Characterization of R132H Mutation-specific IDH1 Antibody Binding in Brain Tumors

Authors


Andreas von Deimling, MD, Department of Neuropathology, Institute of Pathology, Ruprecht-Karls-Universität Heidelberg, Im Neuenheimer Feld 220/221, D-69120 Heidelberg, Germany (E-mail: andreas.vondeimling@med.uni-heidelberg.de)

Abstract

Heterozygous point mutations of isocitrate dehydrogenase (IDH)1 codon 132 are frequent in grade II and III gliomas. Recently, we reported an antibody specific for the IDH1R132H mutation. Here we investigate the capability of this antibody to differentiate wild type and mutated IDH1 protein in central nervous system (CNS) tumors by Western blot and immunohistochemistry. Results of protein analysis are correlated to sequencing data. In Western blot, anti-IDH1R132H mouse monoclonal antibody mIDH1R132H detected a specific band only in mutated tumors. Immunohistochemistry of 345 primary brain tumors demonstrated a strong cytoplasmic and weaker nuclear staining in 122 cases. Correlation with direct sequencing of 186 cases resulted in consensus of 177 cases. Genetic retesting of cases with conflicting findings resulted in a match of 186/186 cases, with all discrepancies resolving in favor of immunohistochemistry. Intriguing is the ability of mIDH1R132H to detect single infiltrating tumor cells. The very high frequency and the distribution of this mutation among specific brain tumor entities allow the highly sensitive and specific discrimination of various tumors by immunohistochemistry, such as anaplastic astrocytoma from primary glioblastoma or diffuse astrocytoma World Health Organization (WHO) grade II from pilocytic astrocytoma or ependymoma. Noteworthy is the discrimination of the infiltrating edge of tumors with IDH1 mutation from reactive gliosis.

INTRODUCTION

Recently, extraordinary high rates of spontaneous mutations in the gene encoding cytosolic nicotinamide adenine dinucleotide phosphate (NADP+) dependent isocitrate dehydrogenase (IDH)1 have been reported in diffuse gliomas World Health Organization (WHO) grades II and III of astrocytic and oligodendroglial lineages as well as in secondary glioblastoma WHO grade IV (sGBM) (1, 6, 18, 21). In contrast, mutations of IDH1 in primary glioblastoma WHO grade IV (pGBM) are rare (1, 2, 6, 15, 18, 21). Indeed, IDH1 mutations have been proposed as a marker to distinguish primary from secondary GBM (14). Mutations in other tumor types are rare (2, 7); however, a recent report described IDH1 mutations in acute myeloid leukemia (11). IDH1 mutations are associated with younger age at diagnosis in diffuse astrocytoma WHO grade II (A II), anaplastic astrocytoma WHO grade III (A III), anaplastic oligoastrocytoma WHO grade III (OA III) and GBM (1, 5, 14, 15, 18, 21). The high frequency and association with both TP53 mutation and 1p19q loss of heterozygosity delineate IDH1 mutation as an early event in diffuse glioma formation, which may affect a common glial precursor cell population of oligodendroglial and astrocytic tumors (1, 16, 18).

IDH1 on chromosomal arm 2q33.3 encodes cytosolic NADP+ specific IDH (13) and catalyzes the cytosolic oxidative decarboxylation of isocitrate to alpha-ketoglutarate, resulting in the production of reduced form of NADP+ (NADPH). Glioma-specific mutations in IDH1 are heterozygous and of somatic origin, and always affected the amino acid arginine at position 132 of the amino acid sequence (1, 6, 15, 18, 21), which belongs to an evolutionary highly conserved region locating to the binding site of isocitrate (15, 22). The side chain of R132 uniquely forms three hydrogen bonds with both the alpha and beta-carboxyl groups of isocitrate (22). IDH1 functions as a homodimer with two enzymatically active sites, and IDH1 mutations have been shown to dominantly inhibit IDH1 catalytic activity (22). In a series of 1010 grade II and III gliomas, 93% of IDH1 mutations were characterized by a base-pair exchange of guanine to adenine (G395A), resulting in a substitution of amino acid arginine by histidine (R132H) (5). In oligodendroglial tumors, the distribution of IDH1 mutations is even more in favor of the R132H type, reaching frequencies of over 95% of total IDH1 mutations (5). Several recent publications report on prognostic implications of IDH1 mutations, with mutated tumors generally showing a better prognosis (14, 15, 19–21).

