(小児)脳腫瘍の組織学

(1-2:Histological feature)
Pediatric brain tumors fall into different categories compared to adult brain tumor where gliomas are the most common type of brain tumor. Pediatric brain tumor is more diverse than adult one(8). The most common type of pediatric brain tumors are high-grade gliomas, medulloblastomas, low-grade gliomas (astrocytic, oligodendroglial and mixed glial-neuronal), ependymomas and brainstem glioma and diffuse intrinsic pontine gliomas. In adults, there are over 120 histological different brain tumors(9) in a finely observed way. Therefore, it is difficult to unravel the whole histological details and some degree of ambiguity remains because of fine analysis from raw data statistically insufficient to find general feature in childhood brain tumor. On the other hand, representative histological type is defined such as neuroepithelial type (asrocytic (pilocytic astrocytoma, well-diff, low grade astrocytoma, anaplastic astrocytoma, glioblastoma multiforme), epndymoma, choroid plexus papilloma, oligodendroglioma, mixed gliomas, medulloblastoma, anaplastic angliogioma, pineal parenchymal tumor of intermediate diff), non-neuroepithelial type(craniopharyngioma, meningioma, schwannoma, lymphoma)(10).
(1-2-1:Histological analysis based on 2 major group)
Neuroepithelial type histologically exhibits low cellularity(low concentration of cancer cell), often contain Rosenthal fibers and present with mural nodules(11). Rosenthal fiber is found in astrocytic process and ae thought to be clumped intermediated filament proteins composed of glial fibrillary acidic protein. This protein condensate is actually observed in neuroepithelial tumor type as dark-colored fiber shape. Additionally, concentration of cancer cell is low meaning low cellularity. The space, which is observed as white color, is relatively high. These histological features may be related to favorable prognosis, which is indicated in dysembryoplastic neuroepithelial tumor in actual clinical and epidemiological case. In fact, decipherable histological feature of high-grade neuroepithelial tumor presents high cellularity. Mural nodules is one of nodules formed on the wall of vascular system like pericyte. Nodules are small film lumps. Therefore, mural nodules can be observed as raised lump on brain vascular wall. However, observation of mural nodules is difficult from histological data.
The histology of non-neuroepithelial tumor isn’t commonly defined, so the histological feature of several subtypes such as craniopharyngioma, meningioma, schwannoma, lymphoma will be explained in detail below. 
Craniopharyngioma are intracranial tumor that is mainly presented in the sellar region and the parasellar region. Sellar region is in deep intracranial region where convergent crossroad of peripheral neural, meningeal, vascular, bone and soft tissue structures. The parasellar region surrounds sellar region and hypothalamus. Histologically, they are benign tumors wit distinct adamantinomatous and papillary subtype. In pediatric brain tumor, adamantinomatous type only arise, and in adult brain tumor, this type is also dominant. Admantinoumatous type includes wet keratin, which is calcium-rich desquamated cell. Desquamated cell is a cell peeled off from epithelial cell. And satellite reticulum is one of the composed layers in admantinoumatous type of craniopharyngioma. Papillary type includes sheets of squamous epithelium(13). Craniopharyngioma is generally considered as a pediatric disease accounting for 1.2-4% of all intracranial tumors. It has a classical bimodal age distribution with an increased incidence rate in 5 to 14 years and 50 to 74 years of age(14).
