Imaging of Esophageal Lymph Node Metastases by …Integrated Systems and Technologies Imaging of Esophageal Lymph Node Metastases by Desorption Electrospray Ionization Mass Spectrometry

Integrated Systems and Technologies

Imaging of Esophageal Lymph Node Metastasesby Desorption Electrospray Ionization MassSpectrometryNima Abbassi-Ghadi1, Ottmar Golf2, Sacheen Kumar1, Stefan Antonowicz1,JamesS.McKenzie2, JuzhengHuang1,NicoleStrittmatter2,HiromiKudo3, EmrysA. Jones2,Kirill Veselkov2, Robert Goldin3, Zoltan Takats2, and George B. Hanna1

Abstract

Histopathological assessment of lymph node metastases(LNM) depends on subjective analysis of cellular morphologywith inter-/intraobserver variability. In this study, LNM fromesophageal adenocarcinoma was objectively detected usingdesorption electrospray ionization-mass spectrometry imaging(DESI-MSI). Ninety lymph nodes (LN) and their primarytumor biopsies from 11 esophago-gastrectomy specimenswere examined and analyzed by DESI-MSI. Images from massspectrometry and corresponding histology were coregisteredand analyzed using multivariate statistical tools. The MSIsrevealed consistent lipidomic profiles of individual tissuetypes found within LNs. Spatial mapping of the profilesshowed identical distribution patterns as per the tissue types

in matched IHC images. Lipidomic profile comparisons ofLNM versus the primary tumor revealed a close association incontrast to benign LN tissue types. This similarity was used forthe objective prediction of LNM in mass spectrometry imagesutilizing the average lipidomic profile of esophageal adeno-carcinoma. The multivariate statistical algorithm developedfor LNM identification demonstrated a sensitivity, specificity,positive predictive value, and negative predictive value of89.5%, 100%, 100%, and 97.2%, respectively, when com-pared with gold-standard IHC. DESI-MSI has the potentialto be a diagnostic tool for perioperative identification of LNMand compares favorably with techniques currently used byhistopathology experts. Cancer Res; 76(19); 5647–56. �2016 AACR.

IntroductionLymphatic spread is the most common route of cancer dis-

semination (1, 2). The presence of metastases in regional lymphnodes (LN) indicates cancer progression and is an importantprognostic factor for long-term survival (3–6). Evaluation oflymph node metastases (LNM) forms an integral component ofcancer staging systems (7, 8), which are routinely used for deci-sion-making in treatment.

Curative therapy for the majority of carcinomas consists ofcomplete surgical resection of the primary tumor and anappropriate regional LN clearance. The sentinel LN is consid-ered the gatekeeper to regional LNs, and is likely to be the firstto harbor metastases (9–11). The concept of the sentinel LNhas been used intraoperatively to guide the extent of LN

clearance in several tumor sites, including breast, thyroid,melanoma, esophageal, and gastric cancer (12–15).It is iden-tified at time of surgery and assessed for the presence of tumor;the histopathological evaluation of which guides the surgeonin deciding whether to proceed with clearance of the rest of theregional LNs.

Intraoperative histopathological analyses of LNs using fro-zen section or touch imprint cytology have sensitivities of 73%and 63% and specificities of 100% and 99% for the identifi-cation of LNM, respectively, when compared with the goldstandard paraffin-embedded tissue-based analysis as deter-mined by meta-analyses (16, 17). Histopathological assess-ment relies on subjective analysis of cellular morphologyresulting in inter- and intraobserver variability (18, 19).Although immuno-histochemistry of paraffin-embedded tissuehas been shown to improve sensitivity for occult metastases inhistopathological analysis, it has a lengthier processing timeand does not suit intraoperative assessment. The presence ofLNM in paraffin sections of sentinel LNs, wrongly diagnosed asnegative for cancer during intraoperative assessment, maynecessitate the patient to have another surgery to removeregional LNs, resulting in patient distress, possible complica-tions, and cost implications for healthcare providers.

Several nonhistopathological techniques have been used toassess LN status intraoperatively. The use of one-step nucleic acidamplification (OSNA) assay (20, 21), qRT-PCR (22), and tech-nologies such as photo-acoustic tomographic imaging (23),hand-held PET probe (24), and Raman spectroscopy (25) havebeen reported in this context. The only method deemed suitablefor clinical application is the OSNA assay in breast cancer, which

1Department of Surgery and Cancer, Imperial College London, 10thFloor QEQM Wing, St Mary's Hospital, London, United Kingdom.2Department of Surgery and Cancer, Imperial College London, 6thFloor Sir Alexander Fleming Building, South Kensington Campus,London, United Kingdom. 3Centre for Pathology, Imperial CollegeLondon, 4th Floor ClarenceWing, St Mary's Hospital, London, UnitedKingdom.

Note: Supplementary data for this article are available at Cancer ResearchOnline (http://cancerres.aacrjournals.org/).

Corresponding Author: Zoltan Takats, Imperial College London, 371 Sir Alex-ander Fleming Building, South Kensington Campus, London SW7 2AZ, UK.Phone: 442075942760; Fax: 442075943226; E-mail: [email protected]

doi: 10.1158/0008-5472.CAN-16-0699

�2016 American Association for Cancer Research.

CancerResearch

www.aacrjournals.org 5647

on June 6, 2020. © 2016 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

Published OnlineFirst June 30, 2016; DOI: 10.1158/0008-5472.CAN-16-0699

http://cancerres.aacrjournals.org/

relies on the molecular identification of cytokeratin-19 (CK19)mRNA. However, when processing a whole LN using the OSNAassay, several important features including size, location, andpattern of LNM as well as extracapsular extension are lost. Inaddition, no residual LN tissue remains after OSNA assay shouldfurther analysis be required.

