DIFFERENT LEVELS IN BLOOD CELL SAMPLES OF EMT-MARKERS FOR THE DIAGNOSIS OF CANCER, IN PARTICULAR OF COLORECTAL

(CRC) AND PANCREATIC (PC) CANCER

Field of the invention

The present invention refers to the field of molecular markers for disease diagnosis, preferably for cancer diagnosis. The present invention is aimed at detecting and measuring mRNA levels of genes involved in epithelial to mesenchymal transition in biological samples, i.e. in peripheral blood samples of tumor patients, to determine the presence of disease, its progression and risk of recurrence.

Background

Circulating tumor cells (CTCs) are an heterogeneous population of cancer cells circulating in the peripheral blood that have been shed from either a primary tumor or its metastasis. Currently, the raw number and corresponding phenotype of CTCs in the whole blood of cancer patients has clinical relevance with respect to patient prognosis. CTC pool includes cells with epithelial, mesenchymal and sternness-like features. It has been shown that cancer intravasation in humans (Min Yu et al., Science, 2013, 339, 580) and animals is coupled with epithelial to mesenchymal transition (EMT), a process driven by transcription factors (TFs; like but not limited to TWIST1, ZEB1, ZEB2, SLUG, SNAIL, PRXX1, etc) and likely reversible (mesenchymal-to- epithelial transition, MET) at metastatic niches and ensuing in modification of cellular epithelial and/or mesenchymal markers (as e.g. CDH1 and Plastin 3). All said transcription factors/genes are involved in epithelial to mesenchymal transition (EMT-markers).

Therefore, the detection of EMT-markers in human blood is expected to provide diagnostic and prognostic clues.

Celesti et al. (Gastroenterology 2013, 145, 647) showed that levels of TWIST1 mRNA were statistically (median value) significantly higher in blood samples from patients with Colo-Rectal Cancer (CRC) than from controls. Unfortunately the distribution of TWIST1 levels in patients and controls partially overlaps, rendering the measurement of TWIST1 mRNA alone not reliable and suitable for diagnosis. Hung Pham et al. (Pancreas 2010, 39, 3, 332-339) suggests that the overexpression of SLUG mRNA in pancreatic cancer tissue samples is correlated to the down regulation of PDGH, resulting in enhancehd PGE2 production. However, the document doesn't teach that the measurement of

SLUG mRNA alone in blood samples is reliable and suitable for diagnosis of pancreatic cancer.

Description of the invention

Conventional, FDA-approved methods (i.e. Cell Search" (Veridex), "Adna Test Breast Cancer Select" and "Adna test Breast Cancer Detect" (AdnaGen AG),

"Biocept Onco CEE" (Biocept Laboratories) detect epithelial antigens expressed, among others, by CTCs. Hence the diagnostic and prognostic value of these tests is limited.

The authors show quantitative RT-PCR assay specific for epithelial- mesenchymal transition (EMT) genes present in circulating cells of tumor patients. Differently from what reported by prior art, no prior separation of circulating cell populations is needed prior to mRNA extraction. This approach allows detecting increased levels of circulating EMT-TF transcripts, irrespective of any antigenic or phenotypic cell feature.

Authors found that high levels of TWIST1, SLUG, and low levels of ZEB1 mRNAs are present in blood cells of CRC patients, whereas high levels of SLUG, TWIST 1 and ZEB2 mRNA are present in blood cells of pancreatic cancer (PC) patients, thus allowing a more reliable and robust diagnosis.

It is therefore an object of the present invention a method for determining if a subject is affected by a cancer comprising the step of assaying a blood cell sample from said subject for the presence of a panel of mRNAs comprising at least SLUG and TWIST 1 mRNA wherein:

a) the increase of mRNA levels of both TWIST1 and SLUG genes but not of ZEB2 gene with respect to control samples is indicative of a colorectal cancer, b) the increase of mRNA levels of all of TWIST1 , SLUG and ZEB2 genes with respect to control samples is indicative of a pancreatic cancer, c) the increase of at least SLUG mRNA level with respect to a first SLUG cut-off is indicative of colorectal or pancreatic cancer,

d) in subjects wherein the SLUG mRNA level is not increased with respect to said first SLUG cut-off, the increase of TWIST1 mRNA level with respect to a TWIST1 cut-off is indicative of colorectal cancer, provided that said subject also shows an increase of SLUG mRNA with respect to a second SLUG cut off, said second SLUG cut-off being lower than the first SLUG cut off, and/or a decrease of CDH1 mRNA level with respect to a CDH1 cut-off,

e) in subjects wherein the SLUG mRNA level is not increased, the increase of TWIST1 mRNA level with respect to a TWIST1 cut-off is indicative of pancreatic cancer, provided that said subject also shows an increase of ZEB2 mRNA level.