Meningioma has widely variety histological subtypes recognized by WHO. As WHO grade 1, meningothelial (syncytial) meningioma, transitional(mixed) meningioma, fibroblastic(fibrous) meningioma, psammomatous meningioma, angiomatous(vascular) meningioma, microcystic meningioma, secretory meningioma, lymphoplasmacyte-rich meningioma, and metaplastic meningioma is categorized. As WHO grade 2, chordoid meningioma, clear-cell meningima and atypical meningioma fall into the histological subtype. As WHO grade 3, papillary meningioma, rhabdoid meningioma and anaplastic(malignant) meningioma are categorized. A peak age of overt meningioma onset is 41 to 50 years of age accounting for about 50% of all incidences and onset of female account for about 70% of all cases(16). Pediatric meningioma is rare. (To be continued~)
Schwannoma
Lymphoma
(1-2-2:Histological analysis based on pediatric cancer type)
(1-2-2-1: Pilocytic astrocytoma and determinant factor of fibrillary cell shape)
Pilocytic astrocytoma(PA) arise from reactive astrocyte. Astrocytoma is most common type of glioma. It is typically located in the midline structure, such as (the most commonly) the cerebellum, optic pathway, hypothalamus, basal ganglion and thalamus. Histological feature of pilocytic astrocytoma is bipolar spindle piloid cells with long fibrillary shape, which is origine of name “pilocytic” astrocytoma(109, fig). This is typically benign and slow growing tumor, corresponding to WHO malignancy grade 1(110). One important factor indicating benign tumor is that pilocytic astrocytoma has long fibrillary structure and typically low to moderate cellularity(111 Fig.2). Pilocytic astrocytoma occurs most commonly in children and younger adults. As indicated in naming of this tumor, pilocytic astrocytoma has distinct cellular shape that is long spindle, fibrillary feature with extremely high aspect ratio like axon. What exists in cytoplasm of this fibrillary portion? For example, the unique morphology of axons are sustained by the organization of the cytoskeleton such as microtubules, neurofilaments and actin(112). As shown in Figure 1 by ref.112, the microtube structure involving skeleton of the axon has many microtubule fascicles(small bundle). This microtube structure and shape of initial and distal segment in axion may be different(112). This cytoskeleton is highly related to molecular motor(112) which is energetically driven by ATP of mitochondrial production. This axon molecular feature could be similar with the fibrillary feature of pilocytic astrocytoma, if so, the distribution and locational stability of mitochondria may play a pivotal role in the maintenance of spindle shape of cell. This positional stability of mitochondria is realized by the ATP periodical fluctuation(113). This spindle feature may be associated with low and moderate cellularity indicating benign tumor, because this spindle structure contributes to repulsion and long distance of main cellular portion including cellular nucleus. In other words, the adhesion among many cancer cells with high cellularity may be positive proportion to malignancy of cancer(114). Therefore, understanding the mechanism determining the distribution of mitochondria, ATP, cytoskeleton(112,113) may confer the novel therapy to decrease cancer cellularity by promotion of spindle structure of glial cell with extremely high aspect ratio including not only this pilocytic astrocytoma, but also the other high-grade glioma, if my hypothesis is relatively correct. This means that the repulsive force among adjacent cancer cells is driven from the “inside” of cancer cell by the controlling of the distribution of ATP, mitochondria, cytoskeleton against astrocyte with originally spindle feature. At least, large production capacity of cytoskeleton is necessary to form the long fibrillary structure like axon(112). The cellular origin of pilocytic astrocytoma(PA) is reactive astrocyte. Reactive astrocyte typically indicates hypertrophy of dendritic structures, which is due to the enhancement of expression of glial fibrillary acidic protein(GFAP)(115). However, this GFAP level of glioblastoma is significantly related to poor prognosis(116), this could be due to conferring the strong mechanical properties or/and enhancement of cellular interaction between tumor cells. At least, we need to analyze histological feature including cellular shape (whether the tumor has long fibrillary shape or not) and cellularity in accordance with overexpression of GFAP. Histologically, pilocytic structure and dendritic structure associated with GFAP is clearly different. Therefore, finding the molecular maker of specific pilocytic structure is crucial.
(1-2-2-2: Pleomorphic xantho-astrocytoma)
Pleomorphic xantho-astrocytoma is superficially located tumor with close relation to the meninges that predominates in the temporal lobe(12). Therefore, in many case, enucleation of this type tumor is possible by surgical procedures. This tumor is assigned as grade 2 by WHO. The typical clinical manifestation is epilepsy(130). Histological feature is pleomorphism clearly indicated names (“pleomorphic”) that spindle cells closely intermingle with small and large mononuclear and multinucleated bizarre tumor giant cells. This spindle portion shows yellow like feature indicated “xantho”-astrocytoma.
