afterwards, the lymphatic vessels bifurcate from the vein. Lymphatic ECs express podophyllotoxin, which promotes platelet aggregation, leading to the separation of lymphangiogenesis vessels from veins.

v) VEGFR-3 activates downstream pathways such as the Akt and p42/p44MAPK pathways, protecting lymphatic ECs from apoptosis and further promoting EC migration (24,25).

These are the classic steps of CL. These steps are also demonstrated in Fig. 2. However, it is important to note that not all lymphangiogenesis originates from the CNV. In the case of dry eye disease (DED), there is a unique condition in which corneal neo-lymphatic vessels are generated, but CNV is not. Multiple studies have revealed that this may be related to the specific cytokine mechanism of dry stress in DED (26), particularly the secretion of IL-17 by Th-17 cells, which induces an increase in VEGF-D (27), and the downstream pathways that lead to an increase in VEGF-3 induced by SP (28). In addition to budding from veins, some studies have suggested that lymphatic vessels may originate from mesenchymal lymphatic stem cells (29), but this theory has not been described for CL and has not been confirmed in mammals. Therefore, it will not be discussed in the present review.

The process of generating CNV and CL are described in detail. There are several similarities between CNV and CL in terms of stimulating factors and activating cells, but they also have their own unique features. The details are presented in Table I.

CNV is a tubular structure surrounded by ECs, with a wall composed of a single layer of ECs surrounded by smooth muscle cells and peripheral cells and an incomplete BM on the outermost layer. CNV arises from the sprouting of veins behind existing capillaries, and the process of its formation can be explained by the tips and stalk cells. In addition, CL involves a tubular structure surrounded by a single layer of lymphatic ECs. The ECs of the lymphatic vessels are surrounded by an incomplete BM, while the lumen is protected from reflux by lymphatic flaps. The formation of CNV occurs by sprouting from existing veins (30–32) Their morphologies are shown in Fig. 3.

Both structures share several similarities in terms of luminal morphology. They are both surrounded by a single layer of ECs, have discontinuous BMs, and sprout from pre-existing veins. However, there are also numerous differences:

i) CL involves the presence of a lymphatic valve that assists the entry and exit of APCs, but it does not involve the presence of blood vessels;

ii) The wall involved in CNV contains pericytes and smooth muscle cells in addition to ECs. On the other hand, the wall of the corneal lymphatic vessel consists of a single layer of cells with a large lumen;

iii) CNV clearly involves red blood cells, while CL does not (30,31). However, in addition to the difference between CNV and the formation of new lymphatic vessels, there are also differences in the signaling pathways that induce new blood vessels and lymphatic vessels; (iv) Studies on CNV indicate that lymphatic vessels affected by VEGF-A/VEGFR-2 are structurally and functionally different from those affected by VEGFR-3 ligands. These vessels exhibit a relatively dilated, leaky, and poorly functional phenotype (26). In terms of new corneal lymphatic vessels, Cao et al (33) reported that the lymphatic vessels affected by bFGF and VEGF-C were larger than those affected by bFGF alone, which may be related to the synergistic effects of different cytokines.

As early as 2002, Cursiefen et al (34,35) reported that although CL occurs simultaneously with CNV, CL degrades earlier than CNV. This result was obtained again in 2006. Although CNV and lymphangiogenesis generally occur together, subtle differences exist in the temporal regularity of angiogenesis in different models.

In the case of bacterial keratitis, CNV resulted in significant proliferation on the seventh day of modelling, but CL increased more slowly on the fourteenth day than did CNV (2). In the case of alkaline burns, although CL vessels grow parallel to CNV vessels, CL vessels sprout approximately at the third day after neovascularization occurs, but at the same time point of detection, the density and length of CL vessels are smaller than those of neovascularization vessels, and there is still a certain lag compared with CNV vessels (36). In the corneal suture model, the situation of CL vessels and CNV is similar to that of alkali burns. However, the generation of CL vessels is greater than that of alkali burns, and even on the third day, the area of CL vessel proliferation is greater than that of CNV (36,37). Although CNV and CL have distinct characteristics in these three types of patients, they are still concomitant overall.

In DED, only CL occurs, without CNV. This suggests that corneal neogenesis and lymphatic vessels grow in parallel or that the preference for blood vessels over lymphatic vessels is not universally applicable. In DED, this phenomenon has been proven to be related to specific cytokine pathways; for example, TH-17 cells secrete IL-17 directly, leading to an increase in VEGF-D, which promotes lymphatic vessel growth (27). SP promotes an increase in downstream VEGF-3 secretion in signaling pathways (28). There is currently no consensus on why CNV does not occur in DED. However, according to the experiments conducted by Chang et al (3), when the dose of bFGF was reduced to 12.5 ng, lymphatic vessel proliferation was significant, accompanied by only a small amount of neovascularization. The process of CNV is highly sensitive to dose, therefore it can be inferred that this phenomenon may be related to the concentration of cytokines that stimulate corneal growth. This is despite the fact that there are now cases where only CL are present. However, by reviewing the relevant articles, no case was found where only CNV occurs without CL. This idea is very novel and provides a direction for the future research.

