Journal of Neuro-Oncology

, Volume 117, Issue 1, pp 175–182

Diffusion-weighted MRI derived apparent diffusion coefficient identifies prognostically distinct subgroups of pediatric diffuse intrinsic pontine glioma

Authors

  • Robert M. Lober
    • Department of Neurosurgery, Lucile Packard Children’s HospitalStanford University
  • Yoon-Jae Cho
    • Department of Neurology, Lucile Packard Children’s HospitalStanford University
  • Yujie Tang
    • Department of Neurology, Lucile Packard Children’s HospitalStanford University
  • Patrick D. Barnes
    • Department of Radiology, Lucile Packard Children’s HospitalStanford University
  • Michael S. Edwards
    • Department of Neurosurgery, Lucile Packard Children’s HospitalStanford University
  • Hannes Vogel
    • Department of Pathology, Lucile Packard Children’s HospitalStanford University
  • Paul G. Fisher
    • Department of Neurology, Lucile Packard Children’s HospitalStanford University
  • Michelle Monje
    • Department of Neurology, Lucile Packard Children’s HospitalStanford University
    • Department of Radiology, Lucile Packard Children’s HospitalStanford University
Clinical Study

DOI: 10.1007/s11060-014-1375-8

Cite this article as:
Lober, R.M., Cho, Y., Tang, Y. et al. J Neurooncol (2014) 117: 175. doi:10.1007/s11060-014-1375-8

Abstract

While pediatric diffuse intrinsic pontine gliomas (DIPG) remain fatal, recent data have shown subgroups with distinct molecular biology and clinical behavior. We hypothesized that diffusion-weighted MRI can be used as a prognostic marker to stratify DIPG subsets with distinct clinical behavior. Apparent diffusion coefficient (ADC) values derived from diffusion-weighted MRI were computed in 20 consecutive children with treatment-naïve DIPG tumors. The median ADC for the cohort was used to stratify the tumors into low and high ADC groups. Survival, gender, therapy, and potential steroid effects were compared between the ADC groups. Median age at diagnosis was 6.6 (range 2.3–13.2) years, with median follow-up seven (range 1–36) months. There were 14 boys and six girls. Seventeen patients received radiotherapy, five received chemotherapy, and six underwent cerebrospinal fluid diversion. The median ADC of 1,295 × 10−6 mm2/s for the cohort partitioned tumors into low or high diffusion groups, which had distinct median survivals of 3 and 13 months, respectively (log-rank p < 0.001). Low ADC tumors were found only in boys, whereas high ADC tumors were found in both boys and girls. Available tissue specimens in three low ADC tumors demonstrated high-grade histology, whereas one high ADC tumor demonstrated low-grade histology with a histone H3.1 K27M mutation and high-grade metastatic lesion at autopsy. ADC derived from diffusion-weighted MRI may identify prognostically distinct subgroups of pediatric DIPG.

Keywords

Diffuse intrinsic pontine glioma (DIPG)DiffusionDiffusion-weighted imaging (DWI)Apparent diffusion coefficient (ADC)MRI

Introduction

Diffuse intrinsic pontine glioma (DIPG) accounts for 10–15 % of all pediatric brain tumors. It affects ~400 children per year in the United States and accounts for the majority of deaths due to childhood brain tumors [1]. It is not surgically cureable and current adjuvant therapies provide limited benefit beyond transient neurological improvement [2, 3].

On MRI, DIPG presents as a heterogeneous mass that enlarges the pons with irregular interfaces with the midbrain and medulla, and ventral extension into the prepontine cistern [4, 5]. These features distinguish uniformly fatal DIPG from the dramatically more favorable focal brainstem gliomas [4, 5]. There are no characteristic enhancement patterns [3, 5] and most studies have shown poor correlation between conventional MRI features and clinical outcome [6, 7]. One large analysis of MRI features across two clinical trials has demonstrated worse outcomes with enhancing tumors [8].

Recent molecular studies have shown that DIPG subsets with distinct biology exist [9] and may harbor unique mutations and genetic expressions different from adult high-grade gliomas and pediatric hemispheric glioblastomas [911]. DIPG are diagnosed by MRI and currently biopsied at very few institutions, which limits the translational usefulness of these studies and highlights the value of an imaging biomarker predictive of clinical behavior.

Studies have shown utility of diffusion MRI, or diffusion weighted imaging (DWI), in predicting tumor pathology, histological grade, and clinical behavior [1220]. DWI may even have a role for tumor monitoring, including reports of normalization of apparent diffusion coefficient (ADC) derived from DWI during chemotherapy and return to pre-therapy levels with tumor recurrence [21, 22]. Early work by Chen et al. [23] demonstrated a positive linear correlation of ADC with survival in DIPG of nine children and described its potential role for DIPG evaluation. More recent work by Poussaint et al. [8] has shown that diffusion changes after radiotherapy are associated with survival in DIPG patients. However, no diffusion metric has yet been defined that stratifies clinically distinct DIPG subsets. In this study, we hypothesized that diffusion imaging can be used to stratify DIPG subsets with distinct clinical behavior.

