Spatially Resolved Whole Transcriptome Analysis of Histologically-Characterized Tissue Microarray of Patient-Matched Primary and Recurrent Glioblastomas to Identify Underlying Mechanisms of Treatment Resistance

PURPOSE/OBJECTIVE(S)

Glioblastoma multiforme (GBM) is a lethal CNS malignancy. Radiation therapy increases overall survival, but tumors often recur in high-radiation dose regions. Additionally, recent investigations have underscored the importance of intra-tumoral heterogeneity as a driver of GBM biology. The purpose of this investigation is to characterize transcriptome differences in primary and recurrent GBM patient clinical samples using a digital spatial profiling approach to better appreciate treatment resistance mechanisms.

MATERIALS/METHODS

To address the lack of understanding of molecular mechanisms of resistance in GBM, patient-matched primary and recurrent GBM pathological specimens were identified within the brain tissue biorepository and tissue cores were selected for generation of a tissue microarray (TMA). Hematoxylin and eosin staining with histomorphological (cellular tumor, pseudopalisading necrosis, invasive edge, and perivascular inflammation) scoring were performed in a blinded fashion for every core. This array was then molecularly characterized using digital spatial profiling of the transcriptome. Quality assurance including filtering of lowly expressed genes followed by downstream analyses of the data were performed using the manufacturer's recommended methods within their Bioconductor library. Gene Set Enrichment Analysis (GSEA) was then performed on the ranked gene lists.

RESULTS

After recommended filtering, 6171 genes and 248 regions of interest remained for downstream analysis representing 22 unique patients across four different tumor histomorphological types. Significance testing revealed 679 genes that were differentially expressed between primary and recurrent tumor samples (at FDR<1%). On GSEA analysis, the chromosomal positional locus that contains genes most strongly up-regulated is 12q14, a locus that was previously identified as genomically amplified in multiple patient-derived xenograft lines after radiation selection. Additionally, recurrent tumors display a transcriptional profile more similar to the mesenchymal subtype, whereas primary tumors have a more classical transcriptional phenotype. The epithelial-to-mesenchymal transition pathway is particularly strongly up-regulated in recurrent tumors.

CONCLUSION

Recurrent selection at previously identified genomic loci and molecular pathways underscores a possible conserved set of pathways for treatment resistance. This analysis has yielded a set of gene and molecular pathways that will guide future work in our lab targeting treatment resistance using novel therapeutics and radiation techniques in GBM. Future directions include assessing the feasibility of mapping these clinical samples onto our previously generated panel of comprehensively characterized patient-derived xenograft lines.

Abstract P6-11-06: Efficacy of carboplatin alone or with ABT888 in an intracranial murine model of BRCA-mutated, basal-like, triple negative breast cancer (TNBC)

Background: TNBC frequently spreads to the central nervous system and is associated with poor outcome. Twenty percent of women with TNBC harbor a BRCA mutation(mut). BRCA dysfunction impairs repair of DNA double-strand breaks by homologous recombination(HR) and sensitizes tumor cells to DNA-damaging cytotoxic agents. Emerging evidence suggests that TNBC responds to platinums(Pt) and drugs potentiating efficiency of DNA damage, such as PARP inhibitors; combination therapy may enhance DNA damage and induce additional cytotoxicity. Recognizing that Pt and many PARP inhibitors penetrate the blood brain barrier(BBB), we evaluated the efficacy of carboplatin(Carbo) alone and with the PARP inhibitor, ABT888(ABT), in murine intracranial(IC) TNBC models characterized by BRCA status.

Methods: Athymic (nu/nu) mice(10wks) were inoculated intracerebrally with 2×105 SUM149(BRCA1mut, pTEN-; basal) cells expressing luciferase. On day14, tumor-bearing mice began treatment with Control(PBS IP weekly), Carbo(50mg/kg IP weekly) and ABT(25mg/kg OG daily) as single agents, or in combination. Survival and IC tumor growth via bioluminescence were compared by treatment. Net PAR levels(PARpg/ml/100ug total protein) were measured in IC tumor tissue(ELISA pharmacodynamic assay) 14days post ABT. Presence of Pt DNA adducts was measured in IC tumor tissue following Carbo+/-ABT by ICP-MS. Survival and IC tumor growth via bioluminescence of the MDA-MB468(BRCAnon-mut, pTEN-; basal) IC model treated with the same regimens were assessed.

