Kuangyu Shi

Acute versus chronic hypoxia: why a simplified classification is simply not enough.

For more than 100 years, it has been suspected that oxygen levels can influence the radiosensitivity of cells (1). Further basic investigations on the importance of hypoxia for radiotherapy were then performed in the 1950s (2). Today, it is generally accepted that diminished oxygen availability leads to substantial limitations in the efficacy of oxygen-dependent treatment modalities, such as standard radiotherapy, some O2-dependent chemotherapy, photodynamic therapy, and immunotherapy (3). Primarily, hypoxia leads to impairment of the proliferative capacity and eventually (in the additional absence of major substrates, e.g., glucose) to death of normal and most tumor cells. On the other hand, a minority of tumor cells can escape hypoxia-induced cell death by triggering alterations in the proteome and/or genome, thus favoring tumor progression (4). This phenomenon has been called the ‘‘Janus face of hypoxia’’ (4). Although a gradual transition of hypoxic effects (also called a continuum of the effects) is most likely, experimental evidence exists that at O2 concentrations <1%, changes in transcription, gene, and protein expression are often induced in tumor cells. At O2 concentrations <0.1%, malignant cells can undergo permanent genomic and epigenomic modifications. However, not only the above-mentioned absolute hypoxia levels but also fluctuating hypoxia levels with time (i.e., cycles of hypoxia–normoxia) and steep spatial oxygen gradients (‘‘4D heterogeneity’’) may be prerequisites for the development of aggressive survival strategies (local invasion, metastasis, and acquired treatment resistance). Taken

Quantitative assessment of hypoxia subtypes in microcirculatory supply units of malignant tumors using (immuno-)fluorescence techniques.