Novel methods of histological classification, based on thechemical organization of tissue samples, usingmass spectrometryimaging (MSI), have gained increasing interest. Tumor versusnormal surrounding tissue have been demonstrated with falsecolor images representative of specific chemical profiles (26, 27)and in close correlation to the matched histopathological image.Commonly used MSI techniques include matrix-assisted laserdesorption/ionization (MALDI), secondary ion mass spectrome-try (SIMS), and desorption electrospray ionization-MSI (DESI-MSI; refs. 28, 29). DESI-MSI is an ideal technology for thedevelopment of targeted objective histology workflows, such asLNM identification, as it requires minimal sample preparation asdirect tissue analysis can be performed under ambient conditions(30). Also, DESI-MSI is a nondestructive technique allowingfurther histopathological analysis if needed.

This study reports the application of DESI-MSI for the identi-fication of esophageal adenocarcinoma (EA) LNM. Comparativeanalysis of the lipidomic profiles of the primary tumor andmetastases versus benign LN tissue was carried out. The similar-ities in lipidomic profiles between primary tumor and LNM wasutilized to identify LNM based on an average primary tumorprofile using multivariate methods. The proposed workflow forintraoperative analysis would involve the retrieval of the LNfollowed by MSI and an integrated protocol for the objectiveidentificationofmetastases basedon its lipid profile, independentof expert histopathological opinion.

Materials and MethodsClinical specimens and patient selection

Approval for the study was obtained from the institutionalethics review committee and Imperial College Healthcare TissueBank (project no. R14120). Consecutive patients undergoingtransthoracic esophago-gastrectomy for EA were recruited intothis study. Exclusion criteria included patients with esophagealsquamous cell carcinoma, malignancy associated with any other

site in the body, liver disease, and patients with signs/symptomsof acute infection.

Manual LNdissectionwas performed immediately after retriev-al of the surgical specimen. The harvested LNswere divided in twoalong their long axis. For each LN, one half was sent for routinehistopathological examination and the other half was snap frozenin liquid nitrogen for MSI. Incision biopsies were also taken fromthe primary tumor for the same purposes.

Sample preparation, DESI-MSI, and reference testCryo-sectioning and DESI-MSI were performed in line with

methods described in our previous study (31).

Histopathological assessment: reference testThe same cohort of LNs used for DESI-MSI were subsequent-

ly stained with hematoxylin and eosin (H&E) and also immu-nostained with the anti-cytokeratin AE1/AE3 antibody (DakoLtd.) to detect the presence of LNM. The samples of primarytumor were only stained with H&E. Digital images wereacquired for bioinformatics analysis using a high-resolutionslide scanner (NanoZoomer2.0-HT, C9600-13 Hamamatsu).Two histopathologists, specializing in esophago-gastric cancer,blinded to the results of the DESI-MSI, assessed the LN sectionsfor the presence of macrometastases (>2 mm) and micro-metastases (0.2 to <2 mm; refs. 32, 33) using standard brightfield microscopy. The identification of LNM was aided by thered/brown staining of the AE1/AE3 anti-cytokeratin stain.Anydisagreement in the results were resolved by a third histopa-thologist. The location of the metastases were mapped onto thedigital images.

Data analysisRefer to Supplementary Materials and Methods for full

methodological details of data analysis with respect to: tis-sue-specific mass spectra extraction; data preprocessing (massrange selection, peak alignment, normalization, de-noising,data averaging); multivariate statistical models; glyceropho-spholipid (GPL) annotation; and individual GPL comparisonbetween tissue types. Tissue-specific mass spectra extraction wasperformed from the full dataset, for the following tissue types:LN parenchyma (LNP), carbon deposits, LN connective tissue

Table 1. Demographics and case characteristics of lymph node sample set

Lymph nodes harvested for MSI

DemographicsHistopathological classification as per

the IHC reference test Histopathology of full resection

Patient ID Age Sex CTxTotal number

of LNs Normal LN Micro-Met Macro-Met LNMR T stage N stage PN invasion LV invasion

TO22 68 M Yes 11 2þ1 ITC 2 7 0 3 2 Yes YesTO26 56 F Yes 8 8 0 0 0 4a 2 Yes YesTO28 78 M No 1 1 0 0 0 1a 1 Yes NoTO29 59 M Yes 5 5 0 0 0 3 1 Yes YesTO30 80 M Yes 14 14 0 0 0 3 1 Yes NoTO31 62 M Yes 9 8 0 1 1a 3 2 Yes NoTO32 57 M Yes 12 7 1 4 0 3 2 Yes YesTO39 38 M Yes 10 9 0 0 1 3 1 Yes NoTO40 50 M Yes 9 4 2 2 1 4a 3 Yes YesTO41 66 M No 7 5 0 0 0 1b 0 No NoTO42 55 F Yes 4 2 0 0 2 3 0 Yes Yes

Abbreviations: CTx, chemotherapy; LV, lymphovascular invasion; Macro-Met, macrometastases; Micro-Met, micrometastases; N-stage, classified as 1 to 3 as per 7thedition of TNM; T stage, classified as 1 to 4b as per 7th edition of TNM staging.aLNMR was present in the same LN, which contained the macrometastases.

Abbassi-Ghadi et al.

Cancer Res; 76(19) October 1, 2016 Cancer Research5648




(LNC), fat, micrometastases (0.2–2 mm), macrometastases (>2mm), metastases with response to chemotherapy (LNMR), andprimary tumor, for the purpose of multivariable analysis.

Data processing for spatial prediction of metastases. For the trainingset, all mass spectra of the nonmalignant LN tissue types (LNP,LNC, carbon deposits, fat) were combined into one class, theprimary tumor (excludingmetastases) into a second class, and theglass slide background as a third class.

