In this study, we aimed to identify novel cytogenetic abnormalities in PMF patients by using high-density SNP-A. Previous studies showed the ability of such technique in detecting novel genetic lesions in both myeloid and lymphoid disorders 
. In particular, SNP-A profiling retains several advantages if compared to metaphase cytogenetic (MC), including the ability to detect lesions even when cells are not actively dividing, a condition favourable to the investigation of chronic myeloid neoplasms (CMN), and definitely higher sensitivity and capacity to detect aUPD. Noteworthy, the latter two features also distinguish SNP-A analysis from CGH. On the other hand, a major limitation of SNP-A, in comparison to MC, is its inability to recognize chromosomal translocations 
Indeed, with this approach we were able to detect a remarkable number of lesions (2,765 CNVs) including aUPD, deletion and amplification, many of which (992/2,765; 36%) turned out to be previously uncovered. Notably, according to our study design, all copy number changes detected by SNP-A paired analysis in the training set, were then validated in a test set. Specifically, the correspondence between the two groups was significant (p
0.0053), with 44% of lesions confirmed in both (Figure S2
Concerning the lesions previously associated with MPN, we identified del(13q), del(20q), and abnormalities on chromosome 1, +9 +8 and del(5q). Actually, we identified the presence of different regions of aUPD on chr9 at JAK2
, on chr 13q, on chr 1 and on chr 11 
In particular, we observed large regions of aUPDs in correspondence to the MPL, CBL and RB1 genes which are commonly mutated or deleted in PMF.
Importantly, as internal validation, we also recognized two lesions previously detected by MC in our cohort (namely del(5q) and trisomy 8). Thus, SPN-A karyotyping was confirmed to be an effective tool for the identification of both cryptic and larger genomic abnormalities 
. Taken together, our results confirmed the presence of a complex and heterogeneous karyotype in PMF patients, basically consistent with previous studies 
and proposed novel CNVs for further characterisation.
As far as newly identified aberrations were concerned, we focused on the lesions which recurred at higher frequency in our series, thus being candidate to retain a potential pathogenic significance. Interestingly, as indicated by other studies using less sensitive techniques 
, PMF was confirmed to have heterogeneous lesions at cytogenetic level. Actually, all patients presented with a complex karyotype, with a minimum number of CNV/patient of 188. Remarkably, however, when a paired analysis was performed in the training set, no lesion was consistently present in all the cases, not even in up to 60% of them. On the other hand, a few new lesions were found in 50% of patients. Among them, we focused on an amplification affecting the cytoband 20p13, since it proved to be the most frequent in the entire series (training + test set), the overall prevalence of such CNV being 55%. Noteworthy, if confirmed in larger series and functionally relevant, it would represent the commonest genomic lesion so far described in PMF, together with JAK2
To better understand the context of this lesion we primarily extend our attention to the adjacent areas, in the 20p13 cytoband. Interestingly, we found that all the patients except one (19/20), have at least one alteration (amplification, deletion or aUPD) occurring in the 20p13 cytoband. Subsequently, we assessed the prevalence of 20p13 abnormalities in an independent validation data set. Specifically, we studied a group of 60 myeloid tumours, basing on previously published data which were available online. Remarkably, we confirmed the presence of lesions affecting 20p13 cytoband in 7/14 MDS/MPN patients, 5/7 PMF patients and 1/39 MPN patient. Of note, this finding was then consistent with the concept that different myeloid malignancies may share common genetic abnormalities 
. In this regard, 20p13 gains were recently described in MDS 
, though occurring at very low frequency. Nevertheless, other loci on chromosome 20 have been found to be involved in myeloid malignancies, though the implicated genes and the exact consequences are poorly known 
Importantly, among all the 20p13 lesions detected in our PMF patients, we could define the CNV occurring with 55% of frequency as a minimally affected region (MAR). In particular the MAR is an amplification spanning over 9,911 bps, in correspondence to the SIRPB1 gene locus.
Subsequently, we investigated whether the copy number amplification within the SIRPB1
gene could be eventually reflected at protein level. SIRPB1 is a member of the signal-regulatory-protein (SIRP) family, belonging to the immunoglobulin superfamily. SIRPB1 is expressed in neurons and a limited subset of hematopoietc cells mainly consisting of macrophages and dendritic cells (DCs) 
. Importantly, the efficient cell-surface expression of SIRPB1 requires the association with DAP12, a dimeric adaptor ITAM- bearing protein, located in the plasma-membrane. Upon stimulation, the SIRPB1/DAP12 complex induces tyrosine phosphorylation of a number of kinases including the mitogen-activated protein kinases (MAPKs), ERK1, and ERK2. In addition, it binds the tyrosine kinase SYK, leading to the activation of the SYK-JAK-STAT intracellular cascade 
To date no involvment of SIRPB1 protein has been reported in any diseases, except for the Alzheimer's disease where its up regulation is responsible to increase the phagocytosis process 
. However, 20p13 abnormalities (amplifications, deletions or translocations) have been reported in sporadic cases of osteosarcoma, adenocarcinoma, Adult T-cell lymphoma/leukemia (HTLV-1+), and chronic myeloid leukemia according to the Cancer Chromosome database (http://www.ncbi.nlm.nih.gov/cancerchromosomes
Indeed, in our series we found SIRPB1 significantly more expressed in patients carrying the genomic abnormality rather than in the other cases. Specifically, in patients not harboring the 20p13 amplification, we found, as attended, a sporadic expression of SIRPB1 mainly restricted to dendritic cells (DCs) and macrophages. Conversely, in patients carrying the 20p13 alteration, we appreciated an intense expression of SIRPB1 in all myeloid cells, this reflecting the occurrence of the lesion in a common myeloid progenitor. Furthermore, of note, in case of 20p13 amplification SIRPB1 expression was not limited to the plasma-membrane, but was extended to the cytoplasm as well. It is conceivable that this phenomenon reflected the complete saturation of the SIRPB1 binding sites on the plasma-membrane leading to an abnormal accumulation in the cytoplasm. Bearing in mind the involvement of SIRPB1 in SYK-JAK-STAT signalling, it is possible that a deregulated SIRPB1 expression and/or function might eventually sustain proliferation and survival of affected cells. Intriguingly it may represent an alternative stimulation in JAK2wt cases, possibly having an additive/synergistic effect in JAK2mut ones. In this regard, the two lesions did not appear to be mutually exclusive, as 20p13 amplification was recorded in both JAK2wt and JAK2mut patients.
In conclusion, we detected a novel recurrent lesion involving 20p13 and SIRPB1 gene in PMF patients. Future studies are definitely warranted in order to further characterize the genomic lesion and its functional consequences, as well as to assess the possible clinical relevance of such lesion.