Quantitative Evaluation Of Perfusion Characteristics And Immunotherapy Response In Cerebral Glioma By MRI

Background and purpose:Gliomas are the most common types of primary brain tumor in adults.They have high morbidity and mortality.Among them,glioblastoma multiforme(GBM)is the most malignant type,with a median survival time of 14.6 months and a 5-year survival rate less than 10%regardless of aggressive treatment strategies.Gliomas are characterized by enrichment of neovascularization and immature blood vessels,accompanied with inefficient blood perfusion.More heterogeneous vascular structure and perfusion emerge while increase of malignancy in glioma.These perfusion characteristics in glioma are driven by multiple angiogenic modes,as well as promotion of physical and chemical factors in local environment at different stage in tumor progression.On the other hand,different clones of tumor cells in sub-regions of glioma enable heterogeneous expression of pro-and anti-angiogenic factors,which contributes to these perfusion characteristics.Pathologically,vascular structure and blood perfusion has been closely associated with glioma grading,prognosis and treatment resistance.Therefore,it is urgent to accurately and non-invasively evaluate the perfusion characteristics in a quantitative manner in glioma.Conventional magnetic resonance imaging(MRI)can reveal complex signals inside the tumor,including parenchyma,cysts,edema,necrosis,and large blood vessels.In conparison,advanced perfusion MR imaging techniques,including dynamic contrast-enhanced(DCE)MRI,intra-voxel incoherent motion(IVIM)MRI and dynamic susceptibility contrast(DSC)MRI can further reflect and quantify the vascular permeability and blood perfusion inside tumor,which are helpful for diagnosis and monitoring treatment response in clinical.Furthermore,current analytic methods used in MRI image start incorporating the voxel-wise information in the region of interest(ROI),post-processed by complicated mathematical algorithms to achieve feature extraction.Conventional subject quantitative method(i.e.Hot-pot method)measures signal-intensity or parameters within several limited ROIs inside tumor.However,a large amount of spatial information has been neglected.Studies have found that quantitative assessment of tumor physiological characteristics based on voxel-wise analysis strategies of MRI images have stronger diagnostic efficacy than traditional qualitative or subjective methods.Consequentially,the advanced perfusion MR imaging techniques accompanied with quantitative image analysis can provide new insights for individual and precise diagnosis and treatment for glioma patients.Currently,it has been found that conventional imaging techniques and quantitative methods show some limitations in precise diagnosis and treatment supervision of glioma patients.At first,prognosis,clinical management and treatment strategy vary among different grade and histological subtypes of gliomas.However,current diagnostic MRI techniques and quantitative methods are less competent in this field.Furthermore,it is imperative to accurately and non-invasively assessing the expression of important molecules by appropriate imaging approaches since the involvement of molecular characteristics into the criteria of2016 WHO classification of central nervous system(CNS)tumors.In addition,recent studies have applied immunotherapy,especially the application of chimeric antigen receptor(CAR)T cell therapy,to target GBM and found unsatisfying therapeutic effcacy.Currently,response assessment in neuro-oncology(RANO)are the standard evaluation methods in clinical trials,in which the imaging criteria mainly include measuring the tumor size and the changes of T2or fluid attenuated inversion recovery(FLAIR)weighted signal in tumor lesion.Therefore,it is prior to develop sensitive and reliable MRI techniques and supervising strategies to quantitative assess individual therapeutic response of CAR-T cells therapy and undermine the mechanisms of treatments resistance.To this end,our study retrospectively performed histogram and texture analysis on DCE-MRI parameter maps of glioma patients,and tested their efficacy in quantitative assessment of perfusion characteristics,for evaluating the typical molecules expression and glioma grading.On the other hand,we quantitatively evaluated the alteration of perfusion characteristics after CAR-T cells treatment by analyzing the parameters maps of dynamic IVIM-MRI and DCE-MRI.Our findings may provide potential imaging bio-markers of perfusion features,applied in precise pre-surgery diagnosis of glioma and monitoring therapeutic response of CAR-T cell therapy in GBM.Materials and Methods:Part I:(1)Forty-two pathologically confirmed glioma patients were included,and conventional MRI,DCE-MRI images and pathological tissue specimens were obtained.