In the above method control samples are samples from normal subject or patients with different cancers respect to PC or CRC.

Preferably, in the method according to the invention the increase of both SLUG and TWIST1 mRNA levels with respect to respective cut-offs is indicative of colorectal or pancreatic cancer.

Preferably, in the method according to the invention, in subject wherein SLUG mRNA level is increased with respect to a first SLUG cut-off and TWIST1 mRNA level is not increased with respect to a TWIST1 cut-off, an increase of ZEB2 mRNA level with respect to a ZEB2 cut off is indicative of pancreatic cancer.

Preferably, in the method according to the invention, in subject wherein SLUG mRNA level is increased with respect to a first SLUG cut-off and TWIST1 mRNA level is not increased with respect to a TWIST1 cut-off, a decrease of CDH1 mRNA level with respect to a CDH1 cut-off, is indicative of colorectal cancer.

Preferably, in the method according to the present invention relevant cut-offs are as follows :

- first SLUG cut off for colorectal cancer 7.93018E-9,

- second SLUG cut off for colorectal cancer 1 .84E-10,

- first SLUG cut off for pancreatic cancer 3.27E-9, - TWIST1 cut off for colorectal cancer 1 .00725E-8,

- TWIST1 cut off for pancreatic cancer 1 .35E-8,

- CDH1 cut off: 7.29E-8,

- ZEB2 cut off for pancreatic cancer: 7.19E-6.

Another object of the invention is a method for discriminating between colorectal and pancreatic cancer among subjects that result to be positive to the above method comprising the step of assaying a blood cell sample from said subjects, wherein the increase of any one of ZEB1 and/or ZEB2 and/or CDH1 mRNAs with respect to a proper control from a patient with colorectal cancer is indicative of a pancreatic cancer.

Preferably, in the abov method ZEB2 mRNA level is increased with respect to a cut off of 4.08E-5 and/or the CDH1 mRNA increased with respect to a cut off of 1 .022E-7.

Another object of the invention is a method for determining the stage of a colorectal cancer in an affected subject comprising the step of assaying a blood cell sample from said subject wherein:

a) the increase of ZEB1 mRNA level with respect to control samples from subjects with colorectal cancer is indicative of a less advanced stage of colorectal cancer and/or

b) the decrease of CDH1 mRNA levels with respect to control samples from subjects with colorectal cancer is indicative of metastatic (i.e. stage IV) disease at diagnosis, and/or

c) the increase of TWIST1 mRNA levels with respect to control samples from subjects with colorectal cancer, is indicative of the development of a metachronous metastasis.

In the above method for determining the stage of a colorectal cancer in an affected subject, in the step a) a control sample may be a sample obtained from a subject known stage of colorectal cancer; in the step b) a control sample may be a sample obtained from a subject with colorectal cancer but without metastatic lesions; in the step c) a control sample may be a sample obtained from a subject with colorectal cancer who did not develop metastatic progression.

In the methods according to the invention said blood cell sample is preferably a Circulating Tumor Cell (CTC) enriched cell sample.

Preferably, said Circulating Tumor Cell (CTC) enriched cell sample is a Peripheral Blood Mononuclear Cell (PBMC) sample.

In the methods according to the invention the levels of mRNAs are preferably obtained by RT-PCR.

Any other method to detect specific mRNA in a blood sample known to the expert in the art are within the scope of the instant invention. Illustrative examples are: PCR amplification methods (QX200™ Droplet Digital™ PCR System, TaqMan Probes), other amplification methods; up to single cell gene expression analysis (miRGE - nCounter®).

The method of the invention is able to detect EMT transcripts in blood samples of humans or animals; in a particular aspect the method comprises the steps of :

a) Ficoll gradient density separation of Peripheral Blood Mononuclear Cells (PBMC),

b) detection of the expression levels of EMT genes using RT-PCR,

c) comparison of the level of the EMT genes from the patient to normal control levels,

d) diagnosing the patient as having a specific tumor if the detected levels of the EMT genes are statistically significantly different (higher or lower than a predetermined cut off value depending on the selected gene) than the control level.

Another object of the invention is the use of a quantitative RT-PCR kit for working the methods as above described:

- retrotranscribing means to get specific cDNAs;

- specific amplification probes for amplifying specific cDNAs;

- detecting means to detect and measure the level of amplified specific cDNAs.

Preferably said amplification probes are able to amplify: - the nt. 781 -835 region of TWIST1 , Acc. No.: NM_000474.3, (SEQ ID No.

1 ) ;

- the nt. 730-830 region of SLUG, Acc. No.: NM_003068.4, (SEQ ID No.

2) ;

- the nt. 1262-1361 region of ZEB2 Acc. No.: NM_014795.3, (SEQ ID No.