Results: In the SUM149 TNBC IC model, Carbo+/-ABT resulted in improved survival compared to control (Table). Single agent ABT did not yield a survival benefit over control. Combination Carbo/ABT showed a modest advantage in survival compared to Carbo alone (p = 0.28). IC tumor mean net PAR levels ± SEM were significantly lower in ABT-treated animals compared to control (182.1 ± 91.3 vs 825.7 ± 174.7, p = 0.0043). Addition of ABT did not impair Carbo BBB penetration; Mean ± SEM Pt-DNA adducts were detected in IC tumor after administration of Carbo and Carbo/ABT (0.91±0.13 and 1.46±0.03 pgPt/ugDNA). In contrast to the SUM149 IC model, there was no survival improvement for the MDA-MB468 IC model treated with Carbo+/-ABT compared to control(p = 0.81).

 SUM149(BRCA1mut)MDA-MB468(BRCA1non-mut)TreatmentNMedian Survival (95%CI), dayspValueNMedian Survival (95%CI), daysControl1736(34-40)N/A535(29-57)ABT1039(30-46)p = 0.22544(33-57)Carbo1658(47-67)p&lt;.0001537(33-54)Carbo/ABT1764(59-75)p&lt;.0001537(33-50)

Conclusions: This study shows Pt analogues, specifically Carbo, represent a promising treatment strategy for TNBC brain metastases with BRCA dysfunction. Improved survival relies on the ability of Pt analogues to penetrate the BBB and higher cytotoxicity may be due to compromised HR in TNBC as compared to models with functional BRCA. Although inhibitory activity of ABT was detected in IC tumors, we did not observe a significant additive effect of ABT to Carbo when used in combination. Confirmatory studies in additional IC TNBC models, including claudin-low subtype, as well as mechanistic studies evaluating DNA damage via γH2AX foci and differential gene expression in response to treatment are ongoing.

Citation Information: Cancer Res 2013;73(24 Suppl): Abstract nr P6-11-06.

Sequence-dependent variations associated with H2A/H2B depletion of nucleosomes

Mechanisms that can alter nucleosome structure to enhance DNA accessibility are of great interest because of their potential involvement in genomic processes. One such mechanism is H2A/H2B release from nucleosomes; it occurs in vivo and is involved in the in vitro activities of several transcription-associated complexes. Using fluorescence approaches based on Förster resonance energy transfer, we previously detected sequence-dependent structure/stability variations between 5S and two types of promoter nucleosomes (from yeast GAL10 or mouse mammary tumor virus promoters). Those variations included differing responses when nucleosomes were diluted to concentrations (sub-nM) known to produce H2A/H2B loss. Here, we show that treatment of these same three types of nucleosomes with the histone chaperone yNAP-1, which causes H2A/H2B release from nucleosomes in vitro, produces the same differential Förster resonance energy transfer responses, again demonstrating sequence-dependent variations associated with conditions that produce H2A/H2B loss. Single-molecule population data indicate that DNA dynamics on the particles produced by diluting nucleosomes to sub-nM concentrations follow two-state behavior. Rate information (determined by fluorescence correlation spectroscopy) suggests that these dynamics are enhanced in MMTV-B or GAL10 compared to 5S particles. Taken together, the results indicate that H2A/H2B loss has differing effects on 5S compared to these two promoter nucleosomes and the differences reflect sequence-dependent structure/stability variations in the depleted particles.

Using atomic force microscopy to study chromatin structure and nucleosome remodeling

Atomic force microscopy (AFM) is a technique that can directly image single molecules in solution and it therefore provides a powerful tool for obtaining unique insights into the basic properties of biological materials and the functional processes in which they are involved. We have used AFM to analyze basic features of nucleosomes in arrays, such as DNA-histone binding strength, cooperativity in template occupation, nucleosome stabilities, nucleosome locations and the effects of acetylation, to compare these features in different types of arrays and to track the response of array nucleosomes to the action of the human Swi-Snf ATP-dependent nucleosome remodeling complex. These experiments required several specific adaptations of basic AFM methods, such as repetitive imaging of the same fields of molecules in liquid, the ability to change the environmental conditions of the sample being imaged and detection of specific types of molecules within compositionally complex samples. Here, we describe the techniques that allowed such analyses to be carried out.