Background and Purpose:Hypoxia is a characteristic of tumors, is known to increase aggressiveness, and causes treatment re-sistance. Traditional classification suggests two types of hypoxia: chronic and acute. Acute hypoxia is mostly caused by transient disruptions in perfusion, while chronic hypoxia is caused by diffusion limitations. This classification may be insufficient in terms of pathogenetic and pathophysiological mechanisms. Therefore, we quantified hypoxia subtypes in tumors based on (immuno-)fluorescent marker distribution patterns in microcirculatory supply units (MCSUs).Material and Methods:Cryosections from hSCC lines (SAS, FaDu, UT-SCC-5, UT-SCC-14, UT-SCC-15) were analyzed. Hypoxia was identified by pimonidazole, perfusion by Hoechst 33342, and endothelial cells by CD31. The following patterns were identified in vital tumor tissue: (1) normoxia: Hoechst 33342 fluorescence around microvessels, no pimonidazole, (2) chronic hypoxia: Hoechst 33342 fluorescence around microvessels, pimonidazole distant from microvessels, (3) acute hypoxia: no Hoechst 33342 fluorescence around microvessels, pimonidazole in immediate vicinity of microvessels, and (4) hypoxemic hypoxia: Hoechst 33342 fluorescence and pimonidazole directly around microvessels.Results:Quantitative assessment of MCSUs show predominance for normoxia in 4 out of 5 tumor lines (50.1–72.8%). Total hypoxia slightly prevails in UT-SCC-15 (56.9%). Chronic hypoxia is the dominant subtype (65.4–85.9% of total hypoxia). Acute hypoxia only accounts for 12.9–29.8% and hypoxemic hypoxia for 1.2–6.4% of total hypoxia. The fraction of perfused microvessels ranged from 82.5–96.6%.Conclusion:Chronic hypoxia is the prevailing subtype in MCSUs. Acute hypoxia and hypoxemic hypoxia account for only a small fraction. This approach enables assessment and recognition of different hypoxia subtypes including hypoxemic hypoxia and may facilitate methods to (clinically) identify and eliminate hypoxia.ZusammenfassungHintergrund und Ziel:Hypoxie ist ein Charakteristikum solider Tumoren, führt zur Tumorprogression und Therapieresistenz. Traditionell werden chronische und akute Hypoxie unterschieden. Erstere beruht vor allem auf Diffusionslimitierungen, letztere bevorzugt auf Perfusionsstörungen. Diese Klassifizierung reicht nicht aus, um pathogenetische und pathophysiologische Mechanismen weiter aufzuklären. Deshalb werden Hypoxiesubtypen in mikrozirkulatorischen Versorgungseinheiten (MCSUs) mit Hilfe von (Immun-)Fluoreszenztechniken identifiziert.Material und Methoden:Gefrierschnitte von xenotransplantierten menschlichen Plattenepithelkarzinomen (SAS, FaDu, UT-SCC-5, UT-SCC-14, UT-SCC-15) werden nach Pimonidazol-Färbung zur Hypoxiemarkierung, Hoechst-33342-Fluoreszenz zum Perfusionsnachweis und CD31-Gefäßdarstellung untersucht. Folgende Muster können in vitalem Gewebe nachgewiesen werden: (1) Normoxie: Hoechst-33342-Fluoreszenz um Gefäße, keine Pimonidazol-Anfärbung; (2) chronische Hypoxie: Hoechst-33342-Fluoreszenz in direkter Gefäßnähe, Pimonidazol in einer gewissen Distanz zu den Gefäßen; (3) akute Hypoxie: Hoechst-33342-Flu-oreszenz fehlt, Pimonidazol in unmittelbarer Gefäßnachbarschaft und (4) hypoxämische Hypoxie: Hoechst-33342-Fluoreszenz und Pimonidazol in direkter Gefäßnachbarschaft.Ergebnisse:Die Verteilungsmuster von Hoechst, Pimonidazol und CD31 in den MCSUs weisen darauf hin, dass in 4 der 5 Tumorlinien normoxische Areale überwiegen (50,1–72,8%). Lediglich UT-SCC-15-Tumoren sind überwiegend (56,9%) hypoxisch. Die Analyse der Hypoxiesubtypen zeigt, dass chronische Hypoxie eindeutig überwiegt (65,4–85,9% der Gesamthypoxie). Auf die akute Hypoxie entfallen lediglich 12,9–29,8% der Gesamthypoxie. Der Anteil der hypoxämischen Hypoxie ist am kleinsten (1,2–6,4% der Gesamthypoxie). Die Fraktion der perfundierten Gefäße beträgt 82,5–96,6%.Schlussfolgerung:Chronische Hypoxie herrscht in mikrozirkulatorischen Versorgungseinheiten der untersuchten Plattenepithelkarzinome vor. Akute und hypoxämische Hypoxie spielen nur eine untergeordnete Rolle. Der experimentelle Ansatz erlaubt erstmalig die Erfassung der hypoxämische Hypoxie im Tumorgewebe und ermöglicht eine Differenzierung verschiedener Hypoxiesubtypen. Die beschriebene Methode könnte die (quantitative) Detektion hypoxischer Areale und klinisch relevante Maßnahmen zur Verbesserung des Oxygenierungsstatus erleichtern.

The lower hippocampus global connectivity, the higher its local metabolism in Alzheimer disease.

Objectives: Based on the hippocampus disconnection hypothesis in Alzheimer disease (AD), which postulates that uncoupling from cortical inputs contributes to disinhibition-like changes in hippocampus activity, we suggested that in patients with AD, the more the intrinsic functional connectivity between hippocampus and precuneus is decreased, the higher hippocampal glucose metabolism will be. Methods: Forty patients with mild AD dementia, 21 patients with mild cognitive impairment, and 26 healthy controls underwent simultaneous PET/MRI measurements on an integrated PET/MR scanner. 18F-fluorodeoxyglucose-PET was used to measure local glucose metabolism as proxy for neural activity, and resting-state functional MRI with seed-based functional connectivity analysis was performed to measure intrinsic functional connectivity as proxy for neural coupling. Group comparisons and correlation analysis were corrected for effects of regional atrophy, partial volume effect, age, and sex. Results: In both patient groups, intrinsic connectivity between hippocampus and precuneus was significantly reduced. Moreover, in both patient groups, glucose metabolism was reduced in the precuneus (AD < mild cognitive impairment < controls) while unchanged in the hippocampus. Critically, the lower connectivity between hippocampus and precuneus was in patients with AD dementia, the higher was hippocampus metabolism. Conclusion: Results provide evidence that in patients with AD dementia, stronger decrease of intrinsic connectivity between hippocampus and precuneus is linked with higher intrahippocampal metabolism (probably reflecting higher neuronal activity). These data support the hippocampus disconnection hypothesis, i.e., uncoupling from cortical inputs may contribute to disinhibition-like changes of hippocampal activity with potentially adverse consequences on both intrahippocampal physiology and clinical outcome.

