Participation of Intussusceptive Angiogenesis in the Morphogenesis of Lobular Capillary Hemangioma

In lobular capillary hemangioma (LCH), misnamed pyogenic granuloma, only sprouting angiogenesis (SA) has been considered. We assess the occurrence of intussusceptive angiogenesis (IA) in LCH and whether IA determines the specific and other focal patterns in the lesion. For this purpose, we study specimens of 120 cases of LCH, using semithin sections (in 10), immunohistochemistry, and confocal microscopy (in 20). In addition to SA, the results in LCH showed (1) intussusceptive phenomena, including pillars/folds and associated vessel loops, which encircled interstitial tissue structures (ITSs). (2) Two types of evolved loops depending on interendothelial contacts from opposite walls: (a) numerous interendothelial contacts, alternating with capillary-sized lumens (main capillary pattern of the lesion) and (b) few interendothelial contacts, wide open lumens, and intravascular transport of pillars/folds, which were arranged linearly, forming septa (focal sinusoidal-like pattern) or were irregularly grouped (focal intravascular papillary endothelial hyperplasia, IPEH-like pattern). In conclusion, we demonstrate that IA participates in synergistic interaction with SA in LCH development and that the prevalence of specific intussusceptive phenomena determines the predominant capillary pattern and associated sinusoidal hemangioma-like and IPEH-like patterns in the lesion, which suggest a role of IA as conditioner of vessel tumour/pseudo-tumour morphology.

findings and similar histogenesis, combining sprouting and intussusceptive angiogenesis 16 . We also proposed that the pattern of the processes outlined above is determined by the prevalence of certain intussusceptive findings. Thus, the morphologic pattern of IPEH and sinusoidal hemangioma depends on the expression and arrangement of pillars and associated structures: pillars (papillae) irregularly grouped in IPEH and incomplete intravascular septa formed by linearly grouped pillars in sinusoidal hemangioma 14,16 .
Lobular capillary proliferations are observed in some vascular tumours/pseudotumours. An example is the lobular capillary hemangioma (LCH), misnamed pyogenic granuloma. LCH grows rapidly, presents mitotic activity and can regress spontaneously. SA has been demonstrated in this process, whereas IA has not been considered. A study to assess whether IA occurs in LCH with such a different lobular capillary pattern from IPEH and sinusoidal hemangioma is not only of interest to explain LCH histogenesis, but also to contribute to the concept that variations of a common intussusceptive mechanism can condition the morphologic expression of vessel tumours/ pseudotumours. To reinforce this hypothesis, this study should be expanded to investigate the presence of zones with other histologic patterns in LCH.
Given the above, the objectives of this study are (a) to assess whether IA participates in LCH and whether there is synergistic interaction with SA, (b) to explore the coexistence of other histologic patterns in LCH lesions, and c) to establish whether the prevalence of certain intussusceptive findings determines the morphological pattern(s) in LCH.

Results
General characteristics of LcH. Neovascularization in all cases of LCH (n: 120) adopted a predominantly lobular pattern (Fig. 1A,B). Zones with sinusoidal hemangioma-like (Fig. 1C) and IPEH-like (Fig. 1D) morphology were also seen. These zones were more evident in 20 cases of which 11 showed a sinusoidal hemangioma-like aspect and 9 an IPEH-like aspect. Occasional, small foci (low presence) of these associated patterns were seen in 32 cases (in 18 sinusoidal hemangioma-like and in 14 IPEH-like). In the remaining cases, no associated patterns were observed (see Table 1). In the predominant LCH pattern, each lobule showed numerous capillary-sized vessels and one variably branched venule (Fig. 1B), generally located in the centre. These vessels were lined by ECs expressing CD34 (Fig. 1B,E). Anti-αSMA + mural cells (vascular smooth muscle cells and pericytes) were observed in the venules (Fig. 1E) and in capillaries (Fig. 1B). The lobules were separated by fibrous septa (Fig. 1A). A mixed inflammatory infiltrate and fibrin deposits were seen, predominantly in cases with ulceration (Fig. 1F). SA events were present, including outward growth from the vessel mother (Fig. 1G), with presence of tip endothelial cells (ECs) sprouting toward the interstitium (Fig. 1G), basal lamina degradation (insert of Fig. 1G), formation of a bilayer of stalk ECs (Fig. 1H), presence of mitoses in the stalk ECs ( Fig. 2A) and in the ECs of the mother vessels (Figs. 2B,C) and in pericytes (Fig. 2D), incorporation of pericytes around the stalk ECs and formation of a new vessel lumen ( Fig. 2E-G) with presence of red blood cells in their lumens (Fig. 2E). In addition, the proliferation index (ki67) was moderate/high in some loops (stalk cells) (insert Fig. 1H) and mother vessels (insert Fig. 2C).