(1-2-2-3: Sub-ependymal giant cell astrocytoma (SEGA))
This intra-ventricular tumor is typically associated with tuberous sclerosis complex. Tuberous sclerosis complex (TSC) (nodule hardening) with calcification in some case is a rare multisystem autosomal dominant (one allele mutation) genetic diseases that cause non-cancerous (non-malignant) tumor. Hence, sub-ependymal giant cell astrocytoma is benign tumor and assigned as grade 1 by WHO, which is taken on exclusively in younger person with tuberous sclerosis. It exhibits proliferation of three cell types; large gemistocytes -like cells with perivascular pseudorosette pattern, long spindle fibrillary astrocytes arranged in broad fascicles and giant cells, some with ganglioid appearance. Gamistoyte is a swollen reactive astrocyte. 
(1-2-2-4: Fibrillary astrocytoma)
Fibrillary astrocytoma is the most common histological subtype of diffuse or low grade astrocytoma (Grade 2 by WHO). The outline of fibrillary astrocytoma is not clearly visible in scans, because the borders of the cancer cell tend to be tiny microscopic fibrillary tentacles intermingling with healthy brain cells. Therefore, complete surgical removal of only fibrillary astrocytoma is difficult. Morphologically there is proliferation of “well-differentiated“ fibrillary astrocytes with elongated, irregular and hyperchromatic nuclei exhibiting angulated contours with many coma-shaped forms that lack nucleoli.
(1-2-2-5: Anaplastic astrocytoma (AA))
Anaplastic astrocytoma occupies an intermediate position between fibrillary astrocytoma and glioblastoma. Features of anaplasia include higher cellularity, greater degree of pleomorphism and increased proliferation. 
((12) note Anaplastic astrocytoma)
The exact number of mitotic figures needed to separate this from FA is still debatable and this feature should be evaluated in relation to the amount of tissue sample examined. While the presence of a single mitotic figure in a small stereotactic biopsy justifies assigning grade III to a tumor, the presence of a single mitotic figure in an ample biopsy after careful searching and deeper sections might not be as relevant. In one study; >3MF/10HPFs was the cut-off value proposed [27]. Ancillary studies might help in defining the proliferative activity of a tumor and can be used to support the diagnosis of AA. Although cut-off values are variable an elevated MIB-1 labeling index (>9%) and proliferative activity as measured by PHH3 mitotic index (>4per 1000 cells) were found to be supportive of AA diagnosis over FA, in which MIB-1 and PHH3 labeling indices were low (≤ 9% and ≤4 per 1000 cells) [27]. In addition; MIB-1 labeling index prognostic value independent from histologic grade was reported [14], which might be indicative of an early anaplastic transformation [31], even in the absence of detectable mitoses.

Medulloblastomas is a primitive neuroectodermal tumor arising from cerebellum. A primitive neuroectodermal tumor are now under the classification of embryonal tumors. Ectoderm is one of the three primary germ layers formed in early embryonic development, which can be differentiated to neuron, oligodendrocyte, astrocyte through neural stem cell that form brain. Medulloblastoma is the most common malignant childhood brain tumor and the most common embryonal tumor. The cellular origin of medulloblastomas is the muti-potent progenitor cell-type of cerebellar ventricular zone – Basket cells, Dtellate cells and Purkinje cells which is classified to GABAergic neuron(56) – that forms the innermost boundary of cerebellum(12). These origin is postulated as the classical medulloblastoma, while the cancer arising from the external granular layers is postulated origin of the desmoplastic medulloblastoma. Desmoplasia is a increased stromal cellularity that describes the formation of fibrous connective tissue. The classical medulloblastoma has densely packed form composed of hyperchromatic, round and oval shaped cells, and carrot shaped nuclei with minimal cytoplasm. Medulloblastoma is classified as grade 4 by WHO. This may be due to the histological fact that this tumor typically has high cellularity (packed form). High malignancy of medulloblastoma may be highly related to genetic predisposition with “germline mutation” such as APC, BRCA2, PALB2, PTCH1, SUFU and TP53(57). 