Changes in the fraction of total hypoxia and hypoxia subtypes in human squamous cell carcinomas upon fractionated irradiation: Evaluation using pattern recognition in microcirculatory supply units

BACKGROUND AND PURPOSE Evaluate changes in total hypoxia and hypoxia subtypes in vital tumor tissue of human head and neck squamous cell carcinomas (hHNSCC) upon fractionated irradiation. MATERIALS AND METHODS Xenograft tumors were generated from 5 hHNSCC cell lines (UT-SCC-15, FaDu, SAS, UT-SCC-5 and UT-SCC-14). Hypoxia subtypes were quantified in cryosections based on (immuno-)fluorescent marker distribution patterns of Hoechst 33342 (perfusion), pimonidazole (hypoxia) and CD31 (endothelium) in microcirculatory supply units (MCSUs). Tumors were irradiated with 5 or 10 fractions of 2 Gy, 5×/week. RESULTS Upon irradiation with 10 fractions, the overall fraction of hypoxic MCSUs decreased in UT-SCC-15, FaDu and SAS, remained the same in UT-SCC-5 and increased in UT-SCC-14. Decreases were observed in the proportion of chronically hypoxic MCSUs in UT-SCC-15, in the fraction of acutely hypoxic MCSUs in UT-SCC-15 and SAS, and in the percentage of hypoxemically hypoxic MCSUs in SAS tumors. After irradiation with 5 fractions, there were no significant changes in hypoxia subtypes. Changes in the overall fraction of hypoxic MCSUs were comparable to corresponding alterations in the proportions of acutely hypoxic MCSUs. There was no correlation between radiation resistance (TCD(50)) and any of the investigated hypoxic fractions upon fractionated irradiation. CONCLUSIONS This study shows that there are large alterations in the fractions of hypoxia subtypes upon irradiation that can differ from changes in the overall fraction of hypoxic MCSUs.

Comparison of (immuno-)fluorescence data with serial [18F]Fmiso PET/CT imaging for assessment of chronic and acute hypoxia in head and neck cancers

PURPOSE Both, acute and chronic hypoxia can have unfavorable impacts on tumor progression and therapy response. The aim of this study was to optimize a macroscopic technique for the quantification of acute and chronic hypoxia (Wang model assessment of serial [(18)F]Fmiso PET/CT imaging) by comparing with a microscopic technique [(immuno-)fluorescence staining in tumor cryosections]. MATERIALS AND METHODS Tumor pieces from the human squamous cell carcinoma lines from the head and neck FaDu and CAL33 were xenografted into the hind leg of NMRI nu/nu mice. Tumor-bearing mice were placed on an in-house developed multi-point fixation system and subjected to two consecutive dynamic [(18)F]Fmiso PET/CTs within a 24h interval. The Wang model was applied to SUV (standard uptake values) to quantify the fractions of acute and chronic hypoxia. Hypoxia subtypes were also assessed in vital tumor tissue of cryosections from the same tumors for (immuno-)fluorescence distributions of Hoechst 33342 (perfusion), pimonidazole (hypoxia), and CD31 (endothelium) using pattern recognition in microcirculatory supply units (defined as vital tumor tissue area supplied by a single microvessel). RESULTS Using our multi-point fixation system, acceptable co-registration (registration errors ε ranged from 0.34 to 1.37) between serial PET/CT images within individual voxels was achieved. The Wang model consistently yielded higher fractions of acute hypoxia than the MCSU method. Through specific modification of the Wang model (Wang(mod)), it was possible to reduce the fraction of acute hypoxia. However, there was no significant correlation between the fractions of acute hypoxia in individual tumors assessed by the Wang(mod) model and the MCSU method for either tumor line (FaDu: r=0.68, p=0.21 and CAL33: r=0.71, p=0.18). This lack of correlation is most-likely due to the difference between the non-linear uptake of [(18)F]Fmiso and the spatial assessment of MCSUs. CONCLUSIONS Whether the Wang model can be used to predict radiation response after serial [(18)F]Fmiso PET imaging, needs to be confirmed in experimental and clinical studies.