3) ;

- the nt. 41 13-4172 region of CDH1 , Acc. No.: NM_004360.3, (SEQ ID No. 5);

- the nt. 2917-3020 region of ZEB1 , Acc. No.: NM_001 128128.2, (SEQ ID No. 6).

Preferably, said amplification probes have essentially the sequences:

TWIST1 -Forward AGCAAGATTCAGACCCTCAAGCT (SEQ ID No. 12);

TWIST1 -Reverse CCTGGTAGAGGAAGTCGATGTACCT (SEQ ID No. 13); SLUG-Forward TGTTTGCAAGATCTGCGGC (SEQ ID No. 14);

SLUG-Reverse TGCAGTGAGGGCAAGAAAAA (SEQ ID No. 15);

ZEB2-Forward GCTACACGTTTGCCTACCGC (SEQ ID No. 16);

ZEB2-Reverse CGATTACCTGCTCCTTGGGTT (SEQ ID No. 17);

CDH1 -Forward G G AACTATG AAAAGTG G GCTTG (SEQ ID No. 20);

CDH1 - Reverse AAATTGCCAGGCTCAATGAC (SEQ ID No. 21 );

ZEB1 -Forward GAAAGTGATCCAGCCAAATGG (SEQ ID No. 22);

ZEB1 -Reverse TGGGCGGTGTAGAATCAGAGT(SEQ ID No. 23).

Detailed description of the invention

The invention shall be described with reference to non-limitative examples. Figure legends

Figure 1. mRNA levels of TWIST1 (A), SLUG (B), ZEB1 (C), in blood of 69 CRC patients as compared to 30 healthy controls.

Figure 2. ROC curves for TWIST1 (A), SLUG (B) mRNA levels in 69 patients with CRC as compared to 30 healthy controls.

Figure 3. Algorithm for diagnosis of CRC based upon the circulating levels of both TWIST1 and SLUG mRNAs (two-steps).

Figure 4. Algorithm for diagnosis of CRC based upon the circulating levels of TWIST1-SLUG plus CDH1 mRNAs (two-steps).

Figure 5. Circulating mRNA levels of ZEB1 in patients with stage I and II CRC (pT1 -4 NO). Figure 6. Circulating mRNA levels of ZEB1 in patients with stage II (pT3 NO) and III (pT3 N1 -2) CRC.

Figure 7. Circulating mRNA levels of CDH1 in patients without (MO) and with (M+) metastatic disease at diagnosis.

Figure 8. High TWIST1 mRNA levels in circulating blood cells are associated with CRC metastatic progression (disease free survival - DFS- by Kaplan- Meier curves).

Figure 9. TWIST1, SLUG and ZEB2 mRNA levels in the blood of PC patients and healthy controls.

Figure 10. ROC curve of SLUG mRNA levels discriminating PC patients (n=24) from healthy controls (n=30).

Figure 11. Algorithm for diagnosis of PC based upon the circulating levels of TWIST1-SLUG plus ZEB2 mRNAs (two-steps).

Figure 12. Different circulating mRNA levels of ZEB1, ZEB2, and CDH1 in patients with PC and CRC.

Materials and Methods

Quantification of Gene Transcripts in unselected blood samples

Peripheral blood (6ml_) was collected in anticoagulants (EDTA, sodium citrate, heparin)-coated vacutainer and stored at 4°C. Peripheral blood was processed within 4 hours of collection and Ficoll-Paque Plus (GE Healthcare, Life Science) gradient separated according to the manufacturer's instructions.

Briefly, 15 mL of Ficoll-Paque was added to a centrifuge tube and diluted PBS (6 mL+29 mL of balanced salt solution) was carefully layered on Ficoll-plaque.

The unmixed solution was centrifuged (400 g for 40 minutes at 20°C). Using a clean Pasteur pipette the upper layer was removed leaving untouched the lymphocyte layer at the interface. Next, by a new Pasteur pipette the PBMC layer (containing circulating tumor cells) was transferred to a clean centrifuge tube and washed with at least 3 volumes (18 ml) of balanced salt solution.

Once re-suspended, cells were centrifuged (300 g for 20 minutes at 20°C), and the supernatant was then removed. After adding 50 mL of balanced salt solution cells pellet was finally centrifuged (200 g for 20 minutes at 20°C). The supernatant was again removed and cells lysed with Qiagen lysis buffer plus β-Mercaptoethanol (1 :100). Total RNA was then isolated using Qiagen

Rneasy-Mini kit according to manufacturer's instructions. Thereafter, total

RNA re-suspended in 60 μΙ_ of dethylphyrocarbonate-treated water (DEPC- water) was treated by DNAse (Ambion, Life Science) to minimize the contamination by genomic DNA.

All RNA preparation and handling steps took place in a laminar flow hood, under RNA-free conditions. RNA concentration was determined by absorbance reading at 260 nm using nanodrop.