Materials and methods

Subjects

All patients diagnosed with DIPG at our children’s hospital between 2001 and 2013 were retrospectively reviewed after approval by our Institutional Review Board (protocol IRB-4223). The study cohort was identified on the basis of the following inclusion criteria: the patient presented with a new DIPG diagnosis, obtained a treatment-naïve diffusion MRI, and had follow-up clinical data, including time of death. The diagnosis was made by MRI features of an expansile pontine lesion usually characterized by T1 hypointensity and T2 hyperintensity following presentation with neurologic signs suggestive of a brainstem lesion or hydrocephalus. Medical records were reviewed for steroid administration, surgery, chemotherapy, radiotherapy, and cerebrospinal fluid (CSF) diversion.

Image acquisition

All patients were examined with brain MR imaging at 1.5 or 3 Tesla (Signa, Discovery 750; GE Medical Systems, Milwaukee, WI, USA). Echo planar DWI was obtained using the following parameters: DWI ASSET, 4 mm, 0 skip, 3-direction, TR/TE 8300/70 (1.5 T); 10,000/86 (3 T), b = 1,000 mm2/s, acceleration factor = 2, FOV 24 cm, matrix 128 × 128. Generation of ADC maps was performed with commercially available software (Functool; GE Medical Systems). As part of our routine tumor brain protocol, additional MR sequences included T1 FLAIR, T2 fast spin echo, FLAIR, GRE, T1 spoiled gradient recalled pre and post contrast, and two additional plane contrast-enhanced T1 spin echo sequences.

Diffusion measurements

Mean ADC measurements of solid tumor were obtained by placing ROI by a board-certified pediatric neuroradiologist (KY) with certificate of added qualification (7-years of experience) blinded to clinical or pathological information. Each ROI was manually drawn along the entire tumor boundary and at all available axial planes, avoiding areas of cyst or cavities, hemorrhage/mineralization, and necrosis as identified by areas of ring enhancement (see example in Fig. 1). A second blinded neuroradiologist (PB) with certificate of added qualification (over 30-years of experience) independently confirmed appropriate ROI placement. A weighted average of the mean ADC based on the ROI size at each axial level was computed for each patient.
https://static-content.springer.com/image/art%3A10.1007%2Fs11060-014-1375-8/MediaObjects/11060_2014_1375_Fig1_HTML.jpg
Fig. 1

a Axial T2-weighted image shows characteristic features of DIPG with a heterogeneous mass enlarging the pons and distorting the middle cerebellar peduncles. b Example of ROI placement over the solid tumor for ADC evaluation. A weighted average of the mean ADC based on the ROI size at each axial level was computed for each patient. Here the tumor is seen with higher mean diffusivity compared to the normal-appearing cerebellum (small ROI)

Autopsy and histological preparations

For the three patients who underwent autopsy, brains were harvested and examined after formalin fixation, overseen by HV. Sequential coronal sections of the cerebrum and transverse sections of the brainstem and cerebellum perpendicular to their long axes were made, and representative sections were taken for histological evaluation. Formalin-stained sections were stained with standard hematoxylin and eosin preparations.

Screening for K27M mutation

Tissue specimens from two cases were prepared and analyzed by YT, YJC, and MM. Genomic DNA was extracted from available frozen post-mortem tumor tissues or control cortex tissues of DIPG patients using DNeasy Blood & Tissue Kit (Qiagen). Extracted DNA used for polymerase chain reaction (PCR) amplification of the genomic regions for targeted Sanger sequencing of the Histone 3.3 (H3F3A) and Histone 3.1 (HIST1H3B) genes. The PCR amplicons were then extracted using QIAquick Gel Extraction Kit (Qiagen) and underwent Sanger sequencing with the primer sequences H3F3A_gDNA-F (aggccgttcgaggtattttt) and H3F3A_gDNA-R (aaatgagggggtaggaacttt). Histone 3 gene analysis was targeted since recent studies have found that somatic mutation of the Histone 3.3 or 3.1 genes are common in pediatric high-grade gliomas, particularly midline tumors [10, 11, 24].

Statistical analyses

All statistical analyses were performed by RL using SPSS Statistics version 21.0 (IBM) with an a priori significance level of 0.05, corrected to 0.003 by the Bonferroni method for 16 statistical tests applied to the same patient sample. Kaplan–Meier curves for survival were compared using the Mantel-Cox log-rank and Breslow tests. Differences in distributions of ADC among gender were assessed by the Kruskal–Wallis test for independent samples. Differences in treatment characteristics (surgery, radiotherapy, chemotherapy, CSF diversion, and steroid administration), among ADC groups and genders were assessed with the Fisher’s exact test. Mean ADC was compared between patients with and without steroid administration using the Mann–Whitney U Test.