The current routine procedure for assessing IDH1 status is DNA sequencing. This technique requires extraction of DNA and elaborate laboratory equipment not available in every neuropathology setting. An alternate method not requiring sequencing but relying on polymerase chain reaction (PCR) and restriction endonucleases recognizing a site composed of mismatched primer and IDH1 mutation has recently been introduced (12). However, this approach still depends on extraction of nucleic acids that is both time-consuming and laborious. Therefore, we designed a mutation-specific IDH1 antibody for the most frequent mutation of the R132H type (3). We here demonstrate that R132H-mutated IDH1 is translated into protein in vivo and can be specifically identified using mIDH1R132H antibody both in Western blot and immunohistochemistry of human brain tumor specimens.

MATERIALS AND METHODS

Antigen peptide, immunization and hybridization

The generation of mouse anti-IDH1R132H (mIDH1R132H) and anti-pan-IDH1 rat (rIDH1) monoclonal antibodies has recently been described (3). In brief, 13 amino acid synthetic peptide of the sequence CKPIIIGHHAYGD (Peptide Specialty Laboratories GmbH, Heidelberg, Germany) representing the IDH1 amino acid sequence from codon 125 to 137 containing the R132H mutation was coupled to keyhole limpet hemocyanin. Seven C57BL/6 mice were immunized with 20 µg of the coupled peptide and boosted after 1 week and 11 weeks. Polyethylene glycol fusion of lymph node cells with mouse myeloma SP2/O cells was performed at week 12. Immunoreaction was enhanced with Freund's adjuvant. For the generation of a pan-IDH1 antibody detecting both mutated and wild type (wt) IDH1, Sprague-Dawley rats were immunized with recombinant protein spanning the region of codon 244 to 594 fused to a hexahistidine tag. The monoclonal antibodies were raised according to the method described by Köhler and Milstein (8). Consecutive subcloning, isotyping and purification were performed following published protocols (4).

Screening of clones

For the mutation-specific antibody, all clones were tested in a first screen for immunoreaction with the 13 amino acid antigen peptide conjugated to ovalbumin by Enzyme Linked Immunosorbent Assay (ELISA) (4). For the second screen, we engineered Human Embryonic Kidney (HEK) 293T cells to transiently overexpress either full length IDH1wt (RefSeq DNA: NM_005896) or full length IDH1R132H protein (with pFLAG-CMV-D11, kindly provided by Sabine Henze, German Cancer Research Center). R132H mutation was introduced in full length IDH1 using QuikChange™ site-directed mutagenesis kit (Stratagene, La Jolla, CA, USA) following the instructions of the manual. The supernatants of all clones positive in the first ELISA screen were used as primary antibody in immunofluorescence of IDH1wt or IDH1R132H expressing HEK 293T cells plated on 96 well dishes (Becton Dickinson Labware, Franklin Lakes, NJ, USA). Cy3-labeled anti-mouse antibodies (Dianova, Hamburg, Germany) were used for visualization. One thousand six hundred fifty-five clones were tested in the second screen. Clones that showed a positive immunoreaction for IDH1R132H but not for IDH1wt overexpressing HEK 293T cells were further tested for applicability in Western blot of frozen tissue and in immunohistochemistry of routinely processed paraffin-embedded tissue.

Human tumor specimens

For Western blot, frozen tissues of tumor samples with known IDH1 mutation status were taken from the archives of the Department of Neuropathology at the University of Heidelberg. The tumor set included 6 A II (2 wt, 2 R132H, 1 R132C and 1 R132G), 2 OA II (1 wt and 1 R132H), 2 oligodendroglioma WHO grade II (O II; 1 wt and 1 R132H), 2 A III (1 wt and 1 R132H), 2 OA III (1 wt and 1 R132H) and 4 GBM (2 wt, 2 R132H).

For immunohistochemical analysis, 345 brain tumors diagnosed at the Department of Neuropathology, University of Heidelberg, Germany, and at the Department of Neuropathology, Burdenko Institute, Moscow, Russia, were investigated, including 53 A II, 31 A III, 10 OA II, 8 OA III, 16 O II, 16 anaplastic O III, 56 pGBM, 7 sGBM, 19 ependymoma WHO grade II (E II; 3 supratentorial, 5 infratentorial and 11 spinal), 47 anaplastic E III (22 supratentorial and 25 infratentorial), 21 pilocytic astrocytoma WHO grade I (PA I), 24 meningioma WHO grade I (M I), 25 medulloblastoma (MB) WHO grade IV (11 children and 14 adult patients) and 12 schwannoma WHO grade I (S I, 7 central and 5 peripheral nervous system). Gender, median age and age range are summarized in Table 1. Of these cases 189 were analyzed for IDH1 mutation by direct sequencing. The majority of genetic data has been reported in two preceding studies (1, 5). Additionally, the cortex and white matter of 10 autopsy cases without neuropathological disease were analyzed as normal brain controls. All tissues were routinely processed and analyzed as full section slides. A II, A III, O II, O III, OA II, OA III, pGBM and sGBM were all from adult patients with intracranial lesions. Informed consent was given for all molecular analyses. Diagnoses were made by histological assessment following the criteria of the WHO classification (10). All tissues were evaluated by at least two neuropathologists.