Properties of nucleosomes in acetylated mouse mammary tumor virus versus 5S arrays

Acetylation is one of the most abundant histone modifications found in nucleosomes. Although such modifications are thought to function mainly in recognition, acetylation is known to produce nucleosome structural alterations. These could be of functional significance in vivo. Here, the basic features of mouse mammary tumor virus (MMTV) promoter nucleosomal arrays reconstituted with highly acetylated histones prepared from butyrate-treated HeLa cells are characterized by atomic force microscopy. Results are compared to previous results obtained with hypoacetylated MMTV and hyper- or hypoacetylated 5S rDNA arrays. MMTV arrays containing highly acetylated histones show diminished intramolecular compaction compared to hypoacetylated MMTV arrays and no tendency for cooperativity in nucleosome occupation. Both features have been suggested to reflect histone tail-mediated internucleosomal interactions; these observations are consistent with that suggestion. 5S arrays show qualitatively similar behavior. Two other effects of acetylation show stronger DNA template dependence. Nucleosome salt stability is diminished in highly acetylated compared to hypoacetylated MMTV arrays, but nucleosome (histone) loading tendencies are unaffected by acetylation. However, highly acetylated histones show reduced loading tendencies on 5S templates (vs hypoacetylated), but 5S nucleosome salt stabilities are unaffected by acetylation. ATP-dependent nucleosome remodeling by human Swi-Snf is similar on hyper- and hypoacetylated MMTV arrays.

Two-component atomic force microscopy recognition imaging of complex samples

Biological complexes are typically multisubunit in nature and the processes in which they participate often involve protein compositional changes in themselves and/or their target substrates. Being able to identify more than one type of protein in complex samples and to track compositional changes during processes would thus be very useful. Toward this goal, we describe here a single-molecule technique that can simultaneously identify two types of proteins in compositionally complex samples. It is an adaptation of the recently developed atomic force microscopy (AFM) recognition imaging technique but involves the tethering of two different types of antibodies to the AFM tip and sequential blocking with appropriate antigenic peptides to distinguish the recognition from each antibody. The approach is shown to be capable of simultaneously identifying in a single AFM image two specific components, BRG1 and beta-actin, of the human Swi-Snf ATP-dependent nucleosome remodeling complex and two types of histones, H2A and H3, in chromatin samples.

Sequence-Dependent Nucleosome Structure and Stability Variations Detected by Förster Resonance Energy Transfer

Nucleosomes, the basic unit of eukaryotic chromosome structure, cover most of the DNA in eukaryotes, including regulatory sequences. Here, a recently developed Förster resonance energy transfer approach is used to compare structure and stability features of sea urchin 5S nucleosomes and nucleosomes reconstituted on two promoter sequences that are nucleosomal in vivo, containing the yeast GAL10 TATA or the major transcription response elements from the mouse mammary tumor virus promoter. All three sequences form mononucleosomes with similar gel mobilities and similar stabilities at moderate salt concentrations. However, the two promoter nucleosomes differ from 5S nucleosomes in (1) diffusion coefficient values, which suggest differences in nucleosome compaction, (2) intrinsic FRET efficiencies (in solution or in gels), and (3) the response of FRET efficiency to high (>or=600 mM) NaCl concentrations, subnanomolar nucleosome concentrations, and elevated temperatures (to 42 degrees C). These results indicate that nucleosome features can vary depending on the DNA sequence they contain and show that this fluorescence approach is sufficiently sensitive to detect such differences. Sequence-dependent variations in nucleosome structure or stability could facilitate specific nucleosome recognition, working together with other known genomic regulatory mechanisms. The variations in salt-, concentration-, and temperature-dependent responses all occur under conditions that have been shown previously to produce release of H2A-H2B dimers or terminal DNA from nucleosomes and could thus involve differences in those processes, as well as in other features.

Deciphering cancer complexities in genetically engineered mice

Because the pRb pathway is disrupted in most solid human cancers, we have generated genetically engineered mouse cancer models by inactivating pRb function in several cell types, including astrocytes and mammary, prostate, ovarian, and brain choroid plexus epithelia. In every case, proliferation and apoptosis are acutely induced, predisposing to malignancy. Cell type dictates the pathways involved in tumor progression. In the astrocytoma model, we developed strategies to induce events in the adult brain, either throughout the tissue or focally. Both K-Ras activation and Pten inactivation play significant roles in progression. In the prostate model, adenocarcinoma progression depends on Pten inactivation. However, nonautonomous induction of p53 in the mesenchyme leads to evolution of both compartments, with p53 loss occurring in the mesenchyme. Thus, studies in these models continue to identify key tumorigenesis mechanisms. Furthermore, we are hopeful that the models will provide useful preclinical systems for diagnostic and therapeutic development.