Ependymomas
(12)
(1-2-3:Comprehensive consideration from histological analysis)
 The pediatric brain tumor is histologically diverse, so presenting the typical features determining malignancy is difficult. However, obviously, the aggregating level of cancer cell defined as cancer cellularity is highly related to cancer grade, which is indicated in oligo-dendroglioma, IDH-mutant astrocytoma and IDH-wildtype astrocytoma, which can be evaluated in comparative histological pictures for each grade(17). If the absolute number of cancer cell per an arbitrary area is large, the proliferation rate of tumor is obviously increased. In addition, accessibility of drug especially toward deep layer decreases. In brain cancer, low-grade tumor is not related to “absolute size” as a tumor tissue, but independency meaning that many cancer cells are highly exposed of stroma. Therefore, the medical model such as CAR immune therapy and cancer evolution model like leukemia could be applied. If histological independency of each cancer cell is related to gene evolution in a manner of highly clonal expansion like leukemia(21), high-grade glioma, which may have both boundary growth mode where there is not grand and adenocarcinoma-like tumor type where there is a gland like hypothalamus have distinct genetic evolution mode of cancer cell, the evolution mode of which may be drastically changed whether several gland exists or not(21). On the other hand, high grade brain cancer like glioma is highly connective among cancer cells, so scrutinizing adhesion mechanism such as morphogenesis, adhesion molecules may be crucial to find both the novel target of medical treatment and take prophylactic measures especially against progression from low-grade to high grade especially for the children congenitally having the several driver gene mutations of brain cancer. As mentioned above, not only neuron, (but also astrocyte, oligodendrocyte, microglia) are the cellular origin of pediatric brain cancer. All of these cell-types may have reginal specific phenotype(4). In a brain, selectively connective mechanism may be important obviously in a neuron. Connective mechanism of high-grade brain tumor is more important than low grade one due to dense distribution. Especially, connective mechanism through surface protein like cell adhesion proteins may be regional specific, which pave the pathway to find the novel regional specific target also in high grade tumor organized arbitrary intercranial region. Notably, the intracranial region progressing to overt brain cancer is more diverse in the pediatric case than in the adult one. Whether regional specific target exists or not affects feasibility to improve medicine against high grade pediatric brain tumor with poor prognosis. Actually, several reports present that several types of protocadherin having highly diverse subtype counted by about 70 subtypes at least is expressed in glioblastoma and glioma(18-20). These evidences present that the part of the surface protein related to cellular connectivity is shared among several cell-types in central nerve system. This pharmaceutical framework for highly target therapy could be obviously applied to medical treatment of adult brain cancer. 
(1-2-4: Statistical and computational analysis of histological data)
Histological analysis includes observer bias especially in the cases that there are only few histological pictures for analysis, observer and analyzing person are different and the detail conditions for histological image are unknown. These cases are true when academic papers from the other institutions can be utilized for histological analysis. To overcome this challenge, machine learning approach is effective due to being capable of large statistical analysis of few million pictures which is difficult by human-powered analysis. However, to discern histological feature and map it along with specific analysis condition, confounding of the signal needs to be prevented. In the case that this observational condition is satisfied, the statistically precise analysis is possible by referring to mapping data(22,23). For example, in the case that mapping data indicates each similarity (such as UMAP representation) of multiplexed conditions such as (as category) genetic, sex, age, ethnic, brain cancer type, metabolomic, proteomic, transcriptomic, intercranial regional specific, cell-type specific traits and (as parameter of analysis), (targeting cellular nuclei or cytoplasmic or plasma membrane), concentrated area of analyzed pictures presents the specific tendency. Therefore, the observer can access the specific histological data showing clear trait, and detailing analysis by human brain is possible in a flexible and intuitive manner which is difficult by machine learning. On the other hand, singular points in mapping data are also significant not to overlook observational discovery. In our goal, (image-based and computational analysis) will be possible for three or four dimensional data such as dynamical two dimensional cellular data(3D), three dimensional cellular data(3D) and dynamical three dimensional cellular data(4D) if computational (calculation and storage) power increases and efficient algorithm are developed.