Two μg of treated RNA were reverse transcripted to cDNA using High

Capacity cDNA Reverse Transcription Kit (Applied Biosystem, Life Science).

Synthesized cDNA was subjected to quantitative RT-PCR to detect and quantify EMT-gene mRNA levels:

TWIST1: NCBI Acc. No. NM_000474.3 (SEQ ID No.1 ),

SLUG or SNAI2: NCBI Acc. No. NM_003068.4 (SEQ ID No.2),

ZEB2 or SIPI: NCBI Acc. No. NM_014795.3 (SEQ ID No.3),

SNAIL: NCBI Acc. No. NM_005985.3 (SEQ ID No.4),

CDH1 : NCBI Acc. No.: NM_004360.3 (SEQ ID No. 5),

ZEB1 : NCBI Acc. No.: NM_001 128128.2 (SEQ ID No. 6),

PRRX1 : NCBI Acc. No.: NM_006902.4 (SEQ ID No. 7),

PLASTIN3: NCBI Acc. No.: NM_005032.6 (SEQ ID No. 8).

In brief, 1 μΙ of cDNA (40ng) was placed in 20 μί of reaction volume containing 12 μΙ of Fast SyberGreen Master Mix, 3 μΙ of housekeeping and target Forward and Reverse Primers mixed at 5μΜ and 4 μΙ of DEPC-water. Specific primer sequences were:

18s-Forward CGC CGC TAG AGG TGA AAT TCT (SEQ ID No. 9),

18s-Reverse CTT TCG CTC TGG TCC GTC TT (SEQ ID No. 10);

to amplify the nt. 1049-1 100 region of 18s, Acc. No.: M10098.1 , NCBI, (SEQ ID No. 1 1 ), as control;

TWIST1 -Forward AGC AAG ATT CAG ACC CTC AAG CT (SEQ ID No. 12); TWIST1 -Reverse CCT GGT AGA GGA AGT CGA TGT ACC T (SEQ ID No. 13);

to amplify the nt. 781 -835 region of TWIST1 , Acc. No.: NM_000474.3, (SEQ ID No. 1 );

SLUG-Forward TGT TTG CAA GAT CTG CGG C (SEQ ID No. 14);

SLUG-Reverse TGC AGT GAG GGC AAG AAA AA (SEQ ID No. 15);

to amplify the nt. 730-830 region of SLUG, Acc. No.: NM_003068.4, (SEQ ID No. 2);

ZEB2-Forward GCT ACA CGT TTG CCT ACC GC (SEQ ID No. 16);

ZEB2-Reverse CGA TTA CCT GCT CCT TGG GTT (SEQ ID No. 17);

to amplify the nt. 1262-1361 region of ZEB2 Acc. No.: NM_014795.3, (SEQ ID No. 3);

SNAIL-Forward CTT CCA GCA GCC CTA CGA C, (SEQ ID No. 18);

SNAIL Reverse CGG TGG GGT TGA GGA TCT(SEQ ID No. 19);

to amplify the nt. 174-244 region of SNAIL, Acc. No.: NM_005985.3 (SEQ ID No. 4).

CDH1 -Forward G G AACTATG AAAAGTG G GCTTG (SEQ ID No. 20);

CDH1 - Reverse AAATTGCCAGGCTCAATGAC (SEQ ID No. 21 );

to amplify the nt. 41 13-4172 region of CDH1 , Acc. No.: NM_004360.3, (SEQ ID No. 5);

ZEB1 -Forward GAAAGTGATCCAGCCAAATGG (SEQ ID No. 22);

ZEB1 -Reverse TGGGCGGTGTAGAATCAGAGT (SEQ ID No. 23);

to amplify the nt. 2917-3020 region of ZEB1 , Acc. No.: NM_001 128128.2,

(SEQ ID No. 6);

PRRX1 -Forward ACACTATCCTGATGCTTT TGTG (SEQ ID No. 24);

PRRX1 -Reverse GAACTTGGCTCTTCGGTTC (SEQ ID No. 25);

to amplify the nt. 395-494 region of PRRX1 , Acc. No.: NM_006902.4 (SEQ ID No. 7);

PLASTIN3-Forward CCTTCCGTAACTGGATGAACTC (SEQ ID No. 26);

PLASTIN3-Reverse GGATGCTTCCCTAATTCAACAG (SEQ ID No. 27) to amplify the nt. 1624-1837 region of PLASTIN3, Acc. No.: NM_005032.6, (SEQ ID No. 8).

Amplification was performed in an ABI 7900 HT Real Time PCR system (Applied Biosystem) using the program: 95°C for 10 minutes; 40 cycles of 95°C for 15s and 60°C for 60s. All samples were analyzed in triplicate.