(2)Nine quantitative parameters,including K^trans,Kep,Ve,Vp,etc were obtained after post-processing based on Tofts,Extended Tofts and Patlak pharmaceutical kinetics(PK)model,and three semi-quantitative parameters were acquired from the time-signal intensity curve,including maximum signal intensity(Max SI),initial area undercurve(IAUC)and Wash-in Slope(Max Slope).(3)ROI of the largest area in tumor lesion was independently contoured by four experienced neuro-radiologist.Parameter maps were analysed by Grey-Level co-occurrence matrix(GLCM).Energy,Entropy,Inertia,Correlation,and Inverse difference moment(IDM)were extracted.(4)CD 34 immunohistochemical staining were performed,and microvascular density(MVD)and microvascular area(MVA)were quantified.(5)One-way analysis of variance(ANOVA)and post-hoc multiple comparisons were used to evaluate the differences of textural features among different grades of gliomas.Mann-Whitney U test was used to evaluate the textural features in discriminating histological subtype of gliomas.(6)Inter-class correlation coefficient(ICC)was used to test the credibility among the four independent data collections by four nuero-radiologists.Receiver operating characteristic curve(ROC)were performed to test the diagnostic efficacy of component textural features,which could be discriminating grade III and IV gliomas.(7)Correlation between texture features and pathological vascular perfusion characteristics was evaluated by Spearman rank correlation coefficient.Part II:(1)Thirty-two pathologically confirmed glioma patients were included,and conventional MRI,DCE-MRI images and pathological tissue specimens were obtained.(2)Four vascular perfusion parameters,K^trans,Kep,Ve and Vp were obtained based on Extended Tofts Linear model.(3)ROI of the largest tumor area on conventional MRI images.Histogram analysis were performed to extract features,including 90^th percentile,standard deviation(SD),skewness and maximum relative frequency(m RF)(4)Tissue factor(TF),VEGF,CD 34,CD 105 and?-SMA were determined by immunohistochemical staining.(5)Correlation between DCE-MRI histogram features and pathological vascular parameters were analyzed using Pearson correlation coefficient.One-way ANOVA was used to evaluate the ability of skewness and m RF in evaluating the expression of TF,since the non-Gaussian distribution of these two histogram features.Part III:(1)Overexpression tumor-specific antigen EGFRv III on U87-maligant glioma(MG)cell line and expression of EGFRv III targeting CAR on human peripheral blood T cells expressing CAR targeting EGFRv III were achieved by lentivirus transduction.(2)CAR-T cells were labeled with ultra-small superparamagnetic particles of iron oxide(USPIO).Prussian blue staining,transmission electron microscopy(TEM)and MRI scanning were used to examine the label efficiency in vitro.The ability of labeled CAR-T cells to kill GBM cells in vitro was tested by cytotoxicity and ELISA assays.(3)The CAR-T cells infiltration in GBM was detected by dynamic T2WI and susceptibility weighted imaging(SWI)MRI.confirmative analysis of Prussian blue and CD3staining was conducted on paraffin sections of mice brain.(4)IVIM-MRI and DCE-MRI scanning were performed at baseline,1,3 and 7 days post CAR-T cell injection,along with conventional CE-T1WI and T2WI.Perfusion parameters,includingfast diffusion coefficient(ADC_fast)and fast flowing blood faction(F_fast),were obtained from IVIM-MRI.Meanwhile,reference model based K^trans could be obtained from DCE-MRI.Then the differences of perfusion characteristics between the core and peripheral region were quantitatively assessed in tumor bearing mice,with or without CAR-T cells treatment.