AFM imaging of protein movements: histone H2A-H2B release during nucleosome remodeling

Being able to follow assembly/disassembly reactions of biomolecular complexes directly at the single molecule level would be very useful. Here, we use an AFM technique that can simultaneously obtain topographic images and identify the locations of a specific type of protein within those images to monitor the histone H2A component of nucleosomes acted on by human Swi-Snf, an ATP-dependent nucleosome remodeling complex. Activation of remodeling results in significant H2A release from nucleosomes, based on recognition imaging and nucleosome height changes, and changes in the recognition patterns of H2A associated directly with hSwi-Snf complexes.

Solution AFM studies of human Swi-Snf and its interactions with MMTV DNA and chromatin

ATP-dependent nucleosome remodeling complexes are crucial for relieving nucleosome repression during transcription, DNA replication, recombination, and repair. Remodeling complexes can carry out a variety of reactions on chromatin substrates but precisely how they do so remains a topic of active inquiry. Here, a novel recognition atomic force microscopy (AFM) approach is used to characterize human Swi-Snf (hSwi-Snf) nucleosome remodeling complexes in solution. This information is then used to locate hSwi-Snf complexes bound to mouse mammary tumor virus promoter nucleosomal arrays, a natural target of hSwi-Snf action, in solution topographic AFM images of surface-tethered arrays. By comparing the same individual chromatin arrays before and after hSwi-Snf activation, remodeling events on these arrays can be monitored in relation to the complexes bound to them. Remodeling is observed to be: inherently heterogeneous; nonprocessive; able to occur near and far from bound complexes; often associated with nucleosome height decreases. These height decreases frequently occur near sites of DNA release from chromatin. hSwi-Snf is usually incorporated into nucleosomal arrays, with multiple DNA strands entering into it from various directions, + or - ATP; these DNA paths can change after hSwi-Snf activation. hSwi-Snf appears to interact with naked mouse mammary tumor virus DNA somewhat differently than with chromatin and ATP activation of surface-bound DNA/hSwi-Snf produces no changes detectable by AFM.

Using atomic force microscopy to study nucleosome remodeling on individual nucleosomal arrays in situ

In eukaryotes, genomic processes like transcription, replication, repair, and recombination typically require alterations in nucleosome structure on specific DNA regions to operate. ATP-dependent nucleosome remodeling complexes provide a major mechanism for carrying out such alterations in vivo. To learn more about the action of these important complexes, we have utilized an atomic force microscopy in situ technique that permits comparison of the same individual molecules before and after activation of a particular process, in this case nucleosome remodeling. This direct approach was used to look for changes induced by the action of the human Swi-Snf remodeling complex on individual, single-copy mouse mammary tumor virus promoter nucleosomal arrays. Using this technique, we detect a variety of changes on remodeling. Many of these changes are larger in scale than suggested from previous studies and involve a number of DNA-mediated events, including a preference for the removal of a complete turn (80 basepairs) of nucleosomal DNA. The latter result raises the possibility of an unanticipated mode of human Swi-Snf interaction with the nucleosome, namely via the 11-nm histone surface.

A statistical thermodynamic model applied to experimental AFM population and location data is able to quantify DNA-histone binding strength and internucleosomal interaction differences between acetylated and unacetylated nucleosomal arrays

Imaging of nucleosomal arrays by atomic force microscopy allows a determination of the exact statistical distributions for the numbers of nucleosomes per array and the locations of nucleosomes on the arrays. This precision makes such data an excellent reference for testing models of nucleosome occupation on multisite DNA templates. The approach presented here uses a simple statistical thermodynamic model to calculate theoretical population and positional distributions and compares them to experimental distributions previously determined for 5S rDNA nucleosomal arrays (208-12,172-12). The model considers the possible locations of nucleosomes on the template, and takes as principal parameters an average free energy of interaction between histone octamers and DNA, and an average wrapping length of DNA around the octamers. Analysis of positional statistics shows that it is possible to consider interactions between nucleosomes and positioning effects as perturbations on a random positioning noninteracting model. Analysis of the population statistics is used to determine histone-DNA association constants and to test for differences in the free energies of nucleosome formation with different types of histone octamers, namely acetylated or unacetylated, and different DNA templates, namely 172-12 or 208-12 5S rDNA multisite templates. The results show that the two template DNAs bind histones with similar affinities but histone acetylation weakens the association of histones with both templates. Analysis of locational statistics is used to determine the strength of specific nucleosome positioning tendencies by the DNA templates, and the strength of the interactions between neighboring nucleosomes. The results show only weak positioning tendencies and that unacetylated nucleosomes interact much more strongly with one another than acetylated nucleosomes; in fact acetylation appears to induce a small anticooperative occupation effect between neighboring nucleosomes.