Quantification of target genes and internal reference gene 18s was performed using the fluorescence emission of SyberGreen. DNA contamination was assessed by performing PCR on the no-reverse transcribed portion of each sample. For all samples fluorescence was detected after 0-40 cycles for the control and marker genes in a single reaction, which allow for the deduction of the cycles at threshold (CT) value for each product. The CT value was considered as the PCR cycle at which a significant increase in fluorescence is detected due to the exponential accumulation of double-strand PCR products. Expression of the target genes was normalized on the expression of the housekeeping gene, 18s to obtain the absolute quantification (2 ^"Δ ° ^Τ ), while the relative quantification was calculated by 2 ^~ΔΔ ° ^Τ method to compare the transcript levels of controls versus patients (Livak Kj, Methods,

2001 ;25(4):402-8).

Statistical analysis

Mann-Whitney U-test was used to asses statistically significant differences in the expression levels of the circulating EMT-TF mRNAs between cancer patients and controls. For the EMT-TFs with significantly higher levels in CRC and Pancreatic cancer patients, sensitivity and specificity were estimated using the optimum cut-off points determined by ROC curves analysis. A p- value of <0.05 was deemed to be of statistical significance. All statistical analysis was conducted using StatDirect software (StatsDirect Ltd, Altrincham CHESHIRE, UK).

For each proposed diagnostic marker the following have been assessed:

• Specificity (also called the true positive rate) measures the proportion of actual positives which are correctly identified as such. Specificity relates to the test's ability to exclude a condition correctly.

• Sensitivity (also called the true negative rate) measures the proportion of negatives which are correctly identified as such. Sensitivity relates to the test's ability to identify a condition correctly.

• The diagnostic odds ratio is a measure of the effectiveness of a

diagnostic test. It is defined as the ratio of the odds of the test being positive if the subject has a disease relative to the odds of the test being positive if the subject does not have the disease.

• The positive and negative predictive values (PPV and NPV

respectively) are the proportions of positive and negative results in statistics and diagnostic tests that are true positive and true negative results.

• The false positive rate is the expectancy of the false positive ratio, which refers to the probability of falsely rejecting the null hypothesis for a particular test

• The false negative rate is the rate of occurrence of negative test

results in subjects known to have the disease or behavior for which an individual is being tested.

• The False Omission Rate is the chance of not satisfying the null hypothesis among those that accept the null hypothesis

• The False discovery rate is the expected proportion of errors among the rejected hypotheses.

• The positive likelihood ratio (LR+) defines how much increase the probability of disease if the test is positive

• the negative likelihood ratio (LR-) defines how much decrease the probability of disease if the test is negative

• The accuracy is the measure defining how close a test result comes to the true value.

Patients and Methods

The pilot study included 69 patients with CRC, and 24 patients with pancreatic cancer, plus 30 healthy control subjects. All samples were processed as above described after obtaining patient informed consent.

Results

1. High levels of TWIST1, SLUG, , and low levels of ZEB1 mRNAs are present in blood cells of CRC patients In healthy controls (n=30), coding mRNAs for ZEB1, ZEB2, CDH1, and SNAIL were detected in all samples (100%). TWIST1 and SLUG mRNAs were detected in 26 (87%) and 1 1 (36.7%) controls, respectively. Plastin3 mRNA was detected in 15 (50%), and PRRX1 mRNA in 4 (13%) samples.

In blood samples from 69 patients with CRC, authors detected TWIST1 (Fisher test vs controls p=0.007), ZEB2 and SNAIL mRNAs in all (100%) cases, and SLUG in 66 (95.7%; Fisher test vs controls p<0.001 ) cases. CDH1 was detectable in 65 out of 66 (98.5%) tested samples. In addition, in 57 tested samples, ZEB1 expression was detectable in 55 (96.5%), Plastin3 in 44 (77.2%), and PRRX1 in 3 (5.3%).

As Shown in Figure 1 , CRC patients had higher circulating levels of TWIST1 (Figure 1A, absolute quantification, median, controls 9.93E-9 vs CRC 3.22 E- 8; p<0.001 ) and SLUG (Figure 1 B, controls 9.09E-13 vs CRC, 1 .83E-08; p<0.001 ) mRNAs. Conversely, ZEB1 levels were lower in CRC patients than in controls (Figure 1 C, controls 5.64E-7 vs CRC, 1 .66E-07; p=0.02). The blood levels of ZEB2, SNAIL, CDH1, Plastin, PRRX1 mRNAs did not differ between cases and controls. Accordingly, as compared to controls, the levels of TWIST1 in blood cells of cancer patients were 2.3 times higher, those of SLUG 386 times higher, and those of ZEB1 5.1 times lower (relative quantification).

Threshold values of TWIST 1 and SLUG mRNAs in blood cells for identifying CRC patients.