(1-2-5:Computational picturing method of tissue and mechanical view of tissue)
When selecting the arbitrary data from mapping data presented by above mentioned statistical cell-based analysis utilizing machine learning, picturing the arbitrary shaped tissue like vesicle, spheroid, sheet and tube in a visual-friendly manner is important to share visual information between researchers and as an academic article. Actually, precise picturing along with raw data remains challenging, the problem of which is clear in the picture of many academic papers including review articles.. For example, anyone and even observer are hard to understand tissue shape or cell-type distribution accurately from raw image data of several types of pediatric brain cancer. Accurate picturing could contribute to find unknown medical fact which efficiently makes innovation, given that promoting to (intuitively and concretely) understand spatial tissue information or cell-type distribution among many researchers in an interdisciplinary way is unrealizable without reaching hard-to-get (at this time) and valuably selective information. To realize this, reducing calculation cost and precise fitting is the main issues to overcome. SimuCell3D is one of global competitive methods to simulate three dimensional tissue from raw data(24). In SimuCell3D, Poisson disc sampling could contribute to reduce the calculation cost, because the suitable point is defined from raw data in the initial square grid, whereby complex cellular shape is defined. The spatial resolution and shape of this complex shape are governed by triangular network. How is each point from initial experimental data defined based on Poisson disc sampling? Poisson disc sampling set the points so as to save the minimal distance of any two point, prevent overcrowding and ensure uniformity. The pointing is iterated to make simulation data and real data fit better. The algorithm to make triangular network based on Poisson disc sampling needs to be reshaped such as edge splitting, edge merger and edge swap(supplementary fig.2 ref.24) especially under large cell deformations. Similar computational model called vertex model have been ever studied actively in many groups(25). As described by model name, vertex model can reduce calculation cost by simplifying the cellular shape (topology) to several vertex points (polygons). However, this simplistic representation of tissues come with the drawback that it cannot adequately represent cells with complex shapes. Additionally, the several vertex model that have been ever developed can be applied to several tissue types such as fluid-like tissue states, embryos, the formation of branched structures and the biased elongation of tissue, but are hard to applied to tissue analysis of intercranial region including neuron, microglia, astrocyte and oligodendrocyte because tissue “isn’t connected” by a face with large area but by elongated fine branch site such as axon (neuron), dendritic site (microglia, astrocyte and oligodendrocyte). Therefore, the definition of three-dimensional morphogenesis including neuron and glial cells remains challenging. Both conventional vertex models and SimuCell3D assumes the connected tissue like epithelial tissue which can be simply represented by the definition of the vertex points. On the other hand, these vertex models and physical deformation model could be applied to definition of three-dimensional morphology of solid-state brain tumor especially for the high-grade tumor with higher cellularity even in the intercranial region. However, this application may remain challenging compared to the normal epithelial tissue due to the complex shape of “each” cancer cell. Key points for in silico tissue analysis of solid-state tumor are how to reduce fitting point on plasma membrane of each cell meaning the degree of discreteness and to predict the appropriate “line and face” between adjacent fitting points like string. However, how much the degree of discreteness is high may relate to reduction of calculation cost, which may be trade-off with accuracy. SimuCell3D makes each cell line fit by deformable model relying on several mechanical parameter such as surface tension, membrane elasticity, pressure difference and bending elasticity. To reduce additional calculation cost, SimuCell3D restricts each cellular shape to spheroid. SimuCell3D relies on physical deformable model that difference between sampling of observational data and (independent and initial) cell-type called default dynamics (24 Table2) is based on mechanical force related to volumetric strain, membrane area change, surface curvature, in addition, and systemic potential energy converges minimized value in an equilibrium state(24 formula(1)). Three-dimensional (assignment and pointing) in each pixel depends on Poisson disc sampling so that any two adjacent sampling points can be assigned over arbitral distance, and the calculation of volume, surface area and surface curvature from real three dimensional observation data is based on triangular