As most of primary NB tumors arise from the AM, investigations on the potential cell population involved in NB origin have mainly been focused on the embryonic adrenal development.
To identify which cell populations involved in AM development may be preferentially subjected to tumor transformation during the development and differentiation of sympatho-adrenal lineage, sc-transcriptomes of human developing sympatho-adrenal tissues and adrenal NB tumors were compared.
Because of the limited tumor samples analyzed by scRNA-seq, all results from these studies have been validated by tracking developmental cell-type-related signatures in bulk RNA-seq datasets (in-house or available into GEO database), to identify which signatures may be shared by patients stratified by age, stage of disease, presence of genetic alterations or prognosis. This strategy made it possible to expand RNA-seq information into broader clinical records.
These scRNA-seq studies highlighted the involvement of specific cell types in NB oncogenesis and the developmental plasticity maintained by tumor cells.
Moreover, Chromatin immunoprecipitation sequencing (ChIP-seq) studies also disclosed the fundamental role of super-enhancers (SEs) and their association with lineage-specific transcription factors (TFs) to form specific Core Regulatory Circuitries (CRCs) for each cell type that underlie MES and ADRN cell identity states.
These molecular aspects are described in detail in paragraph: Core regulatory circuitries and super-enhancers in neuroblastoma.
Sympatho-adrenal cell populations
Most primary NB tumors originating in the AM presented transcriptional similarities with the sympatho-adrenal cell populations and an epigenetic ADRN profile.
Sc-transcriptomes from early human embryos and fetal AGs and from adrenal NBs were compared by Dong et al. [
43] in a study disclosing that several NB expression meta-programs have a transcriptional profile resembling sympathoblasts or, most prevalently, chromaffin cells. By defining gene signatures for undifferentiated and differentiated chromaffin cells, the authors found that most adrenal NB tumor cells transcriptionally reflect a noradrenergic chromaffin cell type and that an undifferentiated chromaffin cell state is mainly associated with malignancy. This supported the hypothesis that heterogeneity in adrenal NB is due to different degrees of proliferation/differentiation in chromaffin cells [
43].
In a subsequent study, Hanemaaijer et al. [
72] reconstructed an atlas of the developing mouse AG at a single-cell level and identified five main cell cluster groups (medulla, cortex, endothelial, stroma, and immune clusters). The medulla group was further divided into seven clusters, including the SCP and the neuroblast clusters [
72]. Data were annotated using expression correlation to the MouseRNAseqData reference dataset (358 bulk RNA samples). The neuroblast cluster resulted the most relevant to NB comparison [
72] and identified highly cycling cells that share markers with sympathoblasts, as previously described [
42]. When compared with NB cell signatures of patients at different disease stages, present in the TARGET NB RNA-seq dataset (172 bulk RNA samples) and in the NB SEQC RNA-seq dataset (498 bulk RNA samples), the signature of the medullary SCP cluster turned out to be associated with NB phenotype severity and made it possible to group NB patients based on disease phenotype [
72]. Specifically, the SCP signature score inversely correlated with
ALK and
MYCN expression, that is associated with poor prognosis, while a high SCP signature score correlated with better overall survival rates. This study also showed that healthy neuroblasts express high levels of MycN and Alk, implying that their loci are more open and susceptible to gene amplifications or mutations [
72].
In another study, Kildisiute et al. [
77] compared sc-transcriptomes from human fetal AG and NB, revealing that NB cancer cells resemble the fetal sympathoblast state [
77]. These findings were also validated in a large number of NB bulk transcriptomes from the TARGET RNA-seq and from the SEQC datasets by integrating canonical features of the NB genome with transcriptional signals. However, the transcriptional similarity between NB cancer cells and sympathoblasts emerged as a common feature regardless of cell cluster diversity, individual patients, and clinical features [
77].