The alkali burn model is a frequently employed model for studying chemical eye trauma in experiments. As early as 2010, Shi et al (36) confirmed the presence of blood vessels and lymphatic vessels in a corneal alkali burn model. The modelling method calculates the dosage according to the different animals, and then an appropriate dose of sodium hydroxide solution is used to directly drip into the eyes or soak the sodium hydroxide solution of this concentration on the upper filter paper and stick it on the animal cornea. After a period of time, the alkaline solution is discharged, or the filter paper is removed (35). Although this model is primarily used to investigate the treatment of corneal alkali burns, it is also utilized to examine the impact of various substances on blood vessels and lymphangiogenesis following an alkali burn. For example, Song et al (38) utilized this model to study the impact of leucine rich α-2-glycoprotein 1 (LRG-1) on CNV and lymphangiogenesis in the cornea (38). Compared with other methods, the model offers several advantages. First, the modelling method is simple and has minimal requirements for reagents and resources. Second, it is reproducible and can significantly reduce interference from factors other than variables. Third, it provides high controllability. Finally, the corneal alkali burn model can effectively simulate the condition of the cornea during alkali burns, which is very useful for studying the regulatory mechanism of corneal angiogenesis, the process of corneal inflammation and fibrosis, and the effects of related drugs. This method also has several limitations, such as its consistency being affected by equipment and drug specifications, the absorption of indoor air CO₂ by NaOH solution impacting the results, and other factors. Overall, this method has more advantages than disadvantages and remains a common technique for mold preparation (39).

Corneal suturing is another widely used method to study the relationship between the two parameters. It was first used to induce CNV and was later also used for studying CNV and the formation of new lymphatic vessels in the cornea (40). Its main molding method involves placing three stitches in the corneal stromal layer using a 10/0 nylon suture under a surgical microscope. The needle is then inserted into the center of the pupil from a distance slightly <2.0 mm from the corneal limbus (specific methods may vary for different animals). It is important to note that the specific method used may vary between animals. At present, this modelling method is mainly applied to study the roles of different cytokines in CNV and CL vessels. For example, Maier et al (41) used a corneal suture model to study the effect of TNF-α receptor deficiency on mouse CNV and lymphangiogenesis. This method is simple to use and prevents corneal damage caused by chemical reagents. The incidence of postoperative infection is low, but it can be influenced by the skill level of the operator and the size of the suture line and may induce protective eye rubbing in experimental animals, leading to model failure. Further research is needed to determine how to avoid these shortcomings.

Infection by various pathogens can lead to corneal inflammation, which in turn triggers the formation of new lymphatic vessels and blood vessels. At present, the most common model of bacterial keratitis is Pseudomonas aeruginosa-induced keratitis (42), while the most common model of viral keratitis is caused by the herpes simplex virus type 1 (HSV-1) (43). Currently, these two models are primarily used to investigate the impact of corneal blood vessels and corneal lymphatic vessels on the prognosis of patients with infectious keratitis. For example, Narimatsu et al (2) used Pseudomonas aeruginosa-induced infectious keratitis to study the impact of CL vessels on the prognosis of infectious keratitis. Moreover, these two models also have applications in studying the role of macrophages in neovascular and lymphatic vessels, as well as the factors that influence the timing of their formation (44). This method can effectively simulate the conditions of infectious keratitis. Because keratitis caused by different pathogens may require different treatment methods, the current approach is not reproducible and cannot establish a standard process. Therefore, further improvement is still needed.

In this novel modelling method, cytokine particles related to CNV and the development of CL vessels are implanted at the corneal edge to study the corresponding roles of these cytokines. The specific methods used are as follows: The sucralfate and cytokines are placed in a centrifuge tube and completely dissolved. A certain amount of hydron dissolved in anhydrous ethanol is added, and the mixture is dried on a nylon mesh to prepare particles containing cytokines. Cytokine particles are implanted into the corneal edge within 1 mm of the corneal stroma. At present, bFGF particles have been successfully implanted into the corneal margin in experimental research, and it has been verified that they can promote the production of blood vessels and lymphatic vessels (45).