The full dataset consisted of 11 samples of primary tumor(one from each patient) and the respective LNs from each tumorspecimen. A variable number of LNS were retrieved for eachprimary tumor, resulting in a total of 90 LNs. Leave-one-patient-out cross-validation was performed in the followingmanner. A training set was created by excluding a primary tumorand its respective LNs from one of the 11 sample sets, whereas theLN samples of the excluded patient represented the test set.Therefore, data from the same patient was not present in thetraining and test sets at the same time. This had been carried outiteratively for each of the patients to complete the leave-one-patient-out cross-validation process.

The m/z values of the training set classes were combined to acommonm/z vector by an in-housepeak-matching algorithm. Them/z values of the test sets were then matched to those of thecommon m/z vector of the training set to enable multivariatecomparison between the two sets. Each mass spectrum wasnormalized to its total ion current by dividing each peak intensityof a mass spectrum by the sum of all peak intensities of the samespectrum. m/z values with zero median intensities for all histo-logical classes were considered noise and thus excluded. Variancestabilizing normalizationwas carriedout by log transformationofthe data (34).

Multivariate classification of the test set pixels was based on acombination of linear discriminant analysis with recursivemaximum margin criterion previously described in detail(35). The primary aim was to classify each pixel (mass spec-trum) of the test set as either healthy, tumor, or glass slide.Because the test set samples do not contain primary tumor, butonly metastases from the primary tumor, the classification ofthe metastases mass spectra of the test set were expected tosucceed based on molecular ion patterns of the primary tumormass spectra.

Component 1 (45.6%)

LNMR

Lymph node connec�ve �ssueLymph node parenchymaMetastases

A

E

D

CB

F

GPC1

PC2

Com

pone

nt 2

(23.

8%)

Log

fold

cha

nge

rela

�ve

to th

e m

ean

profi

le a

cros

s all

sam

ples

Inte

nsity

(a.u

.)

Data

poi

nts

600

500

400

300

200

100

Figure 1.

DESI-MSI of lymph node with macrometastases and metastases response to chemotherapy (LNMR). A, AE1/AE3 IHC image of the LN section post DESI-MSI.B, RMMC (supervised multivariate analysis) pixel classification image: metastases (red), LNMR (orange), lymph node connective tissue (green), lymph nodeparenchyma (blue).C,RMMCscores plot of tissue-specificmass spectra in the600 to 1000m/z range.D,PC 1 score image.E,PC2 score image.F, representativemassspectra of tissue types. G, heatmap showing log2 fold changes relative to the mean profile across all samples.

DESI-MSI of Esophageal Lymph Node Metastases

www.aacrjournals.org Cancer Res; 76(19) October 1, 2016 5649




Multivariate comparison of the two sets was carried out byprojecting the test set mass spectra into the multivariatespace of the training set model. This was achieved by mul-tiplication of the test set mass spectra with the training modelweights to obtain the score values of the test set. The test setscore values were then translated to probabilities by means ofmultinomial logistic regression. If the highest relative class-membership probability of a mass spectrum (pixel) wasbelow 70% (the sum of all class probabilities are 100%),then the pixel was considered an outlier; otherwise it waslabeled as the class with the highest probability. Because thetraining set only contained the three classes (i.e., primarytumor, healthy tissue and glass slide), the pixels of the testsamples were color-coded based on their classification resultinto either cyan (healthy), red (metastasis based on primarytumor molecular ion patterns), white (glass slide), or grey(outlier). The color coding of each pixel on a test set creates atissue class prediction image (TCPI), which allows a spatiallocation/tissue type accuracy comparison with the matchedIHC images of the reference test.

Comparison of outcomes between the IHC reference test and MSItissue classification prediction images. To compare the MSI-basedtissue classification method to the IHC reference test, the thresh-old for a positive diagnosis of metastases on the pixel-wise TCPIshad to be determined. Because the size definition of micro-metastases is 0.2 to 2 mm in its largest dimension, an equivalentvalue was sought with respect to the imaging resolution of thepixels. In this dataset, each pixel represented 75 mm in its lateraldimension and therefore a minimum of three adjacent red pixelswas defined as a criterion for the positive identification of micro-metastases. Anything less than this preanalysis threshold wasconsidered normal LN tissue. This analysis was carried out bytwo independent assessors and disagreements were resolved by athird assessor. The binary outcome of the objective histology(metastases vs. no metastases) was compared with the binaryoutcome of the IHC reference test (metastases vs. no metastases)bymeans of a contingency table. Any LNs identified with isolatedtumor cells (ITCs, <0.2 mm, n ¼ 1) on IHC were classified asnormal LNs due to poor identification accuracies and lack ofclinical significance of this cell type.

Lymph node connec�ve �ssueLymph node parenchymaMicrometastases

A

E

F

G

PC1

PC3

Com

pone

nt 2

(5.2

%)

Component 1 (70.9%)

Log

fold

cha

nge

rela

�ve

to th

e m

ean

profi

le a

cros

s all

sam

ples

D

Inte

nsity

(a.u

.)Da

ta p

oint

s

B C

Figure 2.

DESI-MSI of lymphnodewithmicrometastases.A,AE1/AE3 IHC image of the LN section postDESI-MSI.B,RMMC (supervisedmultivariate analysis) pixel classificationimage: metastases (red), lymph node connective tissue (green), lymph node parenchyma (blue). C,RMMC scores plot of mass spectra in the 600 to 1000m/z range.D, zoomed in image of micrometastases. E, PC 3 score image. F, representative mass spectra of tissue types. G, heatmap showing log2 fold changes relative to themean profile across all samples.






ResultsA total of 90 LNs were harvested from 11 patients with EA.

All LNs and primary tumor samples were included in theanalysis with no exclusions. The demographics and case char-acteristics of the 11 patients and their respective LN harvest areshown in Table 1.