Table 1.  Binding of mIDH1R132H antibody in 345 brain tumors. Abbreviations: IDH = isocitrate dehydrogenase; WHO = World Health Organization.
TumorGender M/FAge median (range)IDH1R132H positive/total
Diffuse astrocytoma WHO II33/2036 (23–72)44/53 (83%)
Anaplastic astrocytoma WHO III11/2038 (29–73)25/31 (81%)
Oligoastrocytoma WHO II4/637 (31–66)9/10 (90%)
Anaplastic oligoastrocytoma WHO III5/341 (32–77)7/8
Oligodendroglioma WHO II12/443 (33–65)16/16 (100%)
Anaplastic oligodendroglioma WHO III8/844 (30–78)14/16 (88%)
Secondary glioblastoma4/344 (33–58)5/7
Primary glioblastoma31/2561 (33–94)2/56 (4%)
Ependymoma WHO II13/642 (1–66)0/19
Ependymoma WHO III23/249 (2–47)0/47
Pilocytic astrocytoma WHO I10/1114 (2–56)0/21
Meningioma WHO I10/1464 (31–88)0/24
Medulloblastoma WHO IV15/1022 (1–61)0/25
Schwannoma WHO I7/555 (16–66)0/12

PCR amplification

A fragment of 129-bp length spanning the catalytic domain of IDH1 including codon 132 was amplified using 60 ng each of the primer IDH1f CGGTCTTCAGAGAAGCCATT and the primer IDH1r GCAAAATCACATTATTGCCAAC. PCR using standard buffer conditions, 20 ng of DNA and GoTaq® DNA Polymerase (Promega, Madison, WI, USA) employed 35 cycles with denaturation at 95°C for 30 s, annealing at 56°C for 40 s and extension at 72°C for 50 s in a total volume of 15 µL.

For confirmation, the primer IDH1fc ACCAAATGGCACCATACGA and primer IDH1rc TTCATACCTTGCTTAATGGGTGT generating a 254-bp fragment at the same PCR conditions were employed.

Direct sequencing

Two microliters of the PCR amplification product was submitted to the sequencing reaction using the BigDye® Terminator v3.1 Sequencing Kit (Applied Biosystems, Foster City, CA, USA). Primer design was based on accession number NM_05896 for IDH1. Twenty-five cycles were performed employing 12 ng of the sense primer IDH1f CGGTCTTCAGAGAAGCCATT, with denaturation at 95°C for 30 s, annealing at 56°C for 15 s and extension at 60°C for 240 s. In case of ambiguous readings, a second round of sequencing analysis was performed using the antisense primer IDH1rc TTCATACCTTGCTTAATGGGTGT and the sequencing reaction conditions as described above. Sequences were determined using the semiautomated sequencer (ABI® 3100 Genetic Analyzer, Applied Biosystems) and the Sequence Pilot version 3.1™ (JSI-Medisys, Kippenheim, Germany) software.

Western blot

For the preparation of protein lysates, frozen tumor samples were cut on a microtome. Tumor content was determined on hematoxylin and eosin-stained slides for a total tumor content of at least 90%. About 30–50 slices of 10 µm were cut and immediately lysed in ice-cold lysis buffer consisting of tissue lysis buffer (Cell Signaling Technology, Danvers, MA, USA) and a complete mini tablet (Roche Applied Science, Indianapolis, IN, USA), vigorously vortexed, followed by sonification for 10 s at 20% power (Sonoplus HD 2070™, BANDELIN electronic, Berlin, Germany). After a 10-min clearing spin, the supernatant was stored at −80°C for further processing. Thirty micrograms of protein diluted in NuPAGE®-sample buffer and reducing reagent (Invitrogen, Carlsbad, Germany) was denatured at 95°C for 5 min and electrophoretically separated on 4%–12% Bis-Tris mini gels (Invitrogen). Blotting to nitrocellulose membranes was followed by blocking in 5% nonfat dried milk for 1 h and incubation with undiluted hybridoma supernatant of mIDH1R132H (internal clone H14) overnight at 4°C. After washing, blots were incubated with either peroxidase-labeled anti-mouse Immunoglobulin G (IgG) antibody (Cell Signaling Technology) or anti-rat IgG antibody (KPL, Gaithersburg, MD, USA) as appropriate, then incubated with LumiGLO™ Peroxidase Chemiluminescent Substrate Kit (KPL) for signal visualization. For loading control, total IDH1 was visualized on the same membrane using undiluted hybridoma supernatant of rIDH1 (clone R41).