Single-molecule recognition imaging microscopy

Atomic force microscopy is a powerful and widely used imaging technique that can visualize single molecules and follow processes at the single-molecule level both in air and in solution. For maximum usefulness in biological applications, atomic force microscopy needs to be able to identify specific types of molecules in an image, much as fluorescent tags do for optical microscopy. The results presented here demonstrate that the highly specific antibody-antigen interaction can be used to generate single-molecule maps of specific types of molecules in a compositionally complex sample while simultaneously carrying out high-resolution topographic imaging. Because it can identify specific components, the technique can be used to map composition over an image and to detect compositional changes occurring during a process.

Development of a fluorescent probe for the study of nucleosome assembly and dynamics

To develop a probe for use in real-time dynamic studies of nucleosomes, core histones (from Drosophila) were conjugated to a DNA-intercalating dye, thiazole orange, by a reaction targeting Cys 110 of histone H3. In the absence of DNA, the conjugated histones are only very weakly fluorescent. However, upon reconstitution into nucleosomes by standard salt dialysis procedures, the probe fluoresces strongly, reflecting its ability to intercalate into the nucleosomal DNA. The probe is also sensitive to the nature of the DNA-histone interaction. Nucleosomes reconstituted by stepwise salt dialysis give a fluorescence signal quite different from that of the species formed when DNA and histones are simply mixed in low salt. In addition, changing either the DNA length or the type of sequence (nucleosome positioning sequences versus random DNA of the same size) used in the reconstitution alters the resulting fluorescence yield. The results are all consistent with the conclusion that a more rigid, less flexible nucleosome structure results in less fluorescence than a looser structure, presumably due to structural constraints on dye intercalation. This probe should be well suited to analyzing nucleosome dynamics and to following factor-mediated assembly and remodeling of nucleosomes in real time, particularly at the single-molecule level.

Nucleosomal arrays can be salt-reconstituted on a single-copy MMTV promoter DNA template: their properties differ in several ways from those of comparable 5S concatameric arrays

Subsaturated nucleosomal arrays were reconstituted on a single-copy MMTV promoter DNA fragment by salt dialysis procedures and studied by atomic force microscopy. Up to an occupation level of approximately eight nucleosomes on this 1900 bp template, salt reconstitution produces nucleosomal arrays which look very similar to comparably loaded 5S rDNA nucleosomal arrays; i.e., nucleosomes are dispersed on the DNA template. Thus, at these occupation levels, the single-copy MMTV template forms arrays suitable for biophysical analyses. A quantitative comparison of the population features of subsaturated MMTV and 5S arrays detects differences between the two: a requirement for higher histone levels to achieve a given level of nucleosome occupation on MMTV templates, indicating that nucleosome loading is thermodynamically less favorable on this template; a preference for pairwise nucleosome occupation of the MMTV (but not the 5S) template at midrange occupation levels; and an enhanced salt stability for nucleosomes on MMTV versus 5S arrays, particularly in the midrange of array occupation. When average occupation levels exceed approximately eight nucleosomes per template, MMTV arrays show a significant level of mainly intramolecular compaction; 5S arrays do not. Taken together, these results show clearly that the nature of the underlying DNA template can affect the physical properties of nucleosomal arrays. DNA sequence-directed differences in the physical properties of chromatin may have important consequences for functional processes such as gene regulation.