With the widespread use of keratoplasty, there is an increasing interest in studying the role of neovascularization and lymphatic vessels in this procedure. The use of animals to simulate human corneal transplantation is also one of the currently available methods for inducing CNV and lymphatic vessel formation. Its main modelling method is to simulate the situation of corneal transplantation. For example, the specific process of simulating penetrating keratoplasty involves replacing and removing the corneas of both the donor and recipient mice, suturing the corneas of the donor mice to the eyes of the recipient mice, using nylon thread sutures, injecting sterile air and performing antibacterial treatment, and finally removing the thread. Similar to other types of surgical methods, the process of human corneal transplantation can be reproduced in animals. The application of this method has focused primarily on studying the relationship between the two in the context of corneal transplantation (46).

In addition to the aforementioned model, there are other models used to induce CNV and CL, such as the incision model and the limbal injury model. The incision model simulates corneal trauma by using a surgical blade after central cyclin labelling of the mouse cornea (47). The limbal injury model involves debridement after corneal injury to disrupt the integrity of the limbal epithelium to study its effect on corneal repair (48), although this approach is less commonly used.

There are typically no blood vessels present in the normal cornea. Due to its avascular nature, the cornea has immune privilege and a lower rejection rate. However, when the cornea is damaged, CNV and CL vessels are produced, forming the immune arms of the cornea. The CL vessels act as the afferent arms, transporting antigen substances and sensitizing lymph nodes, while CNV acts as the efferent arms, promoting activated immune cells to infiltrate the inflammatory site, which undermines the immune privilege of the cornea and reduces the success rate of corneal transplantation (1). According to the study by Zhang et al (77), CNV and CL are more prevalent in cases of corneal transplantation than in other models. This finding also highlights the significance of CNV and CL in the rejection of corneal transplantation.

At present, numerous researchers are also focusing on various methods to inhibit the generation and development of CNV and lymphangiogenesis vessels to improve the success rate of corneal transplantation. Reuer et al (67) utilized the combination of Sema3F and neuropirin to inhibit the activation of VEGFR and the production of CNV and lymphatic vessels, and especially after a period of treatment cessation, the graft survival rate was also very high, indicating that this treatment method has certain clinical application value. Ren et al (97) reported that myeloid-derived suppressor cells increase the expression of the antiangiogenic factor Tsp-1 and decrease the expression of the proangiogenic factors VEGF-A and VEGF-C through the iNOS pathway, inhibiting the generation of CNV and CL vessels. Ren et al (98) found that promoting the effect of myeloid suppressor cells through rapamycin can improve the success rate of corneal transplantation, while Wei et al (99) proposed that Cannabidiol can also enhance the effect of myeloid suppressor cells. In addition to myeloid suppressant cells, Kang et al (100) found that human umbilical cord MCSs and human adipose MSCs could also inhibit the formation of neovascular lymphatic vessels and corneal neovascularity to improve the survival rate of corneal transplantation, which was stronger than human adipose MSCs. Concurrently, Yu et al (101) found that insulin-like growth factor-1 modified mRNA can enhance the role of human adipose MSCs and promote corneal injury repair. Intravenous injection of cytokine-treated bone marrow cells can improve the survival rate of corneal transplant grafts (97). Zhu et al (102) also reported that isolated induced myeloid-derived suppressor cells could inhibit CNV through the iNOS pathway, improving the success rate of corneal transplantation. Shokirova et al (72) found that inhibiting adenylyl cyclase through Gαi/o using the κ opioid receptor agonist nanofuranin could suppress the production of VEGF and produce analgesic effects similar to those of opioid receptor agonists, but it does not have side effects caused by MOR activation, making it an ideal treatment for inflammation after corneal transplantation. Zhu et al (103) proposed that membrane collagen could be crosslinked with riboflavin and UVA to combat corneal inflammation. Subsequently, Hou et al (104) further confirmed that this therapy can promote the regression of mature CNV and CL vessels after transplantation and can be used to improve the survival rate of corneal grafts. Zhang et al (76) proposed that local application of VEGFTrapR1R2 can further inhibit CNV and CL vessels by specifically inhibiting VEGF, thereby improving the survival rate of corneal transplantation (98). Unlike Zhang et al (76), the soluble form of VEGF-C/D used by Salabarria et al (79) is VEGFR-3. It can inhibit the formation of new lymphatic vessels and angiogenesis in the cornea after transplantation, but it cannot promote corneal graft survival. In addition to acting on VEGF itself and its upstream pathway, blocking its downstream pathway is also feasible at present. Dietrich-Ntoukas et al (105) found that CNV can be inhibited by blocking Integrin-α5β1 downstream of VEGF.