Immunohistochemistry

Sections cut to 4 µm with a Microm HM 355 S™ microtome (Thermo Fisher Scientific, Waltham, MA, USA) with an electrical cooled object clamp (Cool-Cut™; Thermo Fisher Scientific) were dried at 80°C for 15 min and stained with anti-IDH1R132H antibody mIDH1R132H (internal clone H14) on a Ventana BenchMark XT® immunostainer (Ventana Medical Systems, Tucson, AZ, USA). The Ventana staining procedure included pretreatment with cell conditioner 2 (pH 6) for 60 min, followed by incubation with undiluted mIDH1R132H hybridoma supernatant at 37°C for 32 min. Antibody incubation was followed by Ventana standard signal amplification, UltraWash, counterstaining with one drop of hematoxylin for 4 min and one drop of bluing reagent for 4 min. For chromogenic detection, ultraView™ Universal DAB Detection Kit (Ventana Medical Systems) was used. Subsequently, slides were removed from the immunostainer, washed in water with a drop of dishwashing detergent and mounted. Double immunohistochemistry was performed using mIDH1R132H as above and polyclonal rabbit anti-glial fibrillary acidic protein (GFAP) antibody (dilution 1:8000, DakoCytomation, Glostrup, Denmark), with ultraView™ Red Detection Kit as the chromogen (Ventana Medical Systems). No chromogen was detected when primary antibody mIDH1R132H was omitted.

Immunoreaction was scored positive when tumor cells showed a strong cytoplasmic staining for mIDH1R132H. A weak diffuse staining and staining of macrophages were not scored positive.

Statistics

Fisher's exact test was used to examine the association of the presence or absence of mIDH1R132H binding in different tumor categories.

RESULTS

Selection of R132H mutation-specific clone and pan-IDH1 clone

For the identification of an R132H-specific IDH1 antibody, 1655 clones were tested in the second screen of which 78 clones showed the desired immunoprofile of detecting HEK 293T cells overexpressing IDH1R132H but not IDH1wt overexpressing cells. While the majority of these 78 clones also exhibited a positive binding pattern in Western blot, only two clones (internal clones H11 and H14) showed specific immunoreaction in paraffin-embedded tissue. As clone H14 showed a more intense signal in mutated cases, this clone was further investigated and characterized.

For the identification of a pan-IDH1 antibody, the second screen was omitted, and rat hybridoma supernatants were directly tested in Western blot and immunohistochemistry, resulting in the identification of rIDH1.

Identification of IDH1-mutated tumors in Western blot

In Western blot, analysis of 16 diffuse gliomas WHO grades II, III and IV (eight IDH1wt and eight IDH1R132H cases) mIDH1R132H detected a strong band at the predicted weight of ∼46 kD in all eight mutated tumors, but none in the eight wt tumors. A second weaker band is detected at ∼95 kD in three gliomas (OA II IDH1R132H, A III IDH1wt and GBM IDH1R132H). As the band comes up in wt and mutated tumors, it is considered as nonspecific. To control for the presence of IDH1 protein in all tumors, blots were re-incubated with the pan-IDH1 antibody rIDH1, resulting in the detection of a strong band at ∼46 kD in all tumors (Figure 1). To investigate the specificity of mIDH1R132H for the IDH1R132H mutation, we examined two diffuse astrocytomas harboring an IDH1R132C mutation or IDH1R132G mutation (Figure 2A). Neither of these was detected by mIDH1R132H. Because equivalent mutations in the highly homologous IDH2 have been described at a low frequency in diffuse astrocytomas, we also investigated if mIDH1R132H reacts with an R to H transition in the isocitrate binding site of IDH2. mIDH1R132H remained highly specific and did not detect the IDH2 mutation (Figure 2A). Frozen tissues of the very rare mutations of the R132L and R132S type were not available; therefore, we generated HEK 293T cells overexpressing these mutated IDH1 forms and tested cell lysates for reaction in Western blot. mIDH1R132H detected a signal in HEK 293T cells overexpressing IDH1R132H, but no signal in any other form of mutated IDH1 (Figure 2B).

Figure 1.

Representative Western blot of IDH1R132H-mutated and wild type (wt) gliomas demonstrating a highly specific band at the predicted molecular weight of 46 kD only in mutated tumors (left panel). On the right panel, the same blot has been incubated with an antibody (rIDH1) detecting both mutated and wt IDH1. Abbreviations: IDH = isocitrate dehydrogenase; A III = astrocytoma World Health Organization (WHO) grade III; OA III = oligoastrocytoma WHO grade III; GBM = glioblastoma.

Figure 2.

