Towards a single-assay approach: a combined DNA/RNA sequencing panel eliminates diagnostic redundancy and detects clinically-relevant fusions in neuropathology

Cohort demographics

The case cohort comprised 233 surgical resections from 231 patients for which the Oncomine panel was completed. Of the 231 patients, 122 were male and 109 were female, and 27 cases were from pediatric patients (age at diagnosis < 21 years). Of 157 cases diagnosed as infiltrating gliomas presenting in adult patients, 14 were oligodendroglioma, 29 were IDH-mutated infiltrating astrocytoma, and 112 were IDH-wildtype infiltrating astrocytomas. Within the latter category, these 112 cases included the following subsets: 99 cases of glioblastoma, IDH-wildtype with histological features of GBM; 9 cases with lower-grade histological features but meeting molecular criteria for GBM, IDH-wildtype; 3 cases of pediatric-type diffuse hemispheric glioma, H3G34-mutant, and 1 case of IDH-wildtype astrocytoma with grade 3 histological features, NOS (i.e. not meeting molecular criteria for GBM). Tremaining 76 cases comprisedentities other than infiltrating gliomas arising in adults (Table 1).

Detection of IDH and TP53 by immunohistochemistry compared to oncomineIDH

Immunohistochemistry for IDH1 R132H IHC stain was performed on 191 of the 233 cohort cases. Of these 191 cases, 152 were negative for IDH1 R132H mutation and 39 were positive. All cases positive for IDH R132H by IHC also had this mutation detected by Oncomine while 5/152 cases which were negative for the R132H mutation by IHC had an alternative IDH1/2 mutation detected by the Oncomine panel (Fig. 1a, b). Of the additional 5 IDH mutations detected by Oncomine but not IHC, none were IDH1 R132H. These included an IDH2 R172K alteration detected in an oligodendroglioma, as well as IDH1 R132G (1 case) and IDH1 R132S (2 cases) mutations, all three of which were seen in IDH-mutant infiltrating astrocytomas (Fig. 1a, b). Finally, a single case revealed an IDH1 I117T alteration by Oncomine. Since this alteration does not have a known pathogenic association and other molecular alterations in this tumor were characteristic of an IDH-wildtype infiltrating astrocytoma, the mutation was regarded as clinically insignificant and not diagnostic of the IDH-mutated class of tumors.

Fig. 1figure 1

Comparison of IDH and TP53 mutations detected by immunohistochemistry (IHC) and targeted next-generation sequencing (NGS) using the Oncomine Panel. A Detection of IDH mutation in infiltrating gliomas by IHC alone, NGS alone, or both, (B) codon change of IDH mutations detected by the Oncomine targeted NGS panel, (C) concordance of TP53 IHC score with TP53 mutation detection on NGS, (D) detection of TP53 mutation by IHC, NGS, both, or neither by class of infiltrating glioma and all remaining diagnoses. (IA_IDH_WT = IDH-wildtype infiltrating astrocytoma in adults; IA_IDH_MUTANT = IDH-mutant infiltrating astrocytoma; OLIGO = oligodendroglioma; NON_IG = non-infiltrating glioma)

Altogether these results yield a 90.7% sensitivity and 100% specificity for IDH1-R132H immunohistochemical staining relative to the detection of IDH1/IDH2 alterations overall as determined by Oncomine sequencing. Importantly, our data in this series is consistent with a sensitivity and specificity of 100% for the Oncomine panel in detecting pathogenic IDH alterations. No case that was negative for IDH alterations by Oncomine showed positive IDH staining, and moreover none of these cases demonstrated evidence of IDH alterations by concurrent NGS testing by alternative platforms (e.g. Foundation Medicine in a subset of cases) or displayed clinical, histological or other molecular features that would raise suspicion of a false-negative sequencing result (data not shown). In this cohort 9.3% of the total pathogenic IDH1/2 mutations (4/43) were missed by IHC, as expected given its specificity for IDH1 R132H.

TP53

IHC staining for p53 was conducted for 194 of the 233 cases. A score of 0–3 was assigned to each case (as described in Methods). Of the 194 p53 IHC stains conducted, the distribution of scores is shown in Fig. 1c, along with the associated TP53 alterations called by Oncomine.