The 90 LNs composed of 65 normal LNs (15 containing carbondeposits from smoking), 5 LNs with micrometastases, 13 LNswithmacrometastases, one LNwith ITCs, 4 LNMR, and 1 LNwithevidence of both macrometastases and LNMR. Successful DESI-MSIwas performed on all LNs and the 11 primary tumor samples.The LN sections subjected to DESI-MSI, were then successfullystained with AE1/AE3 anti-cytokeratin antibody for the purposeof comparative analysis.

Intrasample comparison of LN tissue typesFigures 1A–G and 2A–G show the intrasample comparison of

the lipidomic profile of specific tissue types found withinindividual LNs. Figure 1 shows a single LN with LNP, LNC,macrometastases, and LNMR tissue types. The mass spectralprofiles (600–1000 m/z) of the tissue classes are distinct anddemonstrated in the RMMC scores plot (refer to SupplementaryMaterials and Methods for full methodological details). Theprincipal component (PC) 1 scores image differentiates thebenign LNP and LNC from the malignant macrometastases and

LNMR. The PC 2 scores image further differentiates the LNMRfrom the other LN tissue types in the same spatial orientation asper the IHC image. Figure 2 shows an LN with micrometastases,which is differentiated from the benign LN tissue types on thePC 3 scores image. This demonstrates the resolution limits ofthe instrument in detecting mass spectral changes consistentwith micrometastases. Further examples of individual LN anal-ysis including an LN with carbon deposits and an LN withmacrometastases is provided in Supplementary Figs. S1 and S2,respectively.

Intersample comparison of LN tissue typesIntersample comparison was performed to interrogate the

lipidomic profile of EA primary tumor (n ¼ 11) and itsrespective LNM (n ¼ 19). PCA did not demonstrate distinctclustering of the two tissue types, suggesting that the massspectral profiles are indistinct (Fig. 3A and B). Of the 203 GPLsidentified in the 600 to 1000 m/z range, only two were foundto be significantly different between the LNM and the primarytumor (Supplementary Table S1). The mean relative abun-dance � SD of PG (40:6) was 31.6 � 24.8 and 63.2 � 24.0in primary tumor and metastases, respectively (log2 mean Fc,1.00; q ¼ 0.0265). The mean relative abundance � SD of PG(38:6) was 23.1 � 21.7 and 69.2 � 42.1 in primary tumor andmetastases, respectively (log2 mean Fc, 1.58; q ¼ 0.0306).

Component 1 (20.0%)

B

Component 1 (20.3%)

Component 2 (16.9%)

Com

pone

nt 3

(16.

4%)

C D

Com

pone

nt 2

(14.

3%)

Component 2 (16.2%)

Inte

nsity

(a.u

.)In

tens

ity (a

.u.)

A

PCA

PCA

LNMR

LNM

LNMPrimary tumor

m/z

m/z

Figure 3.

Multivariate analyses comparing lipidomic profiles (m/z 600 to 1000) of LNM (red) versus primary tumor (green) and LNM versus LNMR (blue). A–C, PC analysisscores plot (each point is the average of multiple mass spectra representative of that tissue type within a single sample from one patient). B–D, representativemass spectra of LNM, primary tumor, and LNMR.






Intersample comparison of LNM (n ¼ 19) versus metastaseswith LNMR (n ¼ 5) is shown in Fig. 3C and D. PCA demon-strated separation of the tissue classes and leave one outinternal cross-validation, with Mahalanobis distance classifier,demonstrated a classification accuracy of 100% for metastasesand 80% for LNMR (Supplementary Fig. S3). The overall datasuggest that primary tumor and LNM share similar lipidomicprofiles and that LNMR are distinct from LNM.

PCs analysis comparing lipidomic profiles (m/z 600–1000) ofspecific nonmalignant LN tissue types (LNP, LNC, fat, carbon, andLNMR), LNM, and primary tumor is shown in Fig. 4A–D. The firstPCA scores plot (Fig. 4A) demonstrates a separation of themalignant tissue types (primary tumor and LNM) from thenonmalignant. This effect is demonstrated clearly by the secondPCA scores plot (Fig. 4C) where nonmalignant LN tissue types areclassified into one class. Leave-one-sample-out internal cross-validation (Fig. 5A–E), with Mahalanobis distance classifier,demonstrated a classification accuracy of 96.7% for malignanttissue and 95.3% for nonmalignant (Fig. 5D). ROC curve analysisdemonstrated an AUCof 0.996 (Fig. 5E). The 215GPLs comparedby means of statistical analysis in this data set are shown inSupplementary Table S2. With this proven association we wereable to successfully implement our multivariate algorithm for theprediction of LNM based on an average primary tumor lipidomicprofile of EA.

Spatial prediction of LNM based on primary tumorlipidomic profiles

The contingency table in Supplementary Table S3demonstratesthe full outcomes of our comparative analysis between the IHCreference test and the MSI TCPI. In summary, the TCPIs had anoverall sensitivity of 89.5%, specificity of 91.5%, positive predic-tive value of 73.9%, negative predictive value of 97.0%, and anaccuracy of 91.1% in diagnosing LNM when compared with thegold standard IHC. An example ofmatched IHC and TCPI imagesfor four different types of LNs is shown in Fig. 6. The spatialdistribution of the metastases is comparable between the twodiagnostic modalities. Matched images for each of the 90 LNs areshown in Supplementary Fig. S4. The single LN, which wasclassified as ITC on IHC had the presence of single/double redpixels on the TCPI. All LNs with the presence of LNMR weresuccessfully classified as normal LNs on the TCPIs.

The main discrepancy in agreement between the referenceand reference test was the presence of micrometastases on theedge of six TCPIs but not on IHC. The IHC images do showpositive AE1/AE3 staining in corresponding locations of thesesix LNS; however, the histopathologists attributed this tononspecific edge effect of the sample. If we are also to excludemicrometastases found exclusively on the edge of TCPIs,corresponding to the LN capsule, the results from the contin-gency table are markedly different (Supplementary Table S4).