A. Lysates of tumors with IDH1R132C or IDH1R132G mutation are not detected by mIDH1R132H. One case with IDH2 mutation is also not detected. Right panel shows same blot incubated with rIDH1 to demonstrate presence of IDH1. B. HEK 293T cells overexpressing all variants of IDH1 mutations so far detected show that mIDH1R132H only detects R132H mutation (left panel). Double bands at ∼57 and ∼48 represent IDH1R132H with and without FLAG tag because of two start of translation sites in the plasmid used for overexpression. Additional weaker nonspecific bands are detected at ∼72 kD and ∼95 kD. Right panel shows re-incubation of the same blot with rIDH1 for loading control. Abbreviations: IDH = isocitrate dehydrogenase; O II = oligodendroglioma World Health Organization (WHO) grade II; A II = astrocytoma WHO grade II; OA III = oligoastrocytoma WHO grade III; wt = wild type.

Correlation of IDH1 sequencing and mIDH1R132H antibody staining

Sequencing of 186 gliomas (51 A II, 25 A III, 8 OA II, 7 OA III, 14 O II, 9 O III, 42 pGBM, 6 sGBM, 16 PA I and 8 E II) revealed 95 IDH1R132H mutations, 3 IDH1R132C mutations, 2 IDH1R132G mutations and 1 IDH1R132S mutation, while 85 cases were wt. All mutations were heterozygous. Immunohistochemistry with mIDH1R132H showed an immunoreaction in 94 of 95 of the R132H-mutated cases and, additionally, in 8 of the wt cases (4 A II, 1 A III, 2 OA III and 1 pGBM). The remaining 77 wt cases and the 6 cases with other IDH1 mutations (R132C, R132G and R132S) did not bind mIDH1R132H. In the nine mismatching cases, we performed a second round of IDH1 sequencing. For this analysis, areas with maximal mIDH1R132H binding were selected for DNA isolation, and for the single case without antibody binding, DNA was isolated from an area appearing as solid tumor. The results of this second round of sequencing matched the immunohistochemical findings in every single case, leading to the identification of eight additional R132H mutations and the correction of one falsely scored R132H case to wt. Thus, we detected 102 IDH1R132H mutations, 3 IDH1R132C mutations, 2 IDHR132G mutations and 1 IDH1R132S mutation, and 78 wt cases. Comparable with our previous series, 94% of IDH1 mutations were of the R132H type. After re-sequencing, the sensitivity and specificity of mIDH1R132H staining to detect R132H mutation were both 100%, while sensitivity and specificity of mIDH1R132H to detect all types of IDH1 mutations were 94% (102 of 108) and 100% (78 of 78), respectively.

Characterization of mIDH1R132H antibody staining in brain tumors

A total of 345 formalin-fixed and paraffin-embedded brain tumors were stained on an automated immunostainer (Ventana Medical Systems). Of these, 122 bound mIDH1R132H (see Table 1). Positive cases showed a strong cytoplasmic and often a weaker nuclear staining of tumor cells (Figure 3A), with 15 cases showing an additional diffuse staining of the fibrillary tumor matrix. The diffuse staining was restricted to tumor-containing areas and was thus also considered a specific antibody staining. Cases with isolated nuclear staining were not observed.

Figure 3.

A. Astrocytoma World Health Organization (WHO) grade II (A II) with cytoplasmic as well as nuclear binding of mIDH1R132H (×200). B. Double immunohistochemistry of glial fibrillary acidic protein (red) and mIDH1R132H (brown) of infiltration zone of anaplastic oligodendroglioma showing that reactive astrocytes (red) are not bound by mIDH1R132H (×200). C. IDH1R132G-mutated A II does not exhibit mIDH1R132H binding (×200). D. Infiltration zone of diffuse astrocytoma with tumor cells forming long uni- to bipolar processes (×200). E. Infiltration zone of oligodendroglioma WHO grade III. Although the tumor seems sharply demarcated to normal brain (right), single infiltrating tumor cells can be seen (×200). F. Tumor cells in the infiltration zone of anaplastic astrocytoma showing perivascular accumulation (×200). G. Positively stained tumor cells in close proximity to neurons as an example of perineuronal satellitosis (×400). H. Identification of loose subpial spread of tumor cells distant from tumor bulk by mIDH1R132H (×100). Abbreviation: IDH = isocitrate dehydrogenase.

In positive cases, all histologically identifiable tumor cells were marked positive with mIDH1R132H, whereas endothelial cells, perivascular lymphocytes and residual brain glial cells were negative. Reactive astrocytes were also not recognized by mIDH1R132H (Figure 3B). All 10 normal brain controls from autopsy did not show binding of mIDH1R132H. In 13 of the surgical specimens, a granular staining of neurons was observed, most likely representing nonspecific binding to lipofuscin. When present, macrophages often showed a strong granular cytoplasmic antibody binding even in IDH1wt tumors (Figure 4F). Negative cases often showed a weak diffuse background staining. Because of the intensity of antibody binding in cases with R132H mutation, all negative cases were easily distinguishable from positive cases. Staining of all six cases with other IDH1 mutations was scored negative (Figure 3C). Staining of the three cases with R132C mutation demonstrated slightly more diffuse background than the two cases with R132G or the one case with R132S mutation.