Yeast chromatin structure and regulation of GAL gene expression

Yeast genomic DNA is covered by nucleosome cores spaced by short, discrete length linkers. The short linkers, reinforced by novel histone properties, create a number of unique and dynamic nucleosome structural features in vivo: permanent unpeeling of DNA from the ends of the core, an inability to bind even full 147 bp core DNA lengths, and facility to undergo a conformational transition that resembles the changes found in active chromatin. These features probably explain how yeast can maintain most of its genome in a transcribable state and avoid large-scale packaging away of inactive genes. The GAL genes provide a closely regulated system in which to study gene-specific chromatin structure. GAL structural genes are inactive without galactose but are highly transcribed in its presence; the expression patterns of the regulatory genes can account for many of the features of GAL structural gene control. In the inactive state, GAL genes demonstrate a characteristic promoter chromosomal organization; the major upstream activation sequence (UASG) elements lie in open, hypersensitive regions, whereas the TATA and transcription start sites are in nucleosomes. This organization helps implement gene regulation in this state and may benefit the organism. Induction of GAL expression triggers Gal4p-dependent upstream nucleosome disruption. Disruption is transient and can readily be reversed by a Gal80p-dependent nucleosome deposition process. Both are sensitive to the metabolic state of the cell. Induction triggers different kinds of nucleosome changes on the coding sequences, perhaps reflecting the differing roles of nucleosomes on coding versus promoter regions. GAL gene activation is a complex process involving multiple Gal4p activities, numerous positive and negative cofactors, and the histone tails. DNA bending and chromosomal architecture of the promoter regions may also play a role in GAL regulation. Regulator-mediated competition between nucleosomes and the TATA binding protein complex for the TATA region is probably a central aspect of GAL regulation and a focal point for the numerous factors and processes that contribute to it.

Kinetics of early therapeutic response as measured by quantitative PCR predicts survival in a murine xenograft model of human T cell acute lymphoblastic leukemia

The identification of prognostic parameters and surrogate markers for defining patient risk has been beneficial in effectively guiding therapy and increasing the survival of leukemia patients. It has been hypothesized that the therapeutic response, as measured by a change in tumor burden during therapy, might serve as a new surrogate marker of survival. Here we describe the development of a murine SCID xenograft model of human T cell acute lymphoblastic leukemia (T-ALL), and the use of a sensitive, quantitative PCR assay for the measurement of tumor levels to investigate the relationships between tumor burden quantification, therapeutic response and survival. Animals engrafted with the CCRF-CEM (CEM) human T-ALL cell line develop leukemia that closely resembles the human disease. Quantitative PCR detects the expanding tumor mass in the peripheral blood of the animals several weeks before death. In response to induction therapy with chemotherapeutic agents, both the level of minimal residual disease (MRD) in peripheral blood at the end of therapy and the rate of tumor reduction in peripheral blood during therapy strongly correlated with animal survival. Thus, these surrogate markers, which can be measured during the early stages of therapy, may help improve patient survival through dynamic risk stratification.

Conformational transition in DNA on a cold surface

The contour length of DNA fragments, deposited and imaged on mica under buffer, was measured as a function of deposition temperature. Extended DNA molecules (on Ni- and silane-treated surfaces) contract rapidly with falling temperature, approaching the contour length of A-DNA at 2 degrees C. The contraction is not unique to a specific sequence and does not occur in solution at 2 degrees C or on a surface at 25 degrees C, indicating that it arises from a combination of low temperature and surface contact. It is probably a consequence of reduced water activity at a cold surface.

The yield of routine chest radiography in children with cancer hospitalized for fever and neutropenia

A routine admission chest radiograph (CXR) in pediatric patients with cancer who are admitted to the hospital for fever and neutropenia has been advised because the signs and symptoms of pneumonia may be absent. The authors studied 131 consecutive patient admissions for fever and neutropenia to evaluate the diagnostic yield of routine CXR. All patients had a complete history, physical examination, complete blood count, blood culture, urinalysis, urine culture, and CXR. Patients routinely started ceftazidime monotherapy. Results of the CXR were correlated with the presence or absence of signs and symptoms of respiratory disease. Of 128 CXR results, 26 (20%) were abnormal (13 with known malignant disease, 2 with atelectasis, 3 with peribronchial cuffing, and 8 with pneumonia [6%]). Three patients with pneumonia were asymptomatic. Therefore, only 3 of 128 patients (2.3%) had pneumonia on CXR not suspected by physical examination. None would have had initial therapy modified based on the CXR finding alone. The authors concluded that the incidence of pneumonia in a child with fever and neutropenia is low and that routine CXR at diagnostic evaluation is unnecessary in the asymptomatic ambulatory patient.