We also observed structures typical for IA, the main focus of our study, including pillar-associated structures and pillars/folds. The pillar-associated structures comprised vessel loops (  Table 1). These findings will be described below.
Vessel loops in LcH. Numerous elementary and complex vessel loops, originating from venules or from other loops, were observed in histological sections (Figs. 2H-K and 3A-D). The loops were lined by ECs, which also expressed CD34 (Fig. 2H). Anti-αSMA + pericytes and/or their processes could be present around the ECs (Fig. 2H). SA phenomena were observed in the mother vessels and in the growing loops (see above). Serial semithin sections, and single and whole-mounted images in confocal microscopy revealed that the double layer of the loop ECs was formed by endothelial sheets (Figs. 2I-K and 3A-D). Each loop, which had two segments connecting to the mother vessel ( Fig. 2I-K), dissected and encircled an ITS (Fig. 2H-K). contacts between ecs from opposite walls of the loop. Contacts were observed between ECs from opposite vessel walls of the loops (Fig. 3E-Z). These contacts varied depending on the morphology of the contacting cells, contact extension, symmetry or asymmetry, and distribution. In histologic sections, the contacting ECs could have a triangular, ovoid or flattened morphology ( Fig. 3E-G). Contact surface extension was related to EC morphology: generally small when ECs where triangular (small apical contact) (Fig. 3E) and extensive when ECs were flattened (extensive planar contact) (Fig. 3G,H). In the symmetric contact (kissing contact), ECs protruded from both opposite walls of the vessels (Fig. 3E,F), whereas in the asymmetric contact, ECs only protruded from one side of the vessel wall (Fig. 3I,J). Depending on their distribution, the contacts could appear relatively grouped (multiple contacts) (Fig. 3K) or isolated (single contact) (Fig. 3I). ECs showing a triangular morphology in multiple symmetric contacts acquired a double serrated image. Contacts could also be formed by ECs or by EC protrusions that surrounded connective tissue, acquiring a peg-like image (Fig. 3L). Occasionally, extremely fine transcapillary EC projections adopting a filiform or antenna-like morphology were observed (Fig. 3M). In electron microscopy, some of these structures appeared as endothelial projections separating luminal spaces (Fig. 3N). Serial semithin sections showed the appearance and disappearance of these contacting cells (Fig. 3O-R). Likewise, in confocal microscopy, we also observed the appearance and disappearance in series of individual views ( Fig. 3T-Z) and in whole-mounted view (Fig. 3S) of some of the aforementioned EC contacts. Perforation of endothelial contacts was seen in serial semi-thin sections (Figs. 4A,B). In these perforations (Fig. 4B), the core was formed by cytoplasmic extensions of pericytes, interstitial and inflammatory cells, and by collagen fibrils. The appearance and disappearance of these perforated interendothelial contacts were demonstrated in sequential images in confocal microscopy ( Fig. 4C-G). intravascular pillars in vessel loops with scarce or no interendothelial contacts. Zones with sinusoidal hemangioma-like and ipeH-like morphology. In vessel loops with scarce or no interendothelial contacts and wide lumen, small (≤2.5 µm) and large (>2.5 µm) pillars were formed in the vessel lumen (Fig. 5). Intravascular pillars showed a cover and a core, and were irregularly distributed. The cover was formed by ECs expressing CD34 (Fig. 5). These ECs corresponded to the loop inner layer. The content of the core varied depending on pillar size. The smaller of these structures presented a core that contained collagen material and cell extensions of pericytes and fibroblasts. The larger showed connective tissue components (Fig. 5), occasionally with small capillaries in giant pillars. Nascent pillars formed by extensions of ECs were also seen (Fig. 5C). Ultrastructurally, pillars characteristics were confirmed, observing covering ECs and processes of pericytes and collagen in the pillar core ( Fig. 6A and insert). In confocal microscopy, we observed the appearance and disappearance in series of individual views of neighboring pillars (Fig. 6C-K), as well as their image in whole-mounted view (Fig. 6B).

Loop fragmentation in capillary-sized vessels with numerous interendothelial contacts.