Two of six (33%) cases that were given a score of 0, indicating that a truncating mutation was considered, had TP53 mutations detected by Oncomine. One case indeed harbored a frameshift truncating mutation and the other case harbored 2 distinct missense mutations. Of the cases given a score of 1, 22/125 (18%) had at least one TP53 alteration detected and 5 cases had more than 1 TP53 mutation; in total there were 17 missense mutations, 6 frameshift deletions, 4 splice site alterations, and 2 nonsense mutations. Out of the 30 cases given a p53 staining score of 2, 13 cases had TP53 mutations all of which were missense, 4 of these cases had 2 missense mutated detected. All of the cases scored 3 (33/33; 100%) had either a missense mutation detected (32 cases) or a non-frameshift deletion (N131del; 1 case). Five of these cases had compound heterozygous mutations, with concurrent nonsense seen in 4 cases and a frameshift mutation detected in 1 case.

TP53 mutation was suspected if the IHC was scored as 0 (completely absent staining suggestive of a truncating mutation) or 3 (strong labeling consistent with a missense mutation). When cases with a score of 0 or 3 are considered in aggregate, 35/39 cases ultimately did have TP53 mutations detected by the Oncomine panel, resulting in 89.7% positive predictive value for mutation by IHC. A score of 1 was given if the pattern of labeling was considered most consistent with wildtype TP53 mutation. 103/125 indeed lacked a TP53 mutation as detected by Oncomine, yielding a positive predictive value of 82% for wildtype TP53 given a score of 1. The cases scored as 1 that did ultimately reveal TP53 mutations often harbored either non-missense mutations or harbored compound heterozygous mutations as stated above, potentially accounting for the discrepancy in a majority of these cases. Given that a score of 2 represents an ambiguous staining pattern, as expected roughly half of the cases with a score of 2 (13/30 or 43%) harbored a TP53 mutation and the rest did not. In summary, if the number of cases scored as 2 (ambiguous staining) are added to those with scores 0,1 and 3 that showed discrepant sequencing results, we obtain a total of 56/194 (29%) cases for which p53 IHC failed to predict the sequencing result. The stain was most reliable when strong labeling was present in a majority of tumor cells, suggesting a missense mutation, and was considerably less reliable in predicting TP53 mutations in the context of the other staining patterns.

Prediction of ATRX mutational status based on IDH and TP53 mutations detected by oncomine

The Oncomine panel does not directly assess for alterations in ATRX, the loss of which is closely associated with the presence of concurrent IDH and TP53 mutations in the majority of infiltrating gliomas arising in adults. It has been reported that IDH-mutant astrocytomas have co-occurring TP53 mutations in 94% of cases and loss of ATRX expression in 86% of cases [10].We sought to assess the added value of performing ATRX immunohistochemistry over and above an ATRX prediction metric based upon the status of IDH and TP53 alterations as determined by Oncomine alone in the adult population within our cohort. The ATRX prediction status was compared to results of ATRX IHC as well as Foundation Medicine targeted next-generation sequencing panel results (which directly assesses ATRX) when available (78 cases).

Of the 29 IDH-mutant gliomas without 1p/19q codeletion 28/29 had a TP53 mutation detected. It was predicted that all 28 of the IDH-mutant infiltrating astrocytomas would have loss of ATRX. ATRX IHC for 7 of the 28 cases demonstrated ambiguous staining with variability of labeling across tumor cells and internal control non-neoplastic cells, and were considered uninterpretable; these cases were excluded from further analysis. Out of the 22 IDH-mutated infiltrating astrocytoma cases with interpretable ATRX IHC, 12 showed evidence for loss of ATRX expression in neoplastic cells. Thus, if one were to predict the presence of an ATRX alteration on the basis of Oncomine-detected IDH/TP53 double mutation, this would correlate with immunohistochemically detected ATRX loss of expression in only 55% of cases (Table 3) by our laboratory. Interestingly, when compared to an orthogonal NGS panel (Foundation Medicine) that assessed for ATRX mutation, all 12 of the IDH1/TP53 double mutant cases for which Foundation sequencing was available indeed demonstrated an ATRX mutation. Nine of these alterations were considered known pathogenic alterations and 3 were considered variants of unknown significance (Table 3).