Results

Clinical and diffusion MRI data

Twenty children met the inclusion criteria and were included in the study. Median age at diagnosis was 6.6 (range 2.3–13.2) years, with median survival seven (range 1–36) months. There were 14 boys and six girls. Seventeen patients received radiotherapy, five received chemotherapy, and six underwent CSF diversion. All but one patient succumbed to the disease at the time of study completion.

The median number of ROIs for the cohort was seven (range four to nine), with a median ROI size of 1,016 mm2. The median ADC of 1,295 × 10−6 mm2/s for the cohort partitioned tumors into “low” or “high” ADC groups, which had distinct median survivals of three and 13 months, respectively (Breslow p < 0.001; Mantel-Cox log-rank p < 0.001) (Fig. 2). Low ADC tumors were found only in boys, whereas high ADC tumors were found in both boys and girls (Table 1).
https://static-content.springer.com/image/art%3A10.1007%2Fs11060-014-1375-8/MediaObjects/11060_2014_1375_Fig2_HTML.gif
Fig. 2

Kaplan–Meier curves comparing survival between patients with low and high ADC tumors. Median survival was 3 months for low ADC tumors and 13 months for high ADC tumors

Table 1

Baseline ADC of DIPG stratified by gender (n = 20)

 

Boys (n = 14)

Girls (n = 6)

Low ADC group

10

0

High ADC group

4

6

  

Fisher’s exact test p = 0.011

Mean ADC (10−6 mm2/s)

1,218 ± 188

1,481 ± 117

Mann–Whitney U Test p = 0.006

There was a trend toward worse survival in boys by the Breslow statistic (p = 0.034) and Mantel-Cox log-rank tests (p = 0.070) (Fig. 3a). High ADC tumors in male patients had a survival curve resembling that of female patients, each having a median survival of 13 months (Fig. 3b).
https://static-content.springer.com/image/art%3A10.1007%2Fs11060-014-1375-8/MediaObjects/11060_2014_1375_Fig3_HTML.gif
Fig. 3

a Kaplan–Meier curves comparing survival between boys and girls. Median survival was 5 months for boys and 13 months for girls. b Kaplan–Meier curves for low and high ADC tumors in boys only. Median survival was 3 months for low ADC tumors and 13 months for high ADC tumors

There was no difference in treatment characteristics (surgery, radiotherapy, chemotherapy, and CSF diversion) between ADC groups (p = 1.000) or genders (p = 1.000), although three of the eight patients with low ADC tumors did not complete radiotherapy because of worsening symptoms or a parental decision to switch to palliative care. Of the 20 patients, 16 had detailed information in their medical records regarding steroid administration, which was not different between ADC groups (p = 0.608) and did not affect mean ADC (1,327 ± 196 for steroid administration and 1,230 ± 229 for no steroids; p = 0.428). The feature of contrast enhancement was present in six patients, and was equally distributed between ADC groups.

Cases with tissue correlation

One patient obtained biopsy at initial diagnosis because of atypical enhancement and possible multifocality; and three patients underwent autopsy. Grade III (in the biopsy and one autopsy) and IV histology were seen in three patients 1–2 months after initial presentation with low ADC tumors. Grade II in pons and grade IV histology in distant periventricular brain regions were present at autopsy in one patient 19 months after presentation with a high ADC tumor. Examples are shown in Figs. 4 and 5. Tissue was available for genomic analysis in two patients: one patient with low ADC tumor had wild-type histone H3 loci; one patient had high ADC tumor had K27M mutation at the histone H3.1 locus.
https://static-content.springer.com/image/art%3A10.1007%2Fs11060-014-1375-8/MediaObjects/11060_2014_1375_Fig4_HTML.jpg
Fig. 4

Example of a low ADC tumor. a An 8 year-old boy presented to our hospital with headache, double vision, multiple cranial neuropathies, and imbalance. The tumor showed necrosis and enhancement in the pons. b ADC map demonstrated low baseline (1,183 × 10−6 mm2/s). He began local field radiotherapy but his parents discontinued this after four doses because of progressive symptoms and poor quality of life. He survived 2 months. c Hematoxylin and eosin staining of paraffin-embedded section of the pons taken an autopsy demonstrated grade IV features of microvascular proliferation and pseudopallisading necrosis. Genetic analysis revealed wild-type histone H3 loci

https://static-content.springer.com/image/art%3A10.1007%2Fs11060-014-1375-8/MediaObjects/11060_2014_1375_Fig5_HTML.jpg
Fig. 5