Pillars could appear isolated in the vessel lumen (Figs. 5H,I) or adhered to other structures (Fig. 5F,G), such as the vessel wall and other pillars (Fig. 5J,K). When the intravascular pillars appeared joined together, they adopted a linear arrangement, forming incomplete septa (Fig. 7A-J), or were grouped irregularly (Fig. 7K,L). The septa frequently had a moniliform surface and were covered by CD34 + ECs (Fig. 7A,B) and the core presented αSMA + cells with a pericyte-like aspect, together with other stromal cells. Communications between opposite endothelial cells within each septum and zones of continuity and discontinuity between the pillars that formed each septum were observed in serial semithin sections and in confocal microscopy ( Fig. 7C-J). The zones presented above with a rise in these phenomena showed a sinusoidal hemangioma-like morphology (presence of incomplete septa by linear arrangement of pillars) (Figs. 1C, 7A,B) or IPEH-like morphology (numerous pillars grouped irregularly) (Figs. 1D and 7K,L). In IPEH, large pillars connected by thin pillars were frequently observed (Fig. 7L). www.nature.com/scientificreports www.nature.com/scientificreports/   www.nature.com/scientificreports www.nature.com/scientificreports/

Discussion
In this work, we report the participation of IA in synergistic interaction with SA in LCH and confirm several morphogenic mechanisms involved in the formation of pillars (hallmarks of intussusception) and capillary-like structures. In addition, we demonstrate that the prevalence of some intussusceptive mechanisms over others determines the morphologic patterns in LCH (predominant capillary pattern and focal sinusoidal hemangioma-like and IPEH-like patterns). Based on these findings and on previous results in sinusoidal hemangioma and in IPEH 14,16 , we suggest that intussusceptive mechanisms may participate in the morphogenesis of vascular tumours/pseudotumours in general. Finally, we take into account the possible histogenic implication of the venules in the lesion origin and lobular aspect.
The presence of pillars, folds and associated morphogenic structures, typical for IA, in LCH confirms that IA participates in LCH formation, a role previously only assigned to SA. The demonstration of these intussusceptive findings was undertaken in serial semithin sections, and in successive and whole-mounted images in confocal  www.nature.com/scientificreports www.nature.com/scientificreports/ microscopy. The combination of SA and IA in LCH concurs with that observed by different authors in the chick chorioallantoic membrane, and in developmental avian kidney and lung 2,13,18,19 . This synergistic action between SA and IA has also been observed in the rat femoral vein after PGE2 and glycerol perivenous administration and in the zebrafish caudal vein plexus 15,17 . By these complementary mechanisms, in an initial phase, vessels mainly grow by sprouting, followed by intussusception. The presence of blood red cells in the lumen of the capillary-sized spaces formed in the loops suggests that the loops are generally perfused and kissing or peg-like contacts are established. Secondary structures formed by fusion or splitting of pillars can also occur, originating intravascular meshworks of processes, septa or irregular aggregates (see below).
The intussusceptive precursor findings involved in pillar morphogenesis in LCH coincide with those previously proposed by pioneering authors studying intussusception 3,6,10-13,20 and their frequency in LCH indicates that this lesion is an appropriate substrate for the study of these morphogenic mechanisms. Thus, the findings in this work support pillar formation by kissing endothelial and peg-like contacts, meso-like intraluminal folds, split intercapillary meshes and vessel loops encircling ITSs.
Loops encircling ITSs and interendothelial contacts between the opposite walls of the loops were common findings in LCH. The frequency of interendothelial contacts between the opposite walls of the loops conditions the LCH patterns (see scheme in Fig. 8 and Table 1). When numerous interendothelial contacts occur in the loops, and the contacts are perforated, the intercalated open lumens of the loops between the contacts persist as capillary-sized vessels (main capillary pattern of the lesion) arranged according to the primitive loop path. When few inter-endothelial contacts occur, the loop lumen is patent, and the inner side of the loop forms the cover of intraluminal pillars (the core corresponds to the ITS encircled by the loop), which may be arranged linearly, forming septa (focal sinusoidal-like pattern) or may be grouped in an irregular form (focal IPEH-like pattern). These intussusceptive mechanisms can be linked to hemodynamic conditions in the loops and mother vessels, as occurs in those described in vascular expansion and branching remodelling 1,6,21-26 , including the influence of intraluminal flow fields 1,23,25,26 . In A2, perforation of an interendothelial contact. In A3, loop segmented in capillary-like structures, acquiring the capillary aspect of LCH. These images only represent one loop. Spatially, however, multiple converging loops occur in the lesion, and the capillary-sized spaces may connect with others in other loops. These findings could explain the future conservation or involution of these spaces. Note that the ITS surrounded by the loop remains outside the mother vessel (extravascular). (B) Several steps of pillar formation in the mother vessel are shown in B1, B2 and B3, with their corresponding microphotographs. The loop