(5)Histlogical staining includes Hematoxylin-eosin(H&E)and Prussian blue(PB)staining.Ki-67 and CD 3 were determined by immunohistochemical staining.(6)Differences between groups were evaluated using one way or two-way ANOVA.ResultsPart ?(1)For all four observers,all textural features(Energy,Entropy,Inertia,Correlation,and IDM)of semi-quantitative parameters and Patlak model-based K^trans displayed good potency in differentiating grade II(n=15),grade III(n=13),and grade IV(n=14)gliomas(p<0.05).Energy,Entropy,and IDM of Patlak-based Vp,Extended Tofts-based Vp,Tofts-based K^trans,and Ve showed significant differences among all grades of gliomas(p<0.05).In addition,Energy and Entropy of Extended Tofts-based K^trans,Correlation of Extended Tofts-based Kep,and Energy of Tofts-based Kep also showed significant differences among all grades of gliomas(p<0.05).(2)Post-hoc tests indicated that Entropy of semi-quantitative IAUC and Max SI,Tofts-based Ve,Patlak-based K^trans and Vp showed significant differences between grade II(n=15)from grade III(n=13)gliomas(p<0.01).IDM of model-free parameters IAUC and Patlak-based K^trans and Vp could also significantly discriminate between grade II and grade III gliomas in four measurements(p<0.01).Besides,Entropy and IDM of semi-quantitative parameters Max SI and Max Slope,Extended Tofs-based Vp,and Patlak-based Vp could significantly differentiate grade III gliomas from grade IV gliomas in all measurements(p<0.01).(3)Energy and Entropy of IAUC could differentiate AA from AOD by observer A(p<0.05).No other textural feature could discriminate between any subtypes of grade II and III gliomas in any other measurements(p>0.05).(4)For textural features which showed significant differences between grade III and IV gliomas,diagnostic efficacy was tested.For Entropy of Max SI,the AUCs are 0.813,0.846,0.857,and 0.863 for observer A,B,C,and D,respectively.Sensitivity and specificity of Entropy of Max SI in HGGs discrimination are 78.57%,92.31%for observer A,71.43%,92.31%for observer B,C,and D.In terms of Entropy of Max Slope,AUCs are 0.857,0.797,0.808,and 0.802 for the four observers.Sensitivity and Specificity are 71.43%,92.31%for observer A,92.86%,69.23%for observer B and C,85.71%,69.23%for observer D,respectively.As the results of measurement of quantitative parameters,Entropy of Extended Tofts-based Vp and Patlak-based Vp showed highest AUC,sensitivity and specificity(0.885,85.71%,and 84.62%for observer A,Extended Tofts-based Vp,and for observer B,Patlak-based Vp).IDM of Extended Tofts-based Vp and Patlak-based Vp displayed similar results(AUC,sensitivity,and specificity were 0.901,85.71%and 84.62%for observer A,Extended Toftsbased Vp;and 0.896,92.86%and 84.62%for observer D,Patlak-based Vp)Despite the high AUC,sensitivity,and specificity,Entropy and IDM of Extended Tofts-or Patlak-based Vp revealed relatively lower ICC(0.816 and 0.774 for Entropy,0.803 and 0.810for IDM)than those of Max SI and Max Slope(0.951 and 0.921 for Entropy,0.953 and 0.931for IDM).Comparing the AUCs,no significant difference was found in diagnostic efficiency among the four observers(p>0.05).(5)Eight extracted features that could differentiate between grade III and IV gliomas were correlated with commonly used microvascular perfusion indices.We selectively used the data of observer A.None of textural features positively or negatively correlated with CD34-MVD(p>0.05).In comparison,all of eight features correlated moderately with CD34-MVA(p<0.05).Part ?:(1)Expression of TF showed independence from glioma grade.All low-grade glioma exhibited weak TF expression(100%),as well as one grade III(OA)glioma(14.3%).Five of seven grade III(85.7%)and five of 14(35.7%)grade IV gliomas revealed moderate expression of TF.Nine of 14 grade IV(64.3%)gliomas and one grade III glioma(14.3%)expressed high-level TF.(2)Further,we correlated four generally accepted microvascular perfusion indexes with TF expression.Notably,MVD of CD34 immunoreactive microvessels revealed no significant correlation with TF(r=0.140,p=0.444),while MVA of CD34 indicated moderate correlation(r=0.419,p=0.017).However,both MVD and MVA of CD105 and?