Figure 4.

A. High magnification of astrocytoma World Health Organization (WHO) grade II showing mIDH1R132H positive tumor cells with uni- to bipolar processes (×400). B. In 7 of 31 astrocytoma WHO grade III, fine stellar processes in the majority of tumor cells were seen (×200). C. Gemistocytes often showed a clearing of antibody binding in central cytoplasm (×400). D. mIDH1R132H binding of oligodendroglial tumor bulk. Cellular processes were generally not observed (×200). E. mIDH1R132H showed no binding in pilocytic astrocytoma tumor cells (×200). F. High magnification of a different case of pilocytic astrocytoma showing positive macrophages. Tumor cells show no binding of mIDH1R132H (×400). G. mIDH1R132H did not bind ependymoma, here WHO III (×200). H. The majority of meningiomas showed nonspecific staining of fibers usually close to blood vessels or around small groups of tumor cells (×200). Abbreviation: IDH = isocitrate dehydrogenase.

Because of the strong binding of mIDH1R132H to all putative tumor cells, this antibody seems especially useful for the identification of infiltrating tumor cells and the assessment of patterns of tumor infiltration. In diffuse astrocytoma (WHO grades II and III), a widespread infiltration of tumor cells in the cortex or white matter was observed in the majority of cases (Figure 3D). The morphology of glioma cells in the infiltration zone often clearly differed from that of the tumor center with tumor cells forming unexpectedly long and thin processes. For oligodendrogliomas, diffuse infiltration strongly resembling that of astrocytic tumors was observed as well as areas of more compact infiltration (Figure 3E). Antibody staining revealed several patterns of infiltration, including perivascular infiltration (Figure 3F) and perineuronal satellitosis (Figure 3G). In several cases, loosely distributed subpial tumor cells were only detected in immunohistochemistry with mIDH1R132H (Figure 3H).

Forty-four of 53 investigated A II (83%) were positive for mIDH1R132H, often demarcating uni- to bipolar tumor cell processes (Figure 4A). A III showed positive binding of 25/31 cases (81%). While the majority of cases strongly resembled the morphology of A II, in seven of the positive cases (28%), antibody binding highlighted very thin stellar processes in the majority of tumor cells (Figure 4B). This pattern was not observed in other tumors of this series. Tumors with a gemistocytic component (both A II and A III) often showed reduced staining in the central cytoplasm of gemistocytes (Figure 4C).

Staining for mIDH1R132H was positive in 16/16 O II and 14/16 O III cases. Oligodendroglial tumors showed an intense cytoplasmic staining with only few visible processes in the tumor bulk (Figure 4D) and often uni- to bipolar processes in the infiltration zone. “Mini gemistocytes” did not show the reduced central immunoreaction as observed for true astrocytic gemistocytes.

In mixed glioma (OA II and OA III), strong mIDH1R132H binding was observed in both astrocytic and oligodendroglial differentiated cells. Nine of 10 OA II and 7/8 OA III were positive.

Among 56 pGBM, 2 positive cases were detected (3%), while 5 of 7 sGBM bound mIDH1R132H. In all positive cases, GBM immunostaining was distributed throughout the whole tumor; viable tumor areas with lack of staining were not observed. Specific patterns of immunoexpression in proximity to necrosis were not observed.

We also investigated other primary brain tumor entities known for low frequencies of IDH1 mutations for mIDH1R132H antibody binding. In line with the low mutation rate, the 21 PA I, 19 E II, 47 E III and 25 MB did not show immunostaining of tumor cells (Figure 4E,G). In 8 of 21 PA I, a moderate number of positive macrophages were detected among the negative tumor cells (Figure 4F). Six of 66 E II and E III showed weak background staining, however, not reaching signal intensity of cases with mutation.

The investigation of non-neuroectodermal brain tumors without mutations of IDH1 to date comprised 24 M I and 12 S I. In both entities, no mIDH1R132H binding of tumor cells was observed. Although tumor cells were negative, 19/24 meningiomas and 5/12 schwannomas demonstrated a peculiar pattern of antibody binding with fine positive-stained fibers within the tumor, usually arranged along tumor blood vessels. In meningioma, the positive fibers were also observed separating small groups of tumor cells (Figure 4H).

