Page 116 - 2022_01-Haematologica-web

P. Kerbs et al.
ples. Notably, LTN1-MX1 was only found in co-occurrence with NRIP1-MIR99AHG. Further recurrence of NRIP1- MIR99AHG was reported by FusionCatcher alone in two patients’ samples (AM-0013-DX, FI-1216-RE).
Based on cDNA availability, we validated the junction of the NRIP1-MIR99AHG fusion transcript by PCR in sample AM-0028-DX. Three cytogenetically normal samples (AM- 0044-DX, AM-0054-DX, AM-0069-DX) were used as neg- ative controls (Figure 4B). Sanger sequencing of the PCR product confirmed a junction spanning sequence which matched the prediction of the RNA-sequencing fusion callers (Figure 4C). Nanopore sequencing of available gDNA from NRIP1-MIR99AHG-positive samples AM- 0028-DX (Figure 4D) and AM-0013-DX (Online Supplementary Figure S6) identified the breakpoints (Online Supplementary Table S6) and confirmed an inversion on the genomic level. With the aim of determining the complete fusion transcript, we generated a customized reference sequence of the inversion based on the identified break- points. Reads from Nanopore cDNA sequencing (median length: 883 bp) of the two NRIP1-MIR99AHG positive sam- ples were mapped to this reference. Only unique mappings were considered to obtain reads spanning the junction of the fusion. We observed high coverage of the custom refer- ence by junction-spanning reads in the two fusion-positive patients (Online Supplementary Figure S7), while there was no coverage in negative controls. NRIP1 includes a consen- sus coding sequence with an open reading frame starting in exon 4, whereas MIR99AHG is non-coding. The identified breakpoint in the NRIP1 locus in AM-0028-DX was located between exons 3 and 4, while the breakpoint in AM-0013- DX was located between exons 1 and 2, consistent with reports from RNA-sequencing fusion callers. In both cases, no annotated open reading frame was included in the puta- tive fusion transcripts. A validation in samples from the Beat AML cohort was not possible because of lack of access to the patients' material. Literature research yielded the report20 of a chronic myelomonocytic leukemia (CMML) patient with trisomy 21. The authors identified an inversion of chromosome 21 with breakpoints in the NRIP1 locus and in a region upstream of MIR125B2 (overlapping with an intronic region of MIR99AHG). We analyzed RNA- sequencing data from this patient (FI-0564-RE) with our fusion detection workflow and found high evidence for a NRIP1-MIR99AHG fusion. In total, NRIP1-MIR99AHG was found in nine (1.1%) of 806 AML patients (AMLCG, n=2; Beat AML, n=5; FIMM, n=1) and one CMML patient.
Increased expression of the 3’ partner gene in NRIP1-MIR99AHG and other fusions
In addition to the detection of fusion transcripts, we examined the expression rate of the single partner genes of a fusion and compared it between samples with and with- out this specific fusion. Sequence coverage of a gene as obtained from mapping but not read coverage of the fusion junction was considered as expression of this gene. Samples harboring a fusion, whose 3' partner gene is usually not expressed or expressed at low levels only, showed increased expression of the 3' partner gene up to the levels of the 5' partner gene, which is expressed at reasonable lev- els regardless of the fusion (Figure 5A, B). We did not observe an increase in the expression of the 3’ partner in fusion events with similar expression rates between the 5’ and 3’ partner genes (Figure 5C, D). Accordingly, MIR99AHG, which is usually not expressed or expressed at
low levels only, showed increased expression levels in NRIP1-MIR99AHG positive samples (Figure 5E). On the other hand, MX1, which is inherently fairly expressed, only showed a slight elevation of expression levels in LTN1- MX1-positive samples (Figure 5F).
Clinical and genetic characteristics of patients with NRIP1-MIR99AHG fusion
All patients found to harbor NRIP1-MIR99AHG had poor survival with a median of 296 days (range, 36-1650 days). Interestingly, most of the patients were male (6/9) and had a median age of 59 years (Online Supplementary Table S7). Karyotyping showed a complex karyotype in four patients and five patients were refractory to intensive induction therapy. Furthermore, three patients showed a gain, and one patient showed a loss of chromosome 21. Unfortunately, we have no information about whether these patients had a constitutional or somatic monosomy/trisomy 21. Cytomorphology was available for three of the nine patients without there being any evidence of megakaryoblastic leukemia (French-American-British classification, M7). Mutational status was available for six of the nine patients, but no apparent pattern was observed. However, recurrently mutated genes among those patients were NRAS (n=2) and ASXL1 (n=2) (Online Supplementary Table S7).
Discussion
The aim of this study was to test the potential of fusion gene detection by RNA-sequencing in several cohorts of AML patients’ samples and to assess its diagnostic applica- bility by comparison to current standard techniques used in clinical routine. Based on our benchmark, the vast majority of true fusions reported by routine diagnostics was also detected by RNA-sequencing, underscoring the high sensi- tivity of this method. Notably, most of the samples in which a true fusion could not be detected by RNA-sequenc- ing had a low read depth (median = 24 million mapped reads), while a minimum of 30 million mapped reads is rec- ommended by the ENCODE consortium21 for general expression analyses and even deeper sequencing for tran- script discovery (e.g., fusion transcripts). Therefore, fusion gene detection was most likely hampered by the low read depth of these samples.
Limitations of fusion gene detection by RNA-sequencing are governed by library preparation steps, read depth, expression rates of the affected genes and the applied bioin- formatic algorithms. On the other hand, Karyotyping is lim- ited to a resolution of 5-10 Mb,22 which hampers the iden- tification of small or cryptic rearrangements as well as rearrangements in specific locations (e.g., centromeric, telomeric).23 Furthermore, break-apart FISH probes identify genomic rearrangements in targeted regions through the visual separation of fluorescent labels. Although this can indicate the rearrangement of a targeted locus, the detection of a specific aberration is still limited by the resolution of microscopic inspection, and the identification of the involved partner locus requires additional assays. In con- trast to break-apart FISH, dual fusion probes target two partner loci and thereby can detect specific rearrangements but are restricted to the candidate loci of interest. In analo- gy, targeted PCR amplification of fusion transcripts requires prior knowledge of the affected genes and the correspon-
108
haematologica | 2022; 107(1)