Authors next determined the optimum cut-off point for high levels of transcripts discriminating CRC patients by ROC curve analysis (Figure 2). By this approach, high levels of TWIST1 (above the cut-off, 1 .01 E-8) had 94.2% sensitivity and 50% specificity (Table 1 and Figure 2A), and high SLUG levels (above the cut-off, 7.93E-9) had 76.8% sensitivity and 90% specificity in discriminating CRC patients from healthy controls (Table 2 and Figure 2B). Table 1. Diagnostic yields, sensitivity and specificity, for colorectal cancer of circulating levels of TWIST1 mRNA above the optimum cut-off values. Oencltort

Prevalence

Condfot} posifm (CRC) Caidtion NegaSwe {CTR}

ea7

m

Test outcome positive 66 15

Test outcome t aw

FOR NPV

Test outcome negatwe 4 15 0210526316 0,79 m TPR, Sensilivity FPR Fatout

1.884057971 0M 0.500 HOC

m ffft TNR.Specifc* 0.81

0.1 15942Q29 0.05S 05

OCR

163

PPV=Positive Predictive Value FPR=False positive rate

FDR=False discovery rate FNR=False negative rate

FOR=False omission rate TNR=True negative rate

NPV= Negative Predictive Value LR+=Positive likelihood ratio

ACC= Accuracy LR-=Negative likelihood ratio

TPR= True positive rate DOR= Diagnostic odds ratio

Table 2. Diagnostic yields, sensitivity and specificity, for colorectal cancer of circulating levels of SLUG mRNA above the optimum cut-off values.

Pwalorw

Concftar po&f ICR Coneffon Negatim (CTRf

0222 m

OCR

as

PPV=Positive Predictive Value FPR=False positive rate

FDR=False discovery rate FNR=False negative rate

FOR=False omission rate TNR=True negative rate

NPV= Negative Predictive Value LR+=Positive likelihood ratio

ACC= Accuracy LR-=Negative likelihood ratio

TPR= True positive rate DOR= Diagnostic odds ratio

Combining circulating levels of TWIST1 and SLUG (with or without CDH1) mRNAs in blood cells for identifying CRC patients. To discriminate CRC patients from healthy individuals we exploited TWIST1 sensitivity in conjunction with SLUG specificity. By mean of this approach we first determined the specimens with high levels (i.e., above the cut-off) of both TWIST1 and SLUG mRNAs (CRC 49/69, 71 %; controls 2/30, 6.6%) (Figure 3). On the other side, low levels (i.e., below the cut-off) of both TWIST1 and SLUG mRNAs were detected only in controls (14/30, 33.3%). In the remaining samples (20 cases and 14 controls), 16 cases and 13 controls showed high- TWIST1-\ow-SLUG mRNA levels, and 4 cases and 1 control with low- TWIST1-h\gh-SLUG mRNA levels. A second order SLUG ROC curve (cut-off 1 .84E-10) was drawn for these 34 samples. According to the second order cut-off, 17 CRC samples showed high-SLL/G levels and 3 low-SLL/G levels, as compared to 7 controls with high-SLL/G levels and 7 controls with low- SLUG levels. Eventually, the overall sensitivity and specificity of the two-steps TWIST1-SLUG algorithm were 95.7% and 70%, respectively. The positive and negative predictive values were 88% and 87.5%, respectively, with a diagnostic odds ratio of 51 .3 (Table 3).

Alternatively, a second order CDH1 ROC curve (cut-off 7.29E-8) could be drawn for the 34 samples with high-TI/WS7 -low-SLL/G and low- TWIST1- h\gh-SLUG mRNA levels. According to the second order CDH1 cut-off, 14 CRC samples showed \ow-CDH1 levels and 6 high-CDH7 levels, as

compared to 4 controls with low-CDH7 levels and 10 controls with high-CDH7 levels (Figure 4). Accordingly, the alternative algorithm combining TWIST1 and SLUG plus CDH1 levels attained 91 .3% overall sensitivity and 80% specificity. The positive and negative predictive values were 91 % and 80%, respectively, with a diagnostic odds ratio of 42 (Table 4). The combination of two or more markers enhances the diagnostic performance of a single marker by matching substantially similar sensitivity with increased specificity, so that the rate of false positive tests is minimized in relationship to the rate of false negative ones, ensuing in improved diagnostic odds ratio and accuracy.

Table 3. Diagnostic yields, sensitivity and specificity, for colorectal cancer of circulating levels of TWIST1 and SLUG mRNAs above the optimum cut-off values, according to diagnostic algorithm 1 . Corrtton

P'Ovafcnco

CantStion positive (CRC) Ctticifef} Negate (CTR)

»,?%

PW FCR

Test outcome positive 65 9

Test outcome 0,88 0,12

FOB NF¥

Test outcome negative a

0,12 088 m TPR,S«siiw|i FPR. Fa*om

319 0957 030C ADO

» R* TNR, Spedfiaty 088

0,082 ao« i.7

∞R<PUGN0SroO∞SRAT!Q)

511

PPV=Positive Predictive Value FPR=False positive rate

FDR=False discovery rate FNR=False negative rate

FOR=False omission rate TNR=True negative rate

NPV= Negative Predictive Value LR+=Positive likelihood ratio

ACC= Accuracy LR-=Negative likelihood ratio

TPR= True positive rate DOR= Diagnostic odds ratio

Table 4. Diagnostic yields, sensitivity and specificity, for colorectal cancer of circulating levels of TWIST1,SLUG plus CDH1 mRNAs above the optimum cut-off values, according to diagnostic algorithm 2.