Component 2 (12.3%)Comp. 1(21.8%)

Com

pone

nt 3

(10.

0%)

Component 1 (20.5%)

Com

pone

nt 2

(13

.7%

)

A

C

B

D

LNMLNPLNCFatCarbon

LNMR

LNMPrimary Tumor

Primary Tumor

Nonmalignant LN tissue

PCA

PCA

m/z

m/z

Figure 4.

Unsupervisedmultivariate analyses comparing lipidomic profiles (m/z600 to 1000)of specific nonmalignant lymphnode tissue typesversusmetastases andprimarytumor. A, PCs analysis scores plot (each point is the average of multiple mass spectra representative of that tissue type within a single sample from one patient).B, average mass spectral profiles of specific tissue types in full dataset. C, PCs analysis scores plot comparing lipidomic profiles of nonmalignant lymph nodetissue types (grouped as one class) versus metastases and primary tumor. D, average mass spectral profiles of nonmalignant LN tissue, metastases, andprimary tumor.






In this case, the TCPIs would have an overall sensitivity of89.5%, specificity of 100%, positive predictive value of 100%,negative predictive value of 97.2%, and an accuracy of 97.7%in diagnosing LNM when compared with the gold standardIHC. As metastases propagate in the subcapsular sinus of LNs,we would not expect to find them in the capsule, unless theyare macrometastases extending out from the parenchyma ofthe LN.

DiscussionIn this study, the capability of DESI-MSI in identifying

specific LN tissue types with accurate spatial distribution andlocalization and without reliance on target-specific reagents (e.g., antibodies) has been demonstrated. The quality of the DESI-MSI was suitable for the distinction of micrometastases, whichare defined by their maximum diameter of 0.2 to 2 mm. Overalllipidomic profile analysis revealed a close association betweenthe primary tumor and metastases in comparison to benign LNtissue types. This association was used for the objective iden-

tification of LN micro- and macrometastases, based on thespatially resolved lipidomic profile of the tumor, with a sen-sitivity of 89.5%, specificity of 100%, positive predictive valueof 100%, and negative predictive value of 97.2% when com-pared with the gold-standard IHC.

DESI-MSI may have a complementary role to histopathologythat may overcome the limitations encountered in currentpractice. When the pathologist is unable to make a definitivediagnosis based on frozen section, DESI-MSI would have asupportive role allowing a definitive intraoperative diagnosis,in the same manner that special staining with IHC supportsparaffin embedded tissue diagnosis. Other potential applica-tions for cancer diagnosis that would benefit from TCPI includeidentification of tumor type in metastases of unknown origin,accurate profiling of samples from fine needle aspiration, oridentification of tissue in histopathologically equivocal cases.In addition, it can be used for postoperative specimen analysis,which often requires extensive resources and manpower toidentify metastases in resected regional LNs. For instance, withesophageal cancer specimens, the number of resected LNs may

Component 1 (19.8%) Component 1

Com

pone

nt 2

Component 1 (20.4%)

Com

pone

nt 2

(14.

1%)

True

cla

ssBA C

Predicted class

3.3%, 1

4.7%, 3

96.7%, 29

95.3%, 61

Com

pone

nt 2

(0.0

%)

True

pos

i�ve

rate

False posi�ve rate

ROC curve, AUC = 0.996

D

E

MalignantNonmalignantIncorrectly classified into malignant groupIncorrectly classified into Nonmalignant group

Figure 5.

Internal cross-validation—lipidomic profile of malignant tissue (tumor and LNM) versus nonmalignant LN tissue types (LNP, LNC, fat, carbon). A, PCs analysisscores plot (each point is the average of multiple mass spectra representative of that tissue type within a single sample from one patient). B, RMMC scores plot.C, leave-one-sample-out cross-validated RMMC scores plot as per confusion matrix. D, confusion matrix of leave-one-sample-out internal cross-validationwith Mahalanobis distance classifier. E, receiver operating characteristic curve.






approach three figures. An automated method may streamlinethe process and improve efficiency.

The following workflow illustrates a potential position of thistechnology in the cancer care pathway. A lipidomic profile of theprimary tumor could be obtained preoperatively through tissuebiopsy or from a spectral database of that particular tumor type.This data, together with database entries obtained by the analysisof healthy LN tissues of other patients, are then used to create amultivariate statistical model using the described algorithms forcreating TCPI. The SLN can be sampled at the start of an operationand processed by a technician in an automated workflow as thesurgeon continues to operate.Once the surgeon is informed aboutthe results, he/she can make a decision regarding completeresection of the regional LNs. This method would objectify theprocess of LNM identification and reduce the burden on histo-pathologists as they do not need to be present at the time of theexamination. Furthermore, analysis is not restricted to a singletumor type, which is the case for OSNA.

Previous studies report the use of other ionization techni-ques for the purpose of MSI of LNM. The use of MALDI-MS hasbeen demonstrated in studies of patients with melanoma andbreast cancer (36, 37). The MSI generated by MALDI primarilydistinguishes areas of varying protein intensities, differentiat-ing LNM from normal lymphocyte regions. Other groups ofcompounds such as trace elements, detected by laser ablation-inductively coupled plasma-mass spectrometry, have also beenused to differentiate areas of metastases within LNs (38). DESI-MSI depends on similar principles of molecular differentia-tion, relying primarily on the detection of varying lipid signalsin the 600 to 1,000 m/z range. In comparison to other MSItechniques, DESI offers certain advantages with respect toclinical applications. It does not require matrix deposition(unlike MALDI) and because it works with full sensitivityunder atmospheric conditions, it has the potential to be usedin an operating theatre with a portable mass spectrometer. Verysimilar lipidomic spectral features can be determined by tech-nologies such as rapid evaporation ionization mass spectrom-etry (REIMS; refs. 39, 40), for in vivo tissue identification ofdifferent types of LN/organ metastases. However, ex vivo anal-ysis with DESI-MSI may be more suitable for the identificationof LNM, as mass spectra can be obtained from discrete areas ata microscopic rather than a macroscopic level. Furthermore,results can be corroborated by histopathological assessment

postoperatively, as the tissue sections remain structurally unal-tered by DESI-MSI.