Table 3 ATRX status prediction using presence and absence of TP53 and IDH mutations compared to ARTX immunohistochemistry and the FoundationOne targeted next-generation sequencing panel

All 14 cases of oligodendroglioma had an IDH mutation detected on Oncomine, but only 2 had TP53 mutations detected. Given the low incidence of ATRX mutations in oligodendrogliomas, none of these cases were predicted to have loss of ATRX on IHC or Foundation Medicine. IHC for ATRX was available for 13 of the oligodendroglioma cases and all showed preservation of ATRX while only 3 cases (including the one for which IHC was not available) had Foundation Medicine results, none of which showed mutations in ATRX. Therefore, prediction accuracy for oligodendroglioma based on diagnosis and alteration status for IDH status and TP53 as determined by Oncomine was 100% compared to both IHC and Foundation Medicine results (Table 3).

Of the 112 IDH-wild type infiltrating astrocytoma cases arising in adults, 38 cases did have TP53 alterations detected by Oncomine. Given the low incidence of ATRX mutations in IDH-wildtype diffuse astrocytomas arising in adults, it was predicted that none of these cases would have ATRX loss by IHC or Foundation Medicine. ATRX IHC was available for 105 of the cases, 3 of which were inconclusive and therefore excluded from further analysis. The 102 cases for which ATRX IHC was available and interpretable, all 102 demonstrated ATRX preservation making the predicted ATRX status relative to ATRX IHC 100% (Table 2). Foundation Medicine results were available for 63 of the 112 cases through which 6 cases were found to have ATRX alterations, 3 were known pathogenic alterations and 3 were classified as VUS. Therefore, the predicted ATRX status was concordant with sequencing in 57/63 (90.5%) cases as compared to Foundation Medicine results (Table 3). Importantly, three of the cases that were discordant between an ATRX prediction metric based upon IDH-wildtype status as determined by Oncomine in adults, and Foundation sequencing, which directly assesses ATRX mutations, included 3 cases of pediatric-type hemispheric glioma, H3G34-mutant, all of which presented in young adults in their third decade of life. Of the two cases for which ATRX IHC was performed among these three, both did not show loss of ATRX by IHC. As discussed below, while the vast majority of IDH-wildtype gliomas in adults do not have ATRX mutations, evidence of ATRX loss, either by IHC or by sequencing panels that assess this gene should prompt consideration of pediatric-type diffuse gliomas, including H3G34-mutant tumors, and other entities including ‘high grade astrocytoma with piloid features’ (HGAP) [7].

In total, when compared to other targeted-NGS panels that include sequencing of ATRX, the accuracy of ARTX status prediction based on the presence or absence of IDH/TP53 mutations alone is 72/78 (92.3%), indicating the redundancy of this data point in the vast majority of adult cases. Relative to IHC, the presence of IDH1/TP53 double mutation correlated with ATRX loss of expression in 127/137 (93%). The independent clinical utility of assessing for ATRX status by immunohistochemistry in adults is unclear, especially considering that sequencing and/or methylation profiling would typically be required to confirm less common tumor diagnoses in the IDH-wt setting that may harbor ARTX alterations. Moreover, it is unclear the extent to which ATRX expression as measured by IHC at the protein level may be reduced independently from DNA-detectable sequencing alterations of the ATRX gene itself, or on the other hand if loss-of-function mutations may occur even when antigenicity relative to commonly used antibodies in clinical practice is preserved in the translated product.