Differentiation of gliomas by mIDH1R132H binding

Classification of gliomas is of major clinical importance for tumor treatment and assessment of prognosis. The high rate of positive A III and the low rate of immunoreactive pGBM make mIDH1R132H a useful marker to distinguish these two entities. A III and pGBM show a significantly different distribution of mIDH1R132H binding (P < 0.0001). The sensitivity of antibody binding for the diagnosis of A III in this setting is 81% (25 of 31 A III), with a specificity of 96% (negative in 54 of 56 pGBM). As evidence is increasing that IDH1 mutation may serve as a differentiation marker for pGBM and sGBM, we also analyzed the potential of mIDH1R132H to differentiate these two entities. The sensitivity and specificity of mIDH1R132H staining to differentiate pGBM and sGBM in our series were 71% (5 of 7) and 96% (54 of 56), respectively. We recently demonstrated the feasibility of differentiating PA I and A II by genetic analysis of IDH1 and human v-raf murine sarcoma viral oncogene homolog B1 (BRAF) (9). As expected, immunohistochemical analysis with mIDH1R132H yielded comparable and highly significant results (P < 0.0001). Forty-four of 53 A II, but none of 21 PA I, were recognized by mIDH1R132H corresponding to a sensitivity of 83% and a specificity of 100%. Because the differentiation of A II from E II, and A III from E III may represent a diagnostic challenge, we included a large series of ependymomas (n = 66). As none of the E II was positive for mIDH1R132H, sensitivity and specificity are the same as when comparing A II and PA I (83% and 100%, respectively). For the differentiation of A III and E III, the sensitivity of antibody binding for the diagnosis of A III is 81% (25 of 31 A III), and the specificity is 100% (negative in 47 of 47 E III).

DISCUSSION

IDH1 mutations are emerging as a major diagnostic and prognostic marker for gliomas. Here we describe the characterization of a highly specific antibody detecting the most frequent IDH1 mutation of the R132H type in Western blot and immunohistochemistry.

Western blot analysis of 16 gliomas demonstrates that mutated IDH1 protein is highly expressed in mutated tumors of low to high WHO grade. mIDH1R132H antibody is highly specific for the R132H mutation and does not detect any other IDH1 mutations described to date. In our series we find no evidence for changes of total IDH1 protein levels comparing wt and IDH1R132H mutated tumors, indicating that there is neither an obvious upregulation of IDH1 in mutated tumors nor a downregulation of IDH1 in wt tumors.

The capability of mIDH1R132H to differentiate mutated and wt IDH1 in paraffin-embedded tissue makes this antibody an invaluable tool for applications in pathology. Several recent studies have demonstrated the importance of IDH1 mutation as a prognostic marker in anaplastic glioma (A III, O III and OA III) (16, 20) and GBM (14, 16, 19, 21). While routine genetic analysis of gliomas for IDH1 mutations is currently not available in the majority of diagnostic institutions, immunohistochemistry is a standard procedure. We could demonstrate that our immunohistochemical approach perfectly matches results of direct sequencing of IDH1 codon R132H and, therefore, immunohistochemistry was adequate to determine genetic status of total IDH1 mutation, with a sensitivity of 94% and a specificity of 100% in this series. The sensitivity of 94% is because of the lack of detection of other types of IDH1 mutations. The generation of further mutation-specific antibodies specific for the next frequent mutations (R132C and R132G) would increase sensitivity close to 100%. In eight cases IDH1 mutation was missed in the first round of direct sequencing but was detected after re-sequencing DNA obtained from tumor regions with higher tumor cell content. We consider this as a general problem of DNA-dependent genetic analysis of diffuse brain tumors where estimation of tumor content may be difficult and the risk of genetically analyzing tissue with low tumor cell content is eminent. Antibody-based detection of IDH1R132H mutation appears to be superior in sensitivity compared with direct sequencing, especially in low-grade diffuse astrocytomas.

Homogenous staining of all identifiable tumor cells with mIDH1R132H gives clear diagnostic implications. The staining of all cells permits the identification of single infiltrating cells in otherwise inconspicuous brain tissue. As we have not observed positive cells in any of the IDH1wt tumors, we consider all stained cells as neoplastic. Reactive astrocytes in tumors and other lesions do not show immunoreaction with mIDH1R132H. The traits of this antibody allow the identification of single tumor cells in very small samples, such as stereotactic biopsies, and in the infiltration zone of diffuse gliomas. The possibility to determine borders of diffuse brain tumors with, to date, unseen accuracy may offer further advances for future brain tumor surgery.

The patterns of infiltration in glioma are stereotypic and have been designated as “secondary structures of Scherer”(17). These patterns include tumor cell spread along white matter tracts, perivascular tumor cell spread, subpial growth and perineuronal satellitosis. mIDH1R132H highlighted these growth patterns in many cases. The clear discrimination of tumor cells and non-tumor cells by mIDH1R132H offers a new tool to study the interaction of glioma cells with reactive astrocytes, inflammatory cells and vessel cells, and may lead to new insights in the cause of these patterns of infiltration.