Several studies comparing LNM with the primary tumor haveutilized `omic disciplines aside from metabolomic/lipidomicprofiling. Perou and colleagues (41) demonstrated that the geno-mic profile of two LNMs were similar to that of multiple primarybreast tumor samples taken from the samepatient. A further studyof 26 patients with breast cancer used a hierarchical clusteringmethod to group samples of primary tumor and paired metas-tases; it found that 92.3% of the cases clustered next to each other,indicating that their overall gene expression profiles were similar(42). The rate of proliferation of primary tumor and associatedLNM has also been studied in 30 patients with breast cancer, bythe identification of antibody labeled S-phase cells. Primarytumor and regional LNM labeling indices correlated strongly, notbeing influenced by age, level of hormone receptors, tumor size,or number of positive nodes (43). In this study, the lipidomicprofiles of the primary tumor were found to be similar to LNM. Atthe individual lipid level, two polyunsaturated PGs were found tobe upregulated in LNM. The functional outcome of these minordifferences is undetermined but is likely to confer a survivaladvantage to cancer cells propagated in an environment foreignto the primary tumor.

Objective classification of LNM based on the average primarytumor lipidomic profile resulted in an accuracy of 97.7%, whichshould encourage further investigation of this method as a diag-nostic test. The sensitivity and specificity of the technique arecomparable to the current gold standard of IHCwithout sufferingfrom its subjectivity. In addition, the reproducibility of DESI-MSIfor the analysis of human cancer tissue has been reported with acoefficient of variance of 18 � 8% (44).

IHChas historical precedence and has stood the test of time butsuffers from the potential consequences of human error. Thequestion that will remain unanswered is the true accuracy of thelipidomic profiling results of DESI-MSI, which may be com-pounded by the subjective nature of the reference test. Neverthe-less, further independent prospective validation blind studies;pathway mapping and cost-effective analysis are needed to pavethe way for the integration DESI-MSI in clinical practice.

Despite its translational potential, there are currently severallimitations for DESI-MSI. In this study, the spatial resolution was75 mm. This is sufficient for identification of micrometastases,which are defined as aggregationsof cancer cells between0.2 and2

Macrometastases

Micrometastases

Normal LN with evidence of carbon deposits

Metastases with complete response to CTX Figure 6.

Examples of matched IHC and TCPI.






mm in their widest dimension. However, identification of ITCs(approximately 20 mm in diameter) is limited at this level ofresolution. The clinical advantage of identifying ITCs continues tobe debated (45) and is thus not commonplace in the staging ofmost epithelial tumors. Nonetheless, an imaging resolution of 40mmhaspreviously beendescribed in the literature (46), and in ourown institution we have used DESI-MSI at a resolution of 20 mm(Supplementary Fig. S5). With the instrument parameters used inthis research, the time required for imaging of very large LNs wasin the order of a few hours. Further development of this technol-ogy at our institution has improved the scan speed by a 30-foldincrease without depreciation of image quality or information.This means that a 10� 20 mm LN cross-section could be imagedwith a resolution of 75 mm in under 10 minutes (SupplementaryFig. S6). The recent advances in spatial resolution and scan speedcoupled with the diagnostic potential of lipid biomarkers makesthe translational application of DESI-MSI a distinct possibility.

ConclusionDESI-MSI can be utilized for the objective identification of

LNM based on its primary tumor lipidomic profile. The accuracyof 97.7% qualifies this technique as a diagnostic test for theidentificationof LNMindependent of histopathological expertise.The use of lipidomic profiling in this context not only confers theadvantage of analytical stability, but delivers a novel panorama ofbiomarkers that can be interrogated for diagnostic purposes.

Disclosure of Potential Conflicts of InterestZ. Takats is a consultant at Waters Corporation, author reports receiving a

commercial research grant from Waters Corporation, also the author is a

consultant/advisory board member for Waters Corporation. No potentialconflicts of interest were disclosed by the other authors.

Authors' ContributionsConceptionanddesign:N.Abbassi-Ghadi, S. Kumar, S. Antonowicz, R.Goldin,Z. Takats, G.B. HannaDevelopment of methodology: N. Abbassi-Ghadi, O. Golf, J. Huang,N. Strittmatter, H. Kudo, E.A. Jones, K. Veselkov, R. Goldin, G.B. HannaAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.):N. Abbassi-Ghadi, J. Huang, N. Strittmatter, H. Kudo,R. Goldin, Z. TakatsAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis):N. Abbassi-Ghadi, O. Golf, S. Kumar, S. Antonowicz,J.S McKenzie, J. Huang, E.A. Jones, K. Veselkov, Z. Takats, G.B. HannaWriting, review, and/or revision of themanuscript:N.Abbassi-Ghadi,O.Golf,S. Kumar, S. Antonowicz, J.S McKenzie, E.A. Jones, R. Goldin, Z. Takats,G.B. HannaAdministrative, technical, or material support (i.e., reporting or organizingdata, constructing databases): N. Abbassi-Ghadi, N. Strittmatter, H. KudoStudy supervision: Z. TakatsOther (funding researcher conducting the study): Z. Takats, G.B. Hanna

Grant SupportThe authors would like to acknowledge funding from the European Research

Council (DESI-JeDI Imaging Starting Grant; MASSLIP Consolidator Grant), theNational Institute of Health Research (Imperial Biomedical Research Centre),the National Institute of Health Research - Diagnostic Evidence CooperativesLondon and the EU Horizon 2020 metaspace project.