Detection of chromosome and gene copy number alterations using oncomineEGFR amplification detection concordance by FISH and oncomine

EGFR amplification was detected by FISH in 39/151 cases for which this assay was performed (25.8%). Oncomine did not detect EGFR amplification in any of the cases with negative FISH results. Of those cases that were called positive for amplification of EGFR by FISH, 35/39 (89.7%) also demonstrated EGFR amplification by Oncomine. The 4 discrepant cases were IDH-wildtype infiltrating astrocytomas which by FISH all had average EGFR signals per nucleus > 4 (range 4.48–6.62) and ratio of EGFR signals/CEP7 signals (a centromeric probe) > 2 (range 2.1 -2.92). Thus, in these ‘discrepant’ cases, tumor cells harbored a relatively low degree of amplification relative to classic cases of EGFR-amplified IDH-wildtype astrocytoma that often harbor 10's to 100's of copies of the gene, often episomally. Interestingly, the copy number of EGFR as inferred by Oncomine analysis for these cases ranged from 2.74 to 4.6 with the copy number ratio of EGFR relative to the average copy number for the remaining genes sequenced by Oncomine on chromosome 7 was between 1.35 and 1.02, more indicative of the increased EGFR copy number being a result of broad chromosomal 7 gain, also a common feature of IDH-wildtype astrocytoma. The biological and prognostic significance of copy number gains and low-level amplification versus high level amplification, and the exact definitional thresholds that should be used as diagnostic criteria (i.e. for gain versus amplification), are not well-defined in the literature and require future studies to further refine. Guidelines published in the cIMPACT-NOW update 3 states that EGFR amplification qualifying for Grade 4 designation of IDH-wildtype astrocytoma in the absence of high-grade histologic features should only be called in the presence of “high-level copy number gains” as established by “clinically validated assays” [9].

1p/19q

The average copy number for genes tested by the Oncomine panel located on chromosome 1p (MTOR, MYCL, MPL, MAGOH, JAK1, and NRAS) was significantly lower in cases of oligodendroglioma (average CN = 1.15 ± 0.14) compared to IDH-mutant infiltrating astrocytoma (average CN = 2.04 ± 0.37; p = 3.65 × 10–13) and IDH-wild type infiltrating astrocytoma (average CN = 1.97 ± 0.16; p = 1.76 × 10–12) but no significant difference was detected between IDH-mutant and IDH-wildtype infiltrating astrocytoma (p > 0.05; Fig. 2a). There was no difference in the average copy number for genes tested on chromosome 1q (BCL9, MCL1, DDR2, and MDM4) between the infiltrating glioma subgroups (p = 0.44; Fig. 2a).

Fig. 2figure 2

Use of average copy number of genes sequenced by the Oncomine panel located on chromosomes 1p, 1q, 19p, 19q in order to detect 1p/19q co-deletion. A Average copy number for genes sequenced on chromosomes 1p, 1q, 19p, and 19q for IDH-mutant infiltrating astrocytoma (IA_IDH_MUTANT), IDH- wild type infiltrating astrocytoma in adults (IA_IDH_WT), and oligodendroglioma (OLIGO). B Map of chromosomes 1 and 19 showing distribution of genes sequenced by the Oncomine panel. C Average copy number for genes on 1p/19q (left panel) and 1q/19p (right panel) for each subgroup of infiltrating glioma where the red line represents suggested cut-off value of 1.65 for average copy number of genes sequenced on 1p and 19q to detected co-deletion. D ROC curve for determination of cut-off value with highest sensitivity (100%) and specificity (98%)

Unexpectedly, a significant increase in the average copy number of genes tested by Oncomine located on 19p (STK11, GNA11, MAP2K2, and JAK3) was detected in IDH-wildtype infiltrating astrocytoma (average CN = 2.44 ± 0.39) compared to IDH-mutant infiltrating astrocytoma (average CN = 2.22 ± 0.29; p = 0.002; Fig. 2a) and oligodendroglioma (average CN = 2.22 ± 0.13; p = 9.96 × 10–5; Fig. 2a). A significant decrease in average copy number for genes tested on 19q (CCNE1 and PPP2R1A) was noted in oligodendroglioma (average CN = 1.23 ± 0.18) compared to IDH-mutant infiltrating astrocytoma (average CN = 1.96 ± 0.29; p = 1.29 × 10–11; Fig. 2a) and IDH-wild type infiltrating astrocytoma (average CN = 2.24 ± 0.38; p = 3.51 × 10–16; Fig. 2a). Interestingly a significant increase in average copy number of 19q was also seen in IDH-wildtype compared to IDH-mutant infiltrating astrocytoma (p = 0.0001; Fig. 2a).