ConcHon

Prevalence

Condion postSve (CR iit n Ne aiim fCTOj

®,7

PV FOR

Test outcome positive m

Testsutoffis i 6

0L91 0.09

WW

Test outcome negative 6 J a

02 06

TPR. SensitMty Β*β, Pa§«jt

4.57 ftii 0..3 AX

m TNR, Specificity 0JB8

αο9 080

PPV=Positive Predictive Value FPR=False positive rate

FDR=False discovery rate FNR=False negative rate

FOR=False omission rate TNR=True negative rate

NPV= Negative Predictive Value LR+=Positive likelihood ratio

ACC= Accuracy LR-=Negative likelihood ratio

TPR= True positive rate DOR= Diagnostic odds ratio

Circulating levels of ZEB1 and CDH1 mRNA in blood cells differ according to TNM features of cancer at diagnosis.

As to local invasion, the levels of ZEB1 mRNA were significantly lower in patients with pT1 -T2 NOMO (i.e., stage I) CRCs than in patients with pT3-T4 N0M0 (i.e., stage II) CRCs (Median, stage I 1 .14E-7 vs stage II 4.51 E-7, overall p=0.001 ) (Figure 5). As to lymph-node metastasis, patients with CRC showing the same depth of local invasion (i.e. pT3) had higher circulating levels of ZEB1 mRNA levels in the absence of nodal invasion (i.e., NO) than when nodal invasion was present at pathological examination (N1 -N2, Median, NO 6.72E-7 vs N1 -2 2.66E-7; p=0.02) (Figure 6). As to the presence of distant metastasis, patients with metastatic disease (M+) at diagnosis, displayed significantly lower circulating levels of CDH1 mRNA than those without metastatic lesions (MO; Median, M+ 2.38E-8 vs MO 5.88E-8, p=0.02) (Figure 7).

Circulating levels of TWIST1 mRNA in blood cells discriminate progression to metastasis over time. In patients without distant CRC metastasis (stage l-lll) at the time of diagnosis (n=54), circulating levels of TWIST1 mRNA were higher (>3.07E-8) in 7 out of 8 (87.5%) patients who later developed post-surgical metastatic progression. Accordingly, TWIST1 levels discriminated a significantly different disease free survival among patients with CRC (Figure 8).

2. High levels of TWIST 1, SLUG, and ZEB2 mRNAs are present in blood cells of pancreatic cancer (PC) patients.

The levels of EMT-TF mRNA in the blood differ between PC patients and healthy individuals. As Shown in Figure 9, compared to healthy controls, 24 PC patients had higher levels of TWIST1 (Figure 9A, absolute quantification, median, controls 9.935E-9 vs PC 3.41 E-8; p<0.001 ) and of SLUG (Figure 9B, controls 9.09E-13 vs PC, 1 .30E-8; p<0.0001 ), as well as of ZEB2 (Figure 9C, controls 1 .69E-6 vs PC 1 .52E-5; p<0.01 ). Accordingly, the circulating levels of TWIST1, SLUG, and ZEB2 mRNAs in blood of PC patients were on average 14, 788, and 5 times higher, respectively, than in controls (relative quantification).

Threshold values of SLUG mRNA in blood identifying PC patients. By

ROC curve analysis, high levels of SLUG (above the cut-off, 3.27E-9) by themselves reached 100% sensitivity and 70% specificity in discriminating cancer patients from healthy controls, reaching 75% positive and 100% negative predictive values, with a diagnostic odds ratio of 130 (Figure 10 and Table 5).

Combining circulating levels of TWIST 1 and SLUG with ZEB2 mRNAs in blood cells for identifying CRC patients. To discriminate PC patients from healthy individuals we exploited TWIST1 in conjunction with SLUG. By mean of this approach we first determined the specimens with high levels (i.e., above the cut-off) of both TWIST1 (cut-off, 1 .35E-8) and SLUG mRNAs (PC 22/24, 91 .6%; controls 6/30, 20%) (Figure 1 1 ). On the other side, low levels (i.e., below the cut-off) of both TWIST1 and SLUG mRNAs were detected only in controls (14/30, 33.3%). In the remaining samples (2 cases and 10 controls), only 8 controls showed high-TI/WST7-low-SLL/G mRNA levels, and

2 cases and 2 control with low-TI/WST7-high-SLL/G mRNA levels. A second order ZEB2 ROC curve (cut-off 7.19E-8) was drawn for these 12 samples. According to the second order cut-off, 1 PC sample showed high-ZES2 levels and 1 IOW-ZES2 levels, as compared to 7 controls with high-ZES2 levels and

3 controls with low-ZES2 levels. Eventually, the overall sensitivity and specificity of the two-steps TWIST1-SLUG plus ZEB2 algorithm were 95.8% and 76.7%, respectively. The positive and negative predictive values were 76.7% and 95.8%, respectively, with a diagnostic odds ratio of 75.57 (Table 7).