Exploring the quantitative relationship between metabolism and enzymatic phenotype by physiological modeling of glucose metabolism and lactate oxidation in solid tumors.

Molecular imaging using PET or hyperpolarized MRI can characterize tumor phenotypes by assessing the related metabolism of certain substrates. However, the interpretation of the substrate turnover in terms of a pathophysiological understanding is not straightforward and only semiquantitative. The metabolism of imaging probes is influenced by a number of factors, such as the microvascular structure or the expression of key enzymes. This study aims to use computational simulation to investigate the relationship between the metabolism behind molecular imaging and the underlying tumor phenotype. The study focused on the pathways of glucose metabolism and lactate oxidation in order to establish the quantitative relationship between the expression of several transporters (GLUT, MCT1 and MCT4), expression of the enzyme hexokinase (HK), microvasculature and the metabolism of glucose or lactate and the extracellular pH distribution. A computational model for a 2D tumor tissue phantom was constructed and the spatio-temporal evolution of related species (e.g. oxygen, glucose, lactate, protons, bicarbonate ions) was estimated by solving reaction-diffusion equations. The proposed model was tested by the verification of the simulation results using in vivo and in vitro literature data. The influences of different expression levels of GLUT, MCT1, MCT4, HK and microvessel distribution on substrate concentrations were analyzed. The major results are consistent with experimental data (e.g. GLUT is more influential to glycolytic flux than HK; extracellular pH is not correlated with MCT expressions) and provide theoretical interpretation of the co-influence of multiple factors of the tumor microenvironment. This computational simulation may assist the generation of hypotheses to bridge the discrepancy between tumor metabolism and the functions of transporters and enzymes. It has the potential to accelerate the development of multi-modal imaging strategies for assessment of tumor phenotypes.

Direct Parametric Image Reconstruction in Reduced Parameter Space for Rapid Multi-Tracer PET Imaging

The separation of multiple PET tracers within an overlapping scan based on intrinsic differences of tracer pharmacokinetics is challenging, due to limited signal-to-noise ratio (SNR) of PET measurements and high complexity of fitting models. In this study, we developed a direct parametric image reconstruction (DPIR) method for estimating kinetic parameters and recovering single tracer information from rapid multi-tracer PET measurements. This is achieved by integrating a multi-tracer model in a reduced parameter space (RPS) into dynamic image reconstruction. This new RPS model is reformulated from an existing multi-tracer model and contains fewer parameters for kinetic fitting. Ordered-subsets expectation-maximization (OSEM) was employed to approximate log-likelihood function with respect to kinetic parameters. To incorporate the multi-tracer model, an iterative weighted nonlinear least square (WNLS) method was employed. The proposed multi-tracer DPIR (MT-DPIR) algorithm was evaluated on dual-tracer PET simulations ([18F]FDG and [11C]MET) as well as on preclinical PET measurements ([18F]FLT and [18F]FDG). The performance of the proposed algorithm was compared to the indirect parameter estimation method with the original dual-tracer model. The respective contributions of the RPS technique and the DPIR method to the performance of the new algorithm were analyzed in detail. For the preclinical evaluation, the tracer separation results were compared with single [18F]FDG scans of the same subjects measured two days before the dual-tracer scan. The results of the simulation and preclinical studies demonstrate that the proposed MT-DPIR method can improve the separation of multiple tracers for PET image quantification and kinetic parameter estimations.