In order to determine a cut-off value for average copy number that would be indicative of 1p/19q co-deletion, a receiver operating characteristic (ROC) curve was created by combining the average copy number for all genes sequenced across chromosome 1p and 19q. A cut-off value of 1.65 for average copy number resulted in a sensitivity of 100% and specificity of 98% for detection of 1p/19q co-deletion relative to FISH (Fig. 2c, d). When looking at the average copy number for genes sequenced on chromosome 1p and 19q for each case, there is a distinct difference in the distribution of average copy number for cases of oligodendroglioma versus IDH-mutant and –wild type infiltrating astrocytoma that corresponds to the cut-off at a value of 1.65 (Fig. 2c, d).

CDKN2A copy number alterations

The Oncomine panel detected 86 alterations in the CDKN2A gene over all cases. The majority of these alterations were isolated copy number losses in CDKN2A (77/86; 89.5%) along with 2 cases where evidence for a putative deletion more broadly over the 9p chromosomal arm was detected (2/86; 2.3%). Three cases had a nonsense variant detected (3.5%) while 2 had a missense mutation detected (2.3%). Additionally, a frame shift deletion was detected in one case and a splice site variation in another (1.2% each). The distribution of tumor classes for each alteration detected in CDKN2A by the Oncomine panel can be seen in Fig. 3a and b.

Fig. 3figure 3

CDKN2A alterations detected on the Oncomine targeted NGS panel. A Distribution tumor classes in which each type of alterations detected in CDKN2A was found (DMG_H3K27M = Diffuse midline glioma, H3K27M mutant; GG = Ganglioglioma; IA_IDH_MUTANT = IDH-mutant infiltrating astrocytoma; IA_IDH_WT = IDH-wildtype infiltrating astrocytoma in adult patients; IA_IDH_WT_PEDIATRIC = Infiltrating astrocytomas in pediatric patients, IDH-wildtype; MALIGNANT_NEOPLASM_NOS = malignant neoplasm, not otherwise specified; MENINGIOMA = meningioma; NET_NOS = Neuroepithelial tumor, not otherwise specified; OLIGO = Oligodendroglioma). B Types of CDKN2A alterations found in each subgroup of infiltrating glioma. C Spread of copy number of CDKN2A by tumor class where each point represents a case which is colored by the CDKN2A alteration class called by the Oncomine panel. D distribution of copy number of CDKN2A for subgroups of infiltrating glioma

Out of the 219 cases that the copy number of CDKN2A was recorded, loss of CDKN2A was detected in 77 (35.2%). Of these, the vast majority were IDH-wildtype infiltrating astrocytoma (67/77; 87.0%) with IDH mutant infiltrating astrocytoma being the second most common tumor type for which CDKN2A loss was detected by targeted NGS (5/77; 6.5%). The remaining 5 cases in which loss of CDKN2A was detected comprised one ganglioglioma, one pediatric-type IDH-wildtype infiltrating astrocytoma, one meningioma, one neuroepithelial tumor not otherwise specified (NOS), and one ‘malignant neoplasm, NOS’ (Fig. 3a, b).

Of 106 cases of IDH-wildtype infiltrating astrocytoma with copy number data for CDKN2A, loss was seen in 67 cases (63.2%). Only 5 of 27 (19%) cases of IDH-mutant cases had loss of CDKN2A detected and 0 of 13 cases of oligodendroglioma had a loss in CDKN2A, although one case demonstrated a frame-shift deletion alteration (Fig. 3b, c).

The average copy number for CDKN2A was 1.09 ± 0.75 in IDH-wildtype infiltrating astrocytoma, 1.73 ± 0.68 in IDH-mutant infiltrating astrocytoma, and 1.94 ± 0.41 in oligodendroglioma. When comparing these values using paired t-tests, the average copy number for IDH-wildtype infiltrating astrocytoma was lower than in IDH-mutant infiltrating astrocytoma (p = 1.02 × 10–4) and oligodendroglioma (p = 1.85 × 10–6). No significant difference was found in the copy numbers between IDH-mutant infiltrating astrocytoma and oligodendroglioma overall (p = 0.24) (Fig. 3d).