In a more recent study, Jansky et al. [
44] elucidated the developmental programs and cell types resembling NB originating in the human AM. Developing human AM samples from seven developmental time points and NB specimens from patients belonging to broad clinical NB subgroups were analyzed by droplet-based single nucleus RNA-seq (snRNA-seq) to discriminate malignant from nonmalignant cells and data were validated in NB bulk transcriptomes from the TARGET and from the SEQC RNA-seq datasets [
44]. First, cells were clustered for expression of known marker genes of infiltrating normal cell types like
PTPRC,
PTPRB and
COL1A1 and for further analysis only malignant cells harboring copy number alterations and lacking expression of normal cell type markers were kept. Differently from results obtained by Dong et al. [
43], adrenergic markers like CARTPT and INSM1 were found expressed in chromaffin cells, while the neuronal markers
NPY,
PRPH,
NTRK1 and
ISL1 were clearly expressed in neuroblasts (Table
2). To support a correct cell type annotation, the authors evaluated the expression of many other sympathetic neuronal marker genes in the neuroblast cluster, including
NEFM,
STMN2,
SYN3 and
ALK (Table
2). Moreover, they found that the well-established chromaffin marker
PNMT was expressed in a subpopulation of mature chromaffin cells [
44].
After determining the transcriptional similarity between malignant NB cells and normal developing AM cell populations, adrenal neuroblasts with high similarity scores were clearly shown to be the best match to NB cells. Then, the authors tried to match the features of normal developmental cell populations to those of the NB risk subgroups defined by genetic alterations:
MYCN-amplified high-risk tumors,
MYCN-nonamplified high-risk tumors (
TERT-rearranged or harboring telomere lengthening), and intermediate/low-risk tumors (lacking telomere maintenance) [
9,
20,
78].
MYCN-amplified NB cells turned out to be the most similar to normal early neuroblasts, with only few cells matching late neuroblasts, while the proportion of cells similar to late neuroblasts was higher in
MYCN-nonamplified high-risk tumors and among low-risk NB cells. These data suggested that low-risk tumors arise at a later time point during development or that a higher degree of dedifferentiation is induced in high-risk NBs, especially, in
MYCN-amplified tumors [
44]. The comparison between normal and NB cells at the single-cell level in the diffusion map embedding [
79] showed a continuous flow of cells across the two cell types, thereby confirming the neuroblast identity of NB tumor cells [
44]. Within the neuroblast population, low-risk tumors were comparable to more differentiated neuroblasts, while high-risk cases tumors mapped closer to the beginning of the neuroblast differentiation trajectory [
44].
Developmental trajectories, spanning the space from SCPs via bridge cells and CPCs to neuroblasts and late neuroblasts (Fig.
3), were also traced in the distinct genetic and epigenetic NB subtypes. Expression analysis also determined cell identity genes whose enhancers, via rearrangements/translocations, were found to activate oncogenes such as
MYCN,
MYC, and
TERT [
44] through a mechanism of enhancer hijacking that juxtaposes ectopic enhancers or SEs controlled by a specific CRC to the gene transcriptional unit [
80,
81]. Remarkably,
MARCH11,
NPY,
EBF1,
HAND2,
ALK,
CCND1 and
EXOC4 genes, all involved in enhancer hijacking events in NB, showed the highest expression in neuroblast trajectory cells, suggesting that the neuroblast lineage is specifically at risk of acquiring genetic alterations promoting neuroblastoma formation [
44].
To further discriminate NB cells from normal neuroblasts, a differential expression analysis of single NB cells clustered by subtype and their closest matching to normal cell types was performed [
44]. It emerged that NB cells of high-risk tumors matched early neuroblasts, while low-risk tumors showed high similarity with their normal counterparts, the late neuroblasts. In low-risk NBs, the expression of developmental genes that define the most differentiated normal neuroblast state, such as
PRPH,
SYN3,
GAP43,
NTRK1, and
SOX4, showed the highest expression. In contrast, markers of less differentiated neuroblasts like
ALK and MEIS2 were expressed at higher levels in high-risk NBs [
44].
Evaluation of AM transcription factor activities in NB subgroups showed low-risk tumors harboring the highest levels of TFAP2B, which is highly expressed in normal neuroblasts, while
MYCN-amplified cells were regulated by MYCN, showing a pronounced oncogenic MYCN signature and a reduced normal neuroblast signature. An inducible
MYCN knockdown was adopted to evaluate whether MYCN itself could suppress differentiation in
MYCN-amplified cells. Indeed, NB cells with reduced expression of
MYCN had higher expression of neuroblast- and late neuroblast-specific genes and showed repression of cell cycle-related genes, indicating that elevated MYCN causes dedifferentiation and proliferative activation. On the other hand, activation of TFAP2B, whose expression is abrogated in high-risk NBs, was able to restore differentiation signatures [
44].