The costs of publication of this articlewere defrayed inpart by the payment ofpage charges. This article must therefore be hereby marked advertisement inaccordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Received March 10, 2016; revised May 13, 2016; accepted May 31, 2016;published OnlineFirst June 30, 2016.

References1. Achen MG, McColl BK, Stacker SA. Focus on lymphangiogenesis in tumor

metastasis. Cancer Cell 2005;7:121–7.2. Nathanson SD. Insights into the mechanisms of lymph node metastasis.

Cancer 2003;98:413–23.3. McGuire WL. Prognostic factors for recurrence and survival in human

breast cancer. Breast Cancer Res Treat 1987;10:5–9.4. Harrison JC, Dean PJ, El-Zeky F, Vander Zwaag R. FromDukes through Jass.

Pathological prognostic indicators in rectal cancer. Hum Pathol 1994;25:498–505.

5. Balch CM, Soong SJ, Gershenwald JE, Thompson JF, Reintgen DS, Casci-nelli N, et al. Prognostic factors analysis of 17,600 melanoma patients:validation of the American Joint Committee on Cancer melanoma stagingsystem. J Clin Oncol 2001;19:3622–34.

6. Jauch KW, Bacha EA, Denecke H, Anthuber M, Schildberg FW. Esoph-ageal carcinoma: prognostic features and comparison between blunttranshiatal dissection and transthoracic resection. Eur J Surg Oncol1992;18:553–62.

7. Edge SB, Byrd DR, Compton CC, Fritz AG, Greene FL, Trotti A. AJCC(American Joint Committee on Cancer) Cancer Staging Manual. 7th ed.New York: Springer; 2010.

8. Astler VB, Coller FA. The prognostic significance of direct extension ofcarcinoma of the colon and rectum. Ann Surg 1954;139:846–52.

9. MortonDL. Technical details of intraoperative lymphaticmapping for earlystage melanoma. Arch Surg 1992;127:392–9.

10. Reintgen D, Cruse CW, Wells K. The orderly progression of melanomanodal metastases. Ann Surg 1994;220:759–67.

11. Fisher B, Fisher ER. The interrelationship of hematogenous and lymphatictumor cell dissemination. Surg Gynecol Obstet 1966;122:791–8.

12. Krag D, Weaver D, Ashikaga T, Moffat F, Klimberg VS, Shriver C, et al.The sentinel node inbreast cancer—amulticenter validation study.NEngl JMed 1998;339:941–6.

13. Morton DL, Thompson JF, Cochran AJ, Mozzillo N, Elashoff R, Essner R,et al. Sentinel-node biopsy or nodal observation in melanoma. N Engl JMed 2006;355:1307–17.

14. Balasubramanian SP, Harrison BJ. Systematic review and meta-analysis ofsentinel node biopsy in thyroid cancer. Br J Surg 2011;98:334–44.

15. Schlag PM, Bembenek A, Schulze T. Sentinel node biopsy in gastrointes-tinal-tract cancer. Eur J Cancer 2004;40:2022–32.

16. Tew K, Irwig L,Matthews A, Crowe P,Macaskill P. Meta-analysis of sentinelnode imprint cytology in breast cancer. Br J Surg 2005;92:1307–17.

17. Liu LC, Lang JE, Lu Y, Roe D, Hwang SE, Ewing CA, et al. Intraoperativefrozen section analysis of sentinel lymph nodes in breast cancerpatients: a meta-analysis and single-institution experience. Cancer2011;117:250–8.

18. Roberts CA, Beitsch PD, Litz CE, Hilton DS, Ewing GE, Clifford E, et al.Interpretive disparity among pathologists in breast sentinel lymph nodeevaluation. Am J Surg 2003;186:324–9.

19. van den BrekelMW, LodderWL, Stel HV, Bloemena E, Leemans CR, van derWaal I. Observer variation in the histopathologic assessment of extranodaltumor spread in lymph node metastases in the neck. Head Neck 2012;34:840–5.

20. TsujimotoM,Nakabayashi K, YoshidomeK, Kaneko T, Iwase T, Akiyama F,et al. One-step nucleic acid amplification for intraoperative detection oflymph node metastasis in breast cancer patients. Clin Cancer Res 2007;13:4807–16.

21. Tamaki Y, Akiyama F, Iwase T, Kaneko T, Tsuda H, Sato K, et al. Moleculardetection of lymph node metastases in breast cancer patients: results of amulticenter trial using the one-step nucleic acid amplification assay. ClinCancer Res 2009;15:2879–84.

22. Nissan A, Jager D, RoystacherM, Prus D, Peretz T, Eisenberg I, et al. RT-PCRassay for the detection ofminimal residual disease in sentinel lymphnodesof breast cancer patients. Br J Cancer 2006;94:681–5.






23. Grootendorst DJ, Jose J, Wouters MW, van Boven H, Van der Hage J, VanLeeuwen TG, et al. First experiences of photoacoustic imaging for detectionofmelanomametastases in resected human lymph nodes. Lasers SurgMed2012;44:541–9.

24. Gonz�alez SJ, Wong J, Gonz�alez L, Brader P, Zakowski M, G€onen M, et al.Novel handheld PET probes provide intraoperative localization of PET-avid lymph nodes. Surg Endosc 2011;25:3214–21.

25. Horsnell J, Stonelake P, Christie-Brown J, Shetty G, Hutchings J, Kendall C,et al. Raman spectroscopy—a new method for the intra-operative assess-ment of axillary lymph nodes. Analyst 2010;135:3042–7.

26. Eberlin LS, Norton I, Dill AL, Golby AJ, Ligon KL, Santagata S, et al.Classifying human brain tumors by lipid imaging withmass spectrometry.Cancer Res 2012;72:645–54.