-SMA immunoreactive vessels correlated well with TF expression.Both dynamic angiogenic indexes,proliferating capillary index(PCI)and Microvascular pericyte converge index(MPI),revealed strong correlations with TF expression(r=0.798,p<0.001 and r=0.835,p<0.001).(3)Among DCE-MRI histogram features,SD of Ve and Vp showed moderate correlation with TF expression(0.4<r<0.6,p<0.001);Mean and 90^thpercentile of Ve and Vp,Mean,SD and 90^thpercentile of K^trans showed strong correlation with TF expression(r>0.6,p<0.001).(4)Statistics revealed that significant difference existed among different TF expression groups(p<0.001).Bonferroni multiple comparison tests revealed that m RF of weak TF expression glioma was significant different from that of other two subgroups(p<0.001)Skewness data also showed significant difference in a same fashion.Part ?:(1)Flow-cytometry assay indicated that 64.35%percent of CAR-T cells were successfully transduced with EGFRv III targeting CAR.(2)With 37.5 ug/m L Fe concentration of UPSIO,in vitro labeling exerts good efficiency and bio-compatibility,and has no negative effect on killing efficacy of CAR-T cells to tumor cells.(3)SWI-MRI showed accumulation of hypo-intensity inside tumor parenchyma 3 day after adoptive transfer of USPIO labeled CAR-T cells into tumor bearing mice.Histopathological staining reveal co-localization of CAR-T(CD 3⁺)cells and iron particles(Prussian blue⁺).(4)Tumor size of CAR-T cells treatment group was significant smaller than control group from day 7 to 14(p<0.05).the survival of CAR-T cells treated mice were significantly prolonged(p<0.05).DWI-MRI indicated that ADC value was significantly elevated in treatment group 3-day post CAR-T cells injection(p<0.05).Immunohistochemistry indicated that expression of Ki-67 index was significantly lower than control group on day 3(p<0.05),accompanied with local necrosis and inflammatory cells infiltration.(5)Dynamic IVIM-MRI scanning indicated that both ADC_fast and F_fast,either in core or periphery site,showed no significant changes during from baseline to day 7 in control group(p>0.05).At each given point,ADC_fastof periphery site were significantly higher than core site(p<0.05).In treatment group,ADC_fastand F_fast showed slight increase,but no significance was detected(p>0.05).In comparison ADC_fastand F_fast showed continuously decrease in coere site(p<0.05).Comparing quantitative perfusion parameters of treatment and control group at each time point,no significant differences were detected in ADC_fastand F_fastvalue in periphery site.However,in core site,ADC_fastand F_fastvalue of treatment group were significantly lower than those of control group from day 1 to day 7(p<0.05).(6)Dynamic DCE-MRI scanning indicated that K^transvalue of both core and periphery site showed no significant difference in control group from day 0 to day 3(p>0.05).However,on day 7,K^transvalue of both core and periphery showed significant increase,compared with previous time points(p<0.05).In treatment group,K^transvalue of periphery sites showed significant increase in on day 7(p<0.05),with no significant deference detected from baseline to day 3(p<0.05),K^transvalue of core site showed significant decrease on each time point after CAR-T cells injection.However,no significant difference was found among K^transvalue of core site on day 1 to day 7(p>0.05).Comparison of K^transvalue of core and periphery sites,we found K^transvalue of core site in treatment group significantly lower than that in control group(p<0.05).In terms of K^transvalue of periphery site,no significant differences were found between treatment and control group,from day 0 to day 7(p>0.05).While K^transvalue of periphery site in control group was significant higher than that of treatment group on day 7(p<0.05).ConclusionIn this study,we find that texture and histogram analysis of the multiple DCE-MRI parameters maps not only could effectively and accurately quantify the perfusion characteristics in glioma,but also pre-operatively grading glioma and evaluate the expression of typical angiogenic molecules.In addition,quantitative perfusion assessment at the early stage of CAR-T cells treatment could reveal the dynamic perfusion characteristics changes at different region of GBM.Our study may provide potential imaging biomarkers for individual and precise diagnosis and therapeutic response assessment in gliomas.