These results showed that NBs are transcriptionally similar to developing adrenal neuroblasts and that the differentiation state of NBs along the normal neuroblast differentiation trajectory is associated with prognosis. Low-risk NBs closely resemble the normal committed neuroblast population and reveal the highest degree of differentiation. In contrast,
MYCN-amplified NBs and NBs with mesenchymal features (see below) were the most undifferentiated subtypes, containing tumor cells with features of early neuroblasts and bridge cells, holding the highest malignant potential [
44].
In another recent scRNA-seq study, Kameneva et al. [
45] compared developmental cell populations of human AM and other sympatho-adrenal regions with the heterogeneous cell types observed in NB. The authors carried out joint analyses of their data on sympatho-adrenal samples and of scRNA-seq data on additional NB biopsies from patients described in other studies [
43,
82].
Results revealed that most adrenergic tumor cells had clear similarity to embryonic sympathoblasts, with a fraction of tumor cells co-aligning with chromaffin cells, mesenchymal cells aligning with the embryonic mesenchymal populations, and Schwann components aligning with the SCP population [
45]. The authors also analyzed bulk RNA-seq data from a large cohort of patients from the SEQC NB dataset with different adrenal NB subtypes and gene expression signatures of embryonic SCPs, chromaffin cells, and proliferative and mature sympathoblasts to point out possible associations with patient survival. The lowest survival rate was associated with the transcriptional signature of proliferative sympathoblasts. This association held regardless of the presence or absence of
MYCN amplification in tumors and was independent of cell cycle genes. The SCP and mature sympathoblast signatures were associated with better prognosis in non-
MYCN-amplified tumors, whereas the embryonic chromaffin cell signature was associated with poor prognosis, suggesting the possible existence of immature chromaffin-like NB cell subtypes. Finally,
MYCN amplification in tumors were found to worsen the prognosis of patients with mature sympathoblast signature NB, while the SCP signature always remained correlated with positive disease outcome [
45].
It also emerged that the SCP population aligned with the Schwann component of tumor samples [
45,
72] and a fraction of tumor cells co-aligned with chromaffin cells [
45]. This is consistent with the microscopic morphology of NB cells that is composed of a population of neuroblasts at different degrees of differentiation and a variable number of normal Schwann cells [
3,
4], depending on the histopathologic features [
83,
84] of each single case.
All the above studies highlight the role of early sympathetic cells or their immediate progenitors in the initiation of NB tumor transformation, particularly in the adrenal localization, where crucial cell fate decision processes occur. The normal development of the peripheral sympathetic nervous system involves many different NC-derived precursors and committed cells that populate various final sites along the routes of migrating NCCs, where they undergo the complete differentiation process. Thus, NB may originate from any of these sites and, likewise adrenal neuroblasts, genetic and epigenetic impairments may affect extra-adrenal neuroblasts committed to the sympathetic chain promoting NB development in para-spinal ganglia. In this regard, it emerged that different genomic aberrations are involved in adrenal NBs vs NBs arising in para-spinal ganglia. A recent study compared genomic and epigenomic data from TARGET project on primary NBs [
18] originating in the AM with those arising from thoracic sympathetic ganglia. This study revealed that adrenal NBs are more likely to harbor structural DNA aberrations including MYCN amplification, whereas thoracic tumors show defects in mitotic checkpoints leading to hyperdiploidy [
85]. These findings confirm that NB tumors arising in different sites are heterogeneous entities, rather than subclasses of NB [
85].
Noteworthy, we should consider that one study [
43] came to different conclusions, showing that most adrenal NB tumor cells transcriptionally reflect a chromaffin cell type and that an undifferentiated chromaffin state is associated with malignancy.