Example of a high ADC tumor. a A 14 year-old boy presented with a high baseline ADC (1,393 × 10−6 mm2/s). b Follow-up showed mixed regional mean ADC values of 1,172 and 1,875 (10−6 mm2/s). He was followed without treatment for 4 months, at which point he developed hydrocephalus, underwent CSF diversion and started treatment. After 1 year, he became wheelchair-bound from right-sided ataxia. He survived a total of 21 months. c Pons specimen at autopsy. d Bland histology in the pons demonstrated mostly grade II histology. e Post-contrast T1-weighted image had shown an enhancing right periventricular lesion late in the clinical course. f A low ADC value was noted (728 × 10−6 mm2/s) corresponding to this periventricular lesion. g Periventricular lesion at autopsy. h Hematoxylin and eosin staining of paraffin-embedded section of periventricular lesion. i High magnification of periventricular lesion showed grade IV histology. Genetic analysis revealed K27M mutation at the histone H3.1 locus

Discussion

DIPG has a dismal prognosis with a median survival under 1 year [2, 25]. However, among DIPG some tumors are unpredictably aggressive, resulting in death within a few months. Recently, gene expression profiles have revealed at least two biological subtypes of DIPG with distinct survival and molecular characteristics [9], and attempts have been made to stratify tumors based on specific mutations, such as K27M in the histone H3.3 or 3.1 genes [11, 26].

Aside from imaging features that distinguish uniformly fatal DIPG and more favorable focal brainstem gliomas, no conventional MRI features have shown to correlate with clinical outcome [47]. Several recent studies have found that MR perfusion and spectroscopy may help assess clinical course in DIPG [27, 28]. Here, we investigated diffusion MRI as a prognostic imaging marker that could predict clinical behavior and thereby help stratify patients for prospective studies and clinical trials.

Diffusion MRI measures the movement of both extracellular and intracellular water within an image voxel [18, 29] and has increasingly been applied in brain tumors and, in some cases, for therapeutic monitoring and prognosis [1216, 19, 21, 22, 2931]. A recent study has inferred based on correlation with perfusion that focal anaplasia in DIPG can be identified with diffusion imaging [32], and that diffusion changes after radiotherapy may correlate with survival [8]. Our results suggest diffusion metrics can define clinically distinct subgroups of pediatric DIPG, with lower tumor ADC at initial presentation predictive of aggressive clinical behavior. Low ADC in clinically more aggressive tumors suggest these may represent higher grade tumors at presentation, similar to prior studies of adult glioma in which ADC values inversely correlated to cell density and nuclear area to total cytoplasm ratio [12, 14, 15]. This is also supported by tissue specimens available in select patients in our cohort that showed grade III or IV histology in low ADC tumors and grade II histology in a high ADC tumor. It is noteworthy, that while the latter patient showed grade II histological features within pons at autopsy, the distant tumor nodules displayed grade IV features, suggesting diverse biology and evolution pattern.

This is also the first study to note potential gender differences in survival of DIPG. There was a trend toward worse survival in boys. Interestingly, low ADC tumors were found only in boys, whereas high ADC tumors were found in both boys and girls (Table 1). The low ADC tumors appeared to account entirely for the trend toward worse survival in the boys of this cohort.

Recent studies have identified somatic mutation of the H3F3A gene encoding the histone 3.3 or HIST1H3B gene encoding Histone 3.1 that results specifically in lysine 27-to-methionine (K27M) substitution in over 30 % of pediatric high-grade gliomas [10] and 70 % of DIPG cases, [24, 26]. Interestingly, K27M gliomas often have distinct midline location pattern, including thalamus and pons with K27M molecular signatures closely resembling mid to late fetal stages of striatum and thalamus that might suggest potential tumor cellular origins [11]. Given frequent K27M mutation in midline pediatric gliomas, we were curious regarding the K27M status in our autopsy samples. Here, we observed K27M mutation in a high ADC tumor but not in low ADC tumor. The significance of this may be better understood with a larger cohort.

This study is limited by its low sample size and retrospective nature, as treatment decisions are often based on many factors, including patient quality of life and parental input regarding goals of care. Although we found that therapies were not different between low and high ADC groups, three of eight patients initiated on with low ADC tumors were unable to complete radiotherapy because of poor quality of life and the wishes of their parents. Steroid administration could potentially alter tumor ADC, but this was not different between low and high ADC groups. Another limitation is the paucity of histological data and the nature of the autopsy specimens that reflects the disease state at the end of the patient’s course, rather than what would potentially be found at the time of diagnosis.

Conclusion

ADC may identify distinct subgroups of pediatric DIPG and may be a useful noninvasive marker for patient selection in clinical trials.

Acknowledgments

The authors express our sincere gratitude to Dr. Terri Haddix for her assistance in histologic preparations and photography of specimens.

Conflict of interest

There are no conflicts of interest or funding source to disclose.

Copyright information

© Springer Science+Business Media New York 2014