network. Hence, the number of edges in SimuCell3D is higher than the other vertex model based on polygons network over triangle, which allows SimuCell3D to shape complex topology. Why is the application of these computational model to brain difficult? At least, the connectivity of neuron and glial cells depends on dendritic structure, which needs to drastically change the physical model. However, neural epithelial tissue and solid-state brain tumor could be simulated by the physical deformable model. In this case, the default dynamics for initial value needs to be adjusted in line with the measurement data. After simulation about three dimensional topology on the outline form of “each cell”, contact state of cell and cellular polarization needs to be defined to determine tissue shape. Hoshen-Kopelman algorithm adopted by Simu3D is based on well-known union-finding algorithms. This algorithm has “disjoint” set data structure. In Simu3D, it is described that Hoshen-Kopelman algorithm is utilized to define unoccupied region. This may be the reason why unoccupied region includes cytoplasm surrounding occupied region (plasma membrane). There are several physical model from start to goal in general physical simulation, such as minimized energy, equilibrium state(timely unchanged state) and (conservation of mass, volume, momentum).  Data to make pointing from observational data is not complete due to existence of unexposed region like connected area, meaning that some prediction for outline form of each cell may be necessary in the simulation. If so, systemic optimization of cellular topology to convergent stage based on several physical model is somewhat needed. The definition of cellular polarization (geometric direction of connected cell) is needed especially for geometrically well-regulated epithelial tissue. To this end, side face with connected mechanism like adhesion protein and apical face needs to be defined. At initialized stage, how are occupied (with plasma membrane) voxel assigned at first? At this time, how could (geometrical data of disjoint (independent) cell, connected tissue data, and setting physical data like surface tension, bending stiffness, repulsion strength, adhesion strength and extracellular matrix properties) be utilized. After that, are these physical parameters made use of as covariant fitting parameter? The surface tension, which is contractible force, and the adhesion mechanism (cadherin), which is tensile force, is balanced(43). This balance determine is degree of stratification of cortex. In the case that the surface tension among adjacent cells is relatively low, the number of cortex layer is low and the cell polarization is maintained. On the other hand, when the surface tension is relatively high, stratification becomes high and cell polarization is somewhat lost((24) Fig.3). The molecular origin supporting the adhesion force (tensile force between cells) is tight junction like E-cadherin(43), so these molecule is one of key molecular factors that determines the tissue type. The origin of surface tension is based on the contractive force in cytoplasm, in which actomyosin plays a role in this mechanics through contacting cytoskeleton with actin (44). In other words, of course, relative force strength between attractive and tensile one governs the main mechanical system, that is, if rigid adhesion mechanism is functioned among the cells, the effectiveness of surface tension in the mechanical system becomes low. In this case, the degree of freedom for the cell shape is large, and the cellular system with high aspect ratio and layered tissue type like epithelial tissue is possible. This mechanical model is applicable to cancer. Tumor tissue with small granular shape has relatively high surface tension compared to adhesion force, and as tumor tissue is developed, adhesion mechanism relatively governs the mechanical system which determine the tissue shape. This give the important novel hint to both maintenance of healthy tissue and prevention of cancer development. As mentioned above, relatively high adhesion force compared to attractive force of each cancer cell may be needed to develop tumor, meaning that relatively high attractive force or relatively small adhesion force may prefer small granular development of tumor which is more exposed of outer space that immune mechanism and cancer drug may be well-functioned. If so, these models may lead to the novel cancer therapy based on the mechanical perspective. However, in co-existence system of normal tissue and cancer system, we need to take into account the effect of nor mal tissue during intervention of cancer system also based on this mechanical views. The promotion for granular shape may make low the rigidness of the normal tissue. Therefore, cancer-specific molecular mechanism related to attractive and tensile force needs to be clarified. In this model, not only inter- or intra-cellular force like cytoskeletons, myosin and adhesion proteins, but also extracellular matrix is necessary to be taken into account.