These incongruent results are basically due to different cell type annotations. In this regard, the expression of CARTPT and other markers in human developing cells has been a subject of much debate. Based on the observation that, during mouse development, Cartpt expression is restricted to sympathoblasts [
41,
42,
46], Dong et al. [
43,
48] attributed CARTPT expression to human sympathoblasts. However, the re-analyses of datasets used by Dong in human embryonic cells [
43] and by Furlan in mouse developing cells [
42] revealed that some clusters, previously designated as human chromaffin cells, resemble mouse developing sympathoblast clusters, whereas the human sympathoblast clusters are likely to resemble to a large extent the mouse chromaffin cells [
40,
47]. Such differences underline the reason why mouse model systems cannot fully recapitulate human development and, consequently, NB oncogenesis.
The studies by Jansky et al. [
44] and by Kameneva et al. [
45] clearly detected the expression of CARTPT in a cluster of chromaffin cells in humans. However, these authors also demonstrated that certain markers of the sympatho-adrenal lineage are highly dynamic during development, in fact there were identified a CPC transient population [
44] and a second transition between sympathoblasts and chromaffin cells [
45] (Fig.
3), highlighting transient expression patterns and overlaps of the two expression programs. Therefore, it is possible that NB may arise from other cell states that assume the transcriptional state of sympathoblasts upon malignant transformation.
Dong et al. [
43] considered that chromaffin cells and sympathoblasts have distinct spatial locations and that the CARTPT+ population was previously shown to be mainly located in the extra-adrenal sites (sympathetic ganglia and SRG) [
43,
48], but not in the AM [
41,
46]. On the other hand, the recent identification of intra- and extra-adrenal sympathoblasts showing neither phenotypic nor transcriptional distinction [
45] changes the perspective.
To broaden the scRNA-seq information and to find a possible association of specific cell types with NB phenotype severity and patient outcome, sc-transcriptomes of the normal sympatho-adrenal populations were compared also with bulk NB RNA-seq transcriptomes, including a large cohort of samples from patients classified by risk subgroups.
Immature or poorly differentiated neuroblasts matched tumor cells of the high-risk group, especially
MYCN-amplified NB cells, and were associated with poor prognosis, while mature or highly differentiated neuroblasts matched NB cells of the low-risk group and
MYCN-nonamplified tumor cells and were associated with good prognosis [
44,
45,
72,
86].
The signature of the medullary SCPs and late SCPs were significantly associated with favorable prognosis [
45,
72] and, consistently, the SCP signature score was inversely correlated with
ALK and
MYCN expression, known to be associated with poor prognosis. Some authors hypothesize that this may be explained by either the presence in good prognosis tumors of healthy SCPs that confer a protective effect [
72] or by the presence in these tumors of other cell types, like Schwannian stromal cells, also originating from SCPs [
44,
72]. Conversely, embryonic chromaffin cells aligned with tumors with unfavorable outcomes, which is consistent with the existence of immature chromaffin-like NB cell subtypes. Notably,
MYCN amplification was found to worsen the prognosis of NB patients with mature sympathoblast signature, while the SCP signature always remained correlated with positive outcome [
45].
Taken together, these findings show that NB can derive from neuroblasts at different stages of differentiation, and that a low degree of differentiation is associated with high-risk tumors, while a higher degree of differentiation is associated with low-risk tumor. Alternatively, a higher degree of de-differentiation may be induced in high-risk NBs, especially in
MYCN-amplified cases [
44].
Mesenchymal cell populations
Most adrenal NB tumors analyzed in the study by Jansky et al. [
44] were classified as ADRN type, yet the malignant cells of three high-risk tumors showed increased MES signature expression and reduced ADRN signature expression and were termed high-risk NBs with MES features, as previously defined [
32].
High-risk NBs with MES features showed similarity to a broader set of normal cell populations including bridge, connecting progenitor and chromaffin cells, in addition to neuroblasts, and were mapped close to the branching point of the chromaffin and neuroblast lineages, within the newly defined CPC population.
ERBB4 and
VGF, which are characteristic of bridge/connecting progenitor cells, were exclusively expressed in high-risk NBs with MES features. In contrast to low-risk NBs that closely resembled the normal committed neuroblast population with the highest degree of differentiation,
MYCN-amplified NBs and NBs with mesenchymal features were the most undifferentiated subtypes, containing tumor cells with features of early neuroblasts and bridge cells that remarkably expand the potential pools of malignant cells [
44].