ZEB1, ZEB2, and CDH1 mRNA levels are higher in blood cells of PC patients than in those of CRC patients. In addition, comparing EMT-gene levels in blood between PC and CRC patients, we found that the levels of ZEB1 (PC 1 .14E-6 vs CRC 1 .66E-7, p=0.002), ZEB2 (PC 1 .52 E-5 vs CRC 6.62 E-6, p=0.03 ) and CDH1 (PC 1 .36E-7 vs CRC 5.73E-8, p=0.001 ) were significantly higher in patients with PC than patients with CRC. By relative quantification, the transcript levels of ZES7 were on average 7 times higher in PC than in CRC patients, while ZEB2 and CDH1 were both 2 times higher in PC than in CRC patients (Figure 12).

Threshold values of CDH1 and ZEB2 mRNA levels in blood cells discriminating pancreatic from colorectal cancer. The cut-off value of CDH1 discriminating patients with pancreatic from those with colorectal cancer was 1 .022E-7 yielding 75.4% specificity for pancreatic cancer (17 positive colorectal cancer positive out of 69) and sensitivity 58.3% (detecting 14 out of 24 pancreatic cancer patients). Subsequently, a second order ZEB2 cut-off (4.08E-5) allowed discriminating 5 out of 17 patients with colorectal cancer (ZEB2 values<4.08E-5) who were not detected by mean of CDH1 levels, while detecting 2 out of 10 patients with pancreatic cancer (ZEB2 values>4.08E-5) who were not detected by mean of CDH1 levels. Overall, combined use of CDH1 and ZEB2 mRNA levels led to discriminate patients with pancreatic cancer from those having colorectal cancer with 82% specificity and 66.7 sensitivity.

Table 5. Diagnostic yields, sensitivity and specificity for pancreatic cancer of circulating levels of SLUG mRNA above the optimum cut-off value.

revalence

Condition positive (PC) Condition Negative (CTR)

44,4%

PPV FDR

Test outcome positive 24 8

Test outcome 0,75 0,25

FOR NPV

Test outcome negative 0 22

0 1

LR+ TPR, Sensitivity FPR, Fall-out

3,75 1,000 0,267 ACC

LR- FNR TNR, Specificity 0,85

0 0,000 0,7

DOR

129,7

PPV=Positive Predictive Value FPR=False positive rate

FDR=False discovery rate FNR=False negative rate

FOR=False omission rate TNR=True negative rate

NPV= Negative Predictive Value LR+=Positive likelihood ratio

ACC= Accuracy LR-=Negative likelihood ratio

TPR= True positive rate DOR= Diagnostic odds ratio

Table 6. Diagnostic yields, sensitivity and specificity for pancreatic cancer of circulating levels of TWIST1 mRNA above the cut-off values.

Condition

Prevalence

Condition positive (PC) Condition Negative (CTR)

44,4%

PPV FDR

Test outcome positive 22 14

Test outcome 0,81 0,39

FOR NPV

Test outcome negative 2 16

0,11 0,89

LR+ TPR, Sensitivity FPR, Fall-out

1,964285714 0,917 0,467 ACC

icity 0,70

DOR

12,57

Table 7. Diagnostic yields, sensitivity and specificity for pancreatic cancer of circulating levels of SLUG, TWIST1 and ZEB2 mRNA above the cut-off values.

Condition

Prevalence

Condition positive (PC) Condition Negative (CTR)

44,4%

PPV FDR

Test outcome positive 23 7

Test outcome 0,77 0,23

FOR NPV

Test outcome negative 1 23

0,041666667 0,958333

LR+ TPR, Sensitivity FPR, Fall-out

4,107142857 0,958 0,233 ACC

LR- FNR TNR, Specificity 0,85

0,054347826 0,042 0,767

DOR

75,57

PPV=Positive Predictive Value FPR=False positive rate

FDR=False discovery rate FNR=False negative rate

FOR=False omission rate TNR=True negative rate

NPV= Negative Predictive Value LR+=Positive likelihood ratio

ACC= Accuracy LR-=Negative likelihood ratio

TPR= True positive rate DOR= Diagnostic odds ratio