Added-value of RNA-seq component of oncomine

The RNA-Seq portion of the Oncomine panel detected 42 RNA-based alterations from the broader cohort of 233 CNS tumor cases, including 157 infiltrating gliomas (Fig. 4a). The RNA-Seq portion of the Oncomine panel failed in 21/233 cases (9.01%) due to either poor RNA quality or low quantity. All detected fusions were confirmed by orthogonal RT-PCR analysis using site-specific primers (data not shown). The alterations detected by the panel included EGFRvIII, (EGFR transcripts with a deletion of exons 2–7) that is characteristic of a subset of IDH-wildtype GBM, and indeed all 19 cases with this transcript detected were IDH-wildtype GBM (Fig. 4a). Interestingly, several additional fusion transcripts were detected by the Oncomine panel, a generic solid tumor panel that was not designed specifically for tumors of the CNS. Some of the detected fusions have been previously reported in infiltrating gliomas, including MET-PTPRZ1, which was detected in 3 cases (2 IDH-mutant and 1 IDH-wildtype case), FGFR1-TACC1 (1 IDH-wildtype case), and FGFR3-TACC3 in 8 cases (6 IDH-wildtype infiltrating astrocytomas, 1 neuroendocrine tumor NOS, and one malignant neoplasm, NOS) (Fig. 4a). Additional less frequently reported alterations were detected in infiltrating gliomas including an NTRK2-ETV6 fusion (1 IDH-mutant astrocytoma) and ROS1-GOPC fusions (2 in IDH-wildtype infiltrating astrocytomas).

Fig. 4figure 4

RNA based alterations detected in the RNA-Seq portion of the Oncomine targeted next-generation sequencing panel. A All detected fusions with column coloration by diagnoses for which each alteration as found (IA_IDH_MUTANT = IDH-mutant infiltrating astrocytoma; IA_IDH_WT = IDH-wildtype infiltrating astrocytoma in adults; MALIGNANT_NEOPLASM_NOS = malignant neoplasm, not otherwise specified; NET_NOS = neuroepithelial tumor, not otherwise specified; PILOCYTIC = pilocytic astrocytoma). B Sanger sequencing confirming RET-PCM1 fusion and chromosomal structure of the fusion

Several of the fusions detected that, to our knowledge, have not been previously reported in CNS tumors included PDGFRA-SCAF11 as well as 2 RET fusions, RET-CCDC6 and RET-PCM1, all of which occurred in cases of IDH-wildtype infiltrating astrocytoma (Fig. 4a, b). Of particular interest is that in both cases with RET fusions, these patients had additional germline alterations in BRCA genes: BRCA2 S1982fs in the RET-PCM1 fused case and BRCA1 splice site 442-1G > T in the RET-CCDC6 fused case. Notably, the RET-PCM1 fusion was detected in a surgically resected recurrent tumor that had followed prior surgery, radiation therapy, and temozolomide treatment. The patient subsequently was treated with one cycle of CCNU through which her tumor progressed. The option of enrolling in a clinical trial for targeted treatment of RET fusion was declined by the patient. The elderly patient harboring the RET-CCDC6 fusion also did not receive additional treatment based on the sequencing results given several clinical factors that led to a decision to forgo additional treatment.

In 5/11 cases of pilocytic astrocytoma a BRAF-KIAA1549 fusion was detected by the panel. One case of pilocytic astrocytoma had a BRAF-KIAA1549 fusion detected by Foundation Medicine in addition to FISH demonstrating tandem duplication of the BRAF locus, but not by the Oncomine panel. This discrepancy with Foundation medicine may be a result of differences in primer design for amplifying transcript reads pertaining only to certain predetermined fusion sites. Of the remaining cases of pilocytic astrocytoma for which BRAF-KIAA1549 fusion was not found, additional drivers such as FGFR4 missense and NF1 truncating mutations were found (1 case each). In only one case of pilocytic astrocytoma were no variants detected by Oncomine.

留言 (0)

沒有登入
gif