Subsequently, Gartlgruber et al. [
87] identified four super-enhancers (SEs) epigenetic subtypes and regulatory cell identities in NB. In addition to three ADRN signatures, 60 NB tumors and 25 cell lines employed in this study were shown to form a common cluster with MES signature. However, differently from NB cell lines, which are clearly distinct in two groups with very high or low MES scores, tumors showed a continuous dependency on MES signature, which is suggestive of cellular heterogeneity that may mask malignant MES clones in NB tumors that are instead enriched during cell culture. This work also disclosed that the MES subtype shares cellular identity with multipotent SCPs and can experimentally be triggered by oncogenic
RAS activation, which indicates that specific mutations can revert neuronal differentiation programs and induce stem cell-like features. Furthermore, mapping of the in-house bulk NB RNA-seq data [
44] from these epigenetic subtypes onto the sc-transcriptomes of mouse developing AGs [
42] revealed phenotypic similarities with distinct cell populations at different stages of differentiation, from SCPs to early neuroblast/chromaffin cells, demonstrating a considerable NB cellular heterogeneity [
87]. These findings are in line with the observation by Kameneva et al. [
45] of NBs with highly heterogeneous composition, including malignant cells of both adrenergic and mesenchymal lineages. Among the identified epigenetic subtypes of NB, the MES subtype distinctly overlapped SCPs, suggesting a high degree of phenotypic and molecular similarities [
87].
In summary, these studies highlight that, besides tumors with clear ADRN signatures, a cluster of adrenergic NBs with increased MES features aligned with embryonic mesenchymal populations and included cells with traits of SCPs, bridge cells, and connecting progenitors in addition to early neuroblasts and immature chromaffin cells [
44,
45,
87]. These results disclosed that adrenal NBs with MES features contain a broad spectrum of undifferentiated cell subtypes that may undergo impairments and functional perturbations that eventually lead to malignant transformation at early stages of development, when sympatho-adrenal and mesenchymal lineages bifurcate from bipotent progenitors (Fig.
3, 2
nd split).
The possible contribution of mesodermal lineage derivatives in the formation of NB was another aspect investigated [
45]. The adrenal cortex originates from an early mesenchymal progenitor, sharing a transcriptional signature with early kidney progenitor cells, and its development proceeds in tight coordination with NC-derived cells. For this reason, this study has compared sc-gene expression signatures of these developing cells with NB transcription signatures from patients with different survival prognosis [
45].
The presence of embryonic adrenocortical and kidney signatures in tumors were associated with poor outcome in non-
MYCN-amplified NB cases, while no outcome difference was detected in
MYCN-amplified NBs. This suggested a role of local microenvironment in the modulation of non-
MYCN-amplified tumors by mesenchymal, cortical, endothelial and kidney cell types, whereas the microenvironment dependencies of tumors with
MYCN amplification resulted distinct [
45].
Very recent insights emerged from the latest sn-transcriptome analysis that compared NB tumor cells from patients classified in different clinical risk groups and stages with healthy cells from postnatal human and mouse AGs [
74]. The analysis was extended to previously published sc-sequencing datasets from other NB tumors from the TARGET and from the SEQC NB RNA-seq datasets [
77] and from mouse [
42] and human [
43] embryonic sympatho-adrenal derivatives to provide a better understanding of the transcriptional basis of the clinical heterogeneity in NB [
74]. The authors identified a new cluster of TRKB+ cholinergic progenitor cells unique in human postnatal AGs sharing a specific gene signature with a cluster of undifferentiated cells of mesenchymal nature, including biomarkers of migratory/progenitor cell states and a significantly high expression of the
NTRK2 gene, encoding TRKB, which is enriched in high-risk NBs [
74]. Moreover, the gene signature of these mesenchymal cells matched to lower patient survival probability and older age-at-diagnosis when assessed in a large cohort of NB patients from NB RNA-seq datasets. Conversely, the transcriptional profile of low-risk NB showed a noradrenergic signature matching the profile of postnatal chromaffin cells as well as of embryonic sympathoblast and chromaffin populations. These results suggest that NBs could also originate from alterations occurring during postnatal development in these cholinergic progenitors. Finally, analyses of cell populations revealed different gene expression programs for worse and better survival in correlation with age at diagnosis, highlighting that the cellular identities and the composition of human NB tumors reflect clinical heterogeneity and outcome [
74].