Other than established markers for glioma diagnostics such as GFAP and microtubule-associated protein 2, IDH1 is a soluble cytosolic protein. This prompted us to expect new morphological aspects in immunohistochemical assessment. A clearly different feature is that mIDH1R132H was found to bind in the cytoplasm as well as the nucleus. Whether mutated IDH1 is truly localized in the nucleus in vivo or whether the soluble protein penetrates the nucleus after surgical removal (sometimes referred to as antigen diffusion) has not been resolved yet. In our investigation of mIDH1R132H binding, no apparent differences between A II and O II were observed. A subset of A III demonstrated evenly spaced mIDH1R132H positive tumor cells, with radial processes morphologically resembling reactive astrocytes (Figure 4B). Interestingly, this specific cellular morphology was observed in both the tumor center and the infiltration zone. In few A II single cells of this morphology were also observed. Without immunohistochemistry, these tumor portions most likely would have been interpreted as reactive change. None of the meningiomas and schwannomas so far revealed IDH1 mutations (1). However, we frequently detected fine positive fibers in these tumors. We consider this nonspecific staining that may be caused by a cross-reaction with a perivascular extracellular matrix protein or a collagen subtype. This staining pattern can easily be recognized.

IDH1 mutations seem to be suitable to distinguish subtypes of gliomas. The high sensitivity and specificity of mIDH1R132H distinguish A III from pGBM (81% and 96% respectively) and may prove to be superior to separation based on the presence or absence of vascular proliferation and/or necrosis (10). Obviously, the latter histological features may be lacking in small or nonrepresentative tumor samples. This may have been the case in two of the six negative A III cases of this series. One of these cases displayed suspected necrosis on Magnetic Resonance Imaging (MRI) and amplification of Epidermal Growth Factor Receptor (EGFR), both typically associated with pGBM, and most likely, the available tissue was not representative. For patients with A III diagnosed by the current WHO classification IDH1 mutation is associated with better survival (16, 20). It is attractive to speculate that a significant number of IDH1wt A III diagnosed by the current WHO criteria may represent “underdiagnosed” or “missed” pGBM cases that did not show the classical histological features of pGBM. Future studies on larger numbers of A III are necessary to clarify to what extent IDH1 mutations may contribute to the classification of gliomas. mIDH1R132H will be a useful tool for this new field of investigation, because over 90% of IDH1-mutated A III are of the R132H type (5).

IDH1 mutation has recently been established as a molecular marker to distinguish pGBM and sGBM (14). Although fewer tumors were investigated in our study, sensitivity and specificity of mIDH1R132H staining to distinguish pGBM and sGBM were almost identical to those of the previous study based on sequencing analysis (respectively 73.3% and 96.3% from Nobusawa et al, and 71% and 96% in this study). Because pGBM with IDH1 mutations share further genetic traits of sGBM, it has been proposed that pGBM with IDH1 mutation actually represent sGBM with an unusually short clinical presentation (14). This is emphasized by the fact that pGBM with IDH1 mutation follows a somewhat more benign course of disease, resembling the slightly better prognosis of sGBM (14, 19). mIDH1R132H antibody staining can serve as an easy screening method for the identification of sGBM clinically presenting as pGBM. Further, mIDH1R132H may also be useful in differentiating A II from PA I, or A II and A III from E II and E III.

With mIDH1R132H, we did detect astrocytic tumor cells resembling reactive astrocytes. However, in none of our specimens did reactive astrocytes capture mIDH1R132H. While this might be more difficult to proof in astrocytic tumors with IDH1 mutations, such observation was evident for all reactive astrocytes in lesions without IDHI mutations examined in this series. We did not detect any positivity in reactive astrocytes accompanying pGBM without IDH1 mutation, E II, E III, MB and PA I. Therefore, this antibody is expected to be of great diagnostic help in differentiating reactive gliosis and diffuse low-grade brain tumors of minimal hypercellularity and with low cytological atypia.

In conclusion, we demonstrate the generation of a highly specific antibody useful for neuropathological diagnostics. Immunohistochemistry with mIDH1R132H is equal, if not superior, to genetic testing for the detection of the most frequent IDH1 mutation of the R132H type. It identifies single infiltrating tumor cells, thereby greatly assisting diagnostics of small tumor samples. Finally, it may serve as a highly reliable differentiation marker to separate A III from pGBM, pGBM from sGBM, A II from PA I, or A II and A III from E II and E III.

ACKNOWLEDGMENTS

This work was supported by the Bundesministerium für Bildung und Forschung (BMBF—01ES0730 and 01GS0883).

Ancillary