27. Guenther S,Muirhead LJ, Speller AV, Golf O, Strittmatter N, RamakrishnanR, et al. Spatially resolved metabolic phenotyping of breast cancer bydesorption electrospray ionization mass spectrometry. Cancer Res 2015;75:1828–37.

28. Vickerman JC. Molecular imaging and depth profiling by mass spectrom-etry—SIMS, MALDI, or DESI? Analyst 2011;136:2199–217.

29. Tak�ats Z,Wiseman JM, Gologan B, Cooks RG.Mass spectrometry samplingunder ambient conditionswith desorption electrospray ionization. Science2004;306:471–3.

30. Wiseman JM, Ifa DR, Venter A, Cooks RG. Ambient molecular imaging bydesorption electrospray ionization mass spectrometry. Nat Protoc2008;3:517–24

31. Abbassi-Ghadi N, Veselkov K, Kumar S, Huang J, Jones E, Strittmatter N,et al. Discrimination of lymph node metastases using desorption electro-spray ionisation-mass spectrometry imaging. Chem Commun (Camb)2004;50:3661–4.

32. Greene FL, Page DL, Fleming ID, Fritz A, Balch CM, Haller DG, et al. AJCCcancer staging manual. 6th ed. New York: Springer; 2002.

33. Sobin LH, Wittekind C. UICC TNM classification of malignant tumours,6th ed. New York, John Willey & Sons, Inc; 2002.

34. Veselkov KA, Vingara LK, Masson P, Robinette SL, Want E, Li JV, et al.Optimized preprocessing of ultra-performance liquid chromatography/mass spectrometry urinary metabolic profiles for improved informationrecovery. Anal Chem 2011;83:5864–72.

35. VeselkovKA,Mirnezami R, StrittmatterN,Goldin RD, Kinross J, Speller AV,et al. Chemo-informatic strategy for imaging mass spectrometry-basedhyperspectral profiling of lipid signatures in colorectal cancer. Proc NatlAcad Sci U S A 2014;111:1216–21.

36. HardestyWM,Caprioli RM. In situmolecular imaging of proteins in tissuesusing mass spectrometry. Anal Bioanal Chem 2008;391:899–903.

37. Seeley EH, Caprioli RM. Imaging mass spectrometry: Towards clinicaldiagnostics. Proteomics Clin Appl 2008;2:1435–43.

38. Hare D, Burger F, Austin C, Fryer F, Grimm R, Reedy B, et al. Elementalbio-imaging of melanoma in lymph node biopsies. Analyst 2009;134:450–3.

39. Balog J, Sasi-Szab�o L, Kinross J, Lewis MR, Muirhead LJ, Veselkov K, et al.Intraoperative tissue identificationusing rapid evaporative ionizationmassspectrometry. Sci Transl Med 2013;5:194ra93.

40. Golf O, Strittmatter N, Karancsi T, Pringle SD, Speller AV, Mroz A, et al.Rapid evaporative ionization mass spectrometry imaging platform fordirect mapping from bulk tissue and bacterial growth media. Anal Chem2015;87:2527–34.

41. Perou CM, Sørlie T, Eisen MB, van de Rijn M, Jeffrey SS, Rees CA, et al.Molecular portraits of human breast tumours. Nature 2000;406:747–52.

42. Feng Y, Sun B, Li X, Zhang L, Niu Y, Xiao C, et al. Differentially expressedgenes between primary cancer and paired lymph node metastases predictclinical outcome of node-positive breast cancer patients. Breast Cancer ResTreat 2007;103:319–29.

43. Goodson WH3rd, Ljung BM, Moore DH2nd, Mayall B, Waldman FM,Chew K, et al. Tumor labeling indices of primary breast cancers and theirregional lymph node metastases. Cancer 1993;71:3914–9.

44. Abbassi-Ghadi N, Jones EA, Veselkov KA, Huang J, Kumar S, StrittmatterN, et al. Repeatability and reproducibility of desorption electrosprayionization-mass spectrometry (DESI-MS) for the imaging analysis ofhuman cancer tissue: a gateway for clinical applications. Anal Methods2015;7:71–80.

45. Patani N, Mokbel K. Clinical significance of sentinel lymph node isolatedtumour cells in breast cancer. Breast Cancer Res Treat. 2011;127:325–34.

46. Vilmos K, Van Berkel GJ. Improved imaging resolution in desorptionelectrospray ionization mass spectrometry. Rapid Commun Mass Spec-trom 2008;22:2639–44.






2016;76:5647-5656. Published OnlineFirst June 30, 2016.Cancer Res Nima Abbassi-Ghadi, Ottmar Golf, Sacheen Kumar, et al. Electrospray Ionization Mass SpectrometryImaging of Esophageal Lymph Node Metastases by Desorption

Updated version

10.1158/0008-5472.CAN-16-0699doi:

Access the most recent version of this article at:

Material

Supplementary

http://cancerres.aacrjournals.org/content/suppl/2016/06/30/0008-5472.CAN-16-0699.DC1

Access the most recent supplemental material at:

Cited articles

http://cancerres.aacrjournals.org/content/76/19/5647.full#ref-list-1

This article cites 43 articles, 8 of which you can access for free at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected]

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://cancerres.aacrjournals.org/content/76/19/5647To request permission to re-use all or part of this article, use this link



http://cancerres.aacrjournals.org/lookup/doi/10.1158/0008-5472.CAN-16-0699

http://cancerres.aacrjournals.org/content/suppl/2016/06/30/0008-5472.CAN-16-0699.DC1

http://cancerres.aacrjournals.org/content/76/19/5647.full#ref-list-1

http://cancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://cancerres.aacrjournals.org/content/76/19/5647


Documents

Imaging of Esophageal Lymph Node Metastases by …Integrated Systems and Technologies Imaging of Esophageal Lymph Node Metastases by Desorption Electrospray Ionization Mass Spectrometry