originating from the mother vessel (B1) becomes permeable (B2), and the ITS and its surrounding endothelium (internal side of the loop) is transported to the mother vessel lumen (although some connections between the pillar and vessel wall may persist). In B3, several pillars can be arranged linearly or irregularly, originating the sinusoidal-hemangioma-like (B3a) or IPEH-like (B3b) pattern, respectively. (2020) 10:4987 | https://doi.org/10.1038/s41598-020-61921-3 www.nature.com/scientificreports www.nature.com/scientificreports/ This work and previous studies 14,16 show that LCH, sinusoidal hemangioma and IPEH present different morphologic patterns and share hallmarks of intussusception, in which vessel loops have an important morphogenic role. A similar participation of vessel loops has been demonstrated in the ovarian vein of the rat after ovariectomy and human tumour xenograft implantation 10,11 and in other human blood and lymphatic vessel diseases, such as hemorrhoidal disease, lymphangiomas/lymphatic malformations and vascular transformation of lymph node sinuses, as well as in some sinuses of developing human foetal lymph nodes 5,14,[27][28][29] . We use the term "piecemeal form of intussusceptive angiogenesis" to describe the mechanism of the formation and intravascular transport of pillars by vessel loops 5,[14][15][16]30 . The morphologic differences of sinusoidal hemangioma, IPEH and LCH depend on the number of interendothelial contacts between the opposite walls of the loops and intravascular pillars formed in/from vessel loops, as well as on the substrate in which vessel loops and pillars form and on the different arrangement of pillars (see Fig. 8 and Table 1). Therefore, the pattern of these lesions depends on the predominance of certain intussusceptive mechanisms. Further studies are required to assess whether intussusception participates in other vessel tumours/pseudotumours and whether the prevalence of certain intussusceptive mechanisms also conditions their morphologic pattern.
Importantly, LCH is a process that can show spontaneous regression with relative frequency. Intussusceptive branching remodelling is a mechanism described in vessel bifurcations, with regression, retraction and atrophy of superfluous vessels [6][7][8] . The transcapillary pillars in LCH could somehow achieve the same effect as the formation of transvascular pillars in vessel bifurcations (now occurring as multiple pillars in smaller vessels), which would explain LCH regression. This issue also requires further study.
The observation of a variably branched venule giving rise to numerous microvessels in each LCH lobule suggests that the lobular pattern of the lesion depends on its multi-venular origin. Namely, each primitive venule originates one lobule, which is separated from other lobules by primitive inter-venular connective tissue. The central or eccentric location of the venule in the lobule can depend on whether capillaries originate from the entire venule girth or from one side, a process that can be observed in initial phases of lobule formation. Subsequent intense capillary formation from other neo-vessels increases the complexity of lobule architecture. In some lobules, the branched venule and new vessels with virtual lumens form a highly cellular structure with intercalated venular and capillary lumens.

Human tissue samples. The archives of Histology and Anatomical Pathology of the Departments of Basic
Medical Sciences of La Laguna University, University Hospital, and Eurofins ® Megalab-Hospiten Hospitals of the Canary Islands were searched for cases of LCH for the period 1990-2017. Paraffin blocks were obtained from surgical specimens of 120 Caucasian patients: 56 males (46.66%) and 64 females (53.33%), ages ranging from 5 to 75 years, predominantly in the second and third decades of life (48 cases). Lesion evolution ranged from 12 days to 3 years. The samples of the 120 cases were studied by conventional histologic techniques. From them, 20 cases with more evident zones of sinusoidal hemangioma-like (n: 11) and IPEH-like (n: 9) patterns were used for immunochemistry procedures and immunofluorescence in confocal microscopy. Likewise, 10 cases were obtained for serial semithin and ultrathin sections. Most of the specimens selected corresponded to the earliest lesions (n: 18). Ethical approval for this study was obtained from the Ethics Committee of La Laguna University (Comité de Ética de la Investigación y de Bienestar Animal, CEIBA2019-0339), including the dissociation of the samples from any information that could identify the patient. The authors therefore had no access to identifiable patient information.
Specimens for semithin sections (1 µm) were fixed in a glutaraldehyde solution, diluted to 2% with sodium cacodylate buffer, pH 7.4, for 6 hours at 4 °C, washed in the same buffer, post-fixed for 2 hours in 1% osmium tetroxide, dehydrated in a graded ethanol series, and embedded in epoxy resin. Serial semithin sections were mounted on acid-cleaned slides, stained with 1% Toluidine blue (Merck ® ), and observed under a Leica ® DM-750 light microscope with an integrated High Definition Camera.
immunofluorescence in confocal microscopy. For immunofluorescence, tissue sections of 6 and 10 µm were obtained. For antigen retrieval, sections were deparaffinized and boiled for 20 minutes in sodium citrate buffer 10 mM (pH 6), rinsed in Trisbuffered saline (TBS, pH 7.6, 0.05 M), and incubated with the following primary antibodies diluted in TBS overnight in a humid chamber at room temperature: mouse monoclonal