Preclinical evaluation of parametric image reconstruction of [18F]FMISO PET: correlation with ex vivo immunohistochemistry

Compared to indirect methods, direct parametric image reconstruction (PIR) has the advantage of high quality and low statistical errors. However, it is not yet clear if this improvement in quality is beneficial for physiological quantification. This study aimed to evaluate direct PIR for the quantification of tumor hypoxia using the hypoxic fraction (HF) assessed from immunohistological data as a physiological reference. Sixteen mice with xenografted human squamous cell carcinomas were scanned with dynamic [18F]FMISO PET. Afterward, tumors were sliced and stained with H&E and the hypoxia marker pimonidazole. The hypoxic signal was segmented using k-means clustering and HF was specified as the ratio of the hypoxic area over the viable tumor area. The parametric Patlak slope images were obtained by indirect voxel-wise modeling on reconstructed images using filtered back projection and ordered-subset expectation maximization (OSEM) and by direct PIR (e.g., parametric-OSEM, POSEM). The mean and maximum Patlak slopes of the tumor area were investigated and compared with HF. POSEM resulted in generally higher correlations between slope and HF among the investigated methods. A strategy for the delineation of the hypoxic tumor volume based on thresholding parametric images at half maximum of the slope is recommended based on the results of this study.

Evaluation of Timepix silicon detector for the detection of 18F positrons

Timepix is an evolving energy and position sensitive pixel detector. It consists of a silicon detector (sensitive layer 300 μm thick) bump-bonded to the Timepix readout chip developed by the Medipix2 collaboration. This study aims to test the feasibility of using the acquired energy and position signals from Timepix for positron imaging. The signals of the commonly used fluorine-18 PET (positron emission tomography) tracer [18F]FDG were measured using Timepix operated both in single particle counting (Medipix) and in time over threshold (TOT) modes. The spatial resolution (SR) was measured using the absorber edge method (AEM) and was calculated from the over-sampled line spread function. The track of a positron in the Timepix detector was characterized as a cluster and the energy weighted centroid of each cluster was considered as readout for the position of the positron incidence. The measurement results were compared with theoretical predictions using Monte-Carlo simulations. In addition, imaging of a tissue slice of a mouse heart was analysed with reference to standard phosphor plate imaging. Our results show that the SR was improved from 177.1±4.1 μm (centroid without energy weighting) to 155.5±3.1 μm μm (centroid with energy weighting). About 12% enhancement of SR was achieved with energy information in TOT mode. The sensitivity of Timepix was 0.35 cps/Bq based on the measurements. The measuring background and the ratio between detected positrons and gamma rays were also evaluated and were found to be consistent with theoretical predictions. A small enhancement of image quality was also achieved by applying energy information to the data of the measured tissue sample. Our results show that the inclusion of energy information could slightly enhance the positron measurement compared to without energy information and the Timepix provides a high SR and sensitivity for positron detection. Thus, Timepix is a potentially effective tool for 2D positron imaging.

Automated Whole-Body Bone Lesion Detection for Multiple Myeloma on 68Ga-Pentixafor PET/CT Imaging Using Deep Learning Methods

The identification of bone lesions is crucial in the diagnostic assessment of multiple myeloma (MM). 68Ga-Pentixafor PET/CT can capture the abnormal molecular expression of CXCR-4 in addition to anatomical changes. However, whole-body detection of dozens of lesions on hybrid imaging is tedious and error prone. It is even more difficult to identify lesions with a large heterogeneity. This study employed deep learning methods to automatically combine characteristics of PET and CT for whole-body MM bone lesion detection in a 3D manner. Two convolutional neural networks (CNNs), V-Net and W-Net, were adopted to segment and detect the lesions. The feasibility of deep learning for lesion detection on 68Ga-Pentixafor PET/CT was first verified on digital phantoms generated using realistic PET simulation methods. Then the proposed methods were evaluated on real 68Ga-Pentixafor PET/CT scans of MM patients. The preliminary results showed that deep learning method can leverage multimodal information for spatial feature representation, and W-Net obtained the best result for segmentation and lesion detection. It also outperformed traditional machine learning methods such as random forest classifier (RF), k-Nearest Neighbors (k-NN), and support vector machine (SVM). The proof-of-concept study encourages further development of deep learning approach for MM lesion detection in population study.