Infectious keratitis is caused mainly by infections caused by bacteria, viruses and fungi. Pseudomonas aeruginosa is a common pathogen of bacterial keratitis and can cause keratitis in individuals who wear contact lenses for a long time. LPS from Pseudomonas aeruginosa can activate VEGF-C through the TLR4 pathway and subsequently induce lymphangiogenesis. The main pathway by which Pseudomonas aeruginosa promotes CL is also VEGF-C/VEGFR-3. However, research has shown that in bacterial keratitis, lymphatic vessels not only play a proinflammatory role but also, to some extent, reduce macrophage infiltration and alleviate corneal edema (2,47,63).

Herpetic keratitis is a type of infectious keratitis caused by HSV-1 (106). It is the primary cause of infectious corneal blindness in developed countries. Unlike bacterial keratitis caused by Pseudomonas aeruginosa, the lymphangiogenesis effect of HSV-1 does not rely on VEGF-3, VEGFR-3, or macrophage infiltration but mainly relies on VEGF-A and VEGFR-2. VEGF-A can also promote CNV. Studies have revealed that lymphangiogenesis promotes the progression of herpetic keratitis, resulting in visual damage by directing destructive inflammatory mediators to infected areas. Compared with the visual impairment caused by HSV-1-induced CNV, CL is more harmful (44,107). In addition to affecting VEGF, HSV-1-induced keratitis is also closely related to chemokines, especially CXCL-10, which is also related to CC chemokines but not as closely related to CXCL-10. Both factors affect the infiltration of leukocytes, which in turn affects the generation of CNV and CL vessels (108).

At present, in the treatment of infectious keratitis, researchers are exploring new treatments in addition to the traditional approach of using antibiotics for anti-inflammatory purposes and using glucocorticoids to inhibit CNV and CL. Zhu et al (103) proposed that corneal collagen can be cross-linked with riboflavin and UVA to treat infectious keratitis, which has the advantages of direct sterilization, reducing inflammatory cell infiltration, and inhibiting CNV and CL vessel formation. Shokirova et al (72) reported that κ opioid receptor agonists can also be used to combat infectious keratitis. This is due to their ability to inhibit transplanted CNV, promote lymphangiogenesis, and provide analgesic effects. For herpetic keratitis, the most commonly used method is to inhibit the generation of VEGF-A. Gurung et al (44) reported that regulating FGF-2 can indirectly inhibit the generation of VEGF-A, and this method can partially preserve the vision of HSV-1-infected mice, which has certain clinical value. However, inhibiting FGF-2 can also lead to nerve degeneration and has potential side effects. However, further research is needed to determine its effectiveness for clinical application. The CC chemokine ligand ACKR-2 is considered to be an important treatment for HSV-1 keratitis. It specifically binds to CC chemokines, reducing the number of free CC chemokines and promoting the resolution of inflammation. Recent findings have shown that it can also bind to CXCL-10, a factor involved in the pathogenesis of HSV-1 keratitis, further demonstrating that ACKR-2 can effectively inhibit HSV-1-induced keratitis (95,108).

The order of formation of CNV and CL is different, but they often occur together in corneal diseases. However, in DED, only lymphangiogenesis occurs. Currently, the mechanism underlying the selective generation of corneal lymphatic vessels in DED is not fully understood. The results of the experiments conducted by Chang et al (3) may be related to differences in the sensitivity of corneal lymphatic vessels and CNV to cytokine dosage. Current treatment methods also mostly exert their effects through the side effects of anti-CNV and anti-CL (28).

Corneal alkali burn injury is a commonly used technique for establishing mouse CNV and CL models. It is detected in some laboratories and chemical plants and can cause accidental damage. Due to the lipophilicity of alkali, it often leads to more severe keratitis than acid burns (109). Li et al (93) reported that knockout of the CXCR-3 gene in an alkali burn model exacerbates CNV and CL in mice. This may be related to the gene's regulation of VEGF function. Song et al (38) reported that LRG1 also regulates the expression of VEGF and VEGFR, as well as CNV and CL, in alkaline burn models. Additionally, Oh et al (110) targeted DDR1 through microRNA-199a/b-5p and inhibited DDR1 expression, ultimately reducing lymphangiogenesis and neovascularization in alkaline burn models. However, the mechanism by which DDR1 promotes CNV and the formation of new corneal lymphatic vessels remains unclear. At present, for the treatment of CNV and CL after alkali burn, Li et al (111) proposed that shark cartilage-derived anti-angiogenic peptide could be used, because it could inhibit VEGF, MMP1 and other factors.

In addition to the aforementioned diseases, CNV and CL also play significant roles in the development of other conditions, such as allergic keratitis and corneal edema, which are associated with angiogenesis and lymphangiogenesis. However, recent advancements in research on this topic have been limited, and the present topic will not be further discussed.