The invention relates to a chip and a method for head & neck cancer prognosis and further the biomarkers that predict the aggressive subset amongst early squamous cancer of tongue, biomarkers that can predict the aggressive subset amongst early squamous cancer of the buccal mucosa, and an analytical process using micro-array based gene expression data to classify metastatic and non-metastatic Squamous cell carcinoma of the head and neck (HNSCC) tumors to aid in the decision making by a Surgeon to dissect or not to dissect the neck of a patient. In other words, this classification process will allow the surgeon to avoid over-treatment as well as under treatment.
BACKGROUND OF THE INVENTION
Squamous cell carcinoma of the head and neck (HNSCC) being the sixth most common cancer worldwide, poses a significant cause for morbidity and mortality, with approximately 540,000 new cases annually worldwide (Vigneswaran and Williams; Oral MaxillofacSurgClin North Am. 2014 May;26(2): 123-41. doi: 10.1016/j.coms.2014.01.001). HNSCC is the highest incidence of all cancer in South East Asia and in India it comprises 32% to 40% of total malignancy in comparison to western countries where it is only 3% to 5% (Vigneswaran and Williams; Oral MaxillofacSurgClin North Am. 2014 May; 26(2):123-41. doi: 10.1016/j.coms.2014.01.001). Despite the recent advances the 5-year survival rate is approximately 50% due to relatively high recurrence rates in the patients and the development of SPTs in the upper aerodigestive tract (Tabor, Brakenhoff et al.;Clin Cancer Res. 2001 Jun;7(6):1523-32). HNSCC arises from the epithelial mucosal region of the upper aerodigestive tract encompassing regions of oral cavity, larynx and pharynx and has variable etiologies and prognosis. Smoking and alcohol consumption are the foremost risk factors for HNSCC. However, the difference in prevalence in different population is attributable to the indigenous habit of chewing a mixture of tobacco, areca nut, lime betel leaf and spices in variety of combinations (Tabor, Brakenhoff et al.; Clin Cancer Res. 2001 Jun;7(6):1523-32). Human papillomavirus (HPV) also is emerging as an important risk factor for the development of HNSCC.
In India, it is the chewing of tobacco quid, and its placement in the gingivo-buccal groove that is responsible for the high incidence overall and the predilection for buccal mucosa cancer. In the West it is mainly alcohol consumption along with smoking that is responsible for causing oral cancer which affects mainly the floor of the mouth and the tongue.

Oral submucous fibrosis,- a premalignant condition, is a typically Indian disease as it is caused due to the habit of chewing areca nut. A very large number of Indian oral cancer cases have associated submucous fibrosis. This is not so in oral cancers in the Western population.
HNSCC has a tendency to metastasize to the cervical lymph nodes. Presence of metastasis to cervical lymph node has a major adverse impact on the prognosis of oral cancer. There are site wise differences in the pattern of lymph node metastases. Level one lymph node is the first echelon of metastasis from cancer of the buccal mucosa whereas level two is the first echelon of metastasis from cancer of the tongue.
Presence of cervical lymph node metastases (N+ neck) may manifest clinically as a palpable mass or may be demonstrated on imaging and a needle biopsy. In a sizeable number of cases of oral cancer there may be micro metastases that are neither clinically palpable nor are demonstrable on imaging. These cases may pass off as ' N zero ' (NO) though they are truly ' N+ 'cases.
Treatment of oral squamous cancer must necessarily address the issue of management of the neck.
Surgery is the mainstay of treatment for oral cancer. An early primary tumour viz stage T1 (tumour size < 2cms) or stage T2 (tumour size 2-4 ems) is generally resected intra orally. If the neck is N+, clinically or on imaging, a radical or modified radical neck dissection is performed.
Nearly 30% of the T1 /T2, NO (N zero) cases may be harbouring occult cervical lymph node metastases. If the neck is not dissected, the neck disease will manifest within weeks or months as palpable metastatic nodes. Resection at that stage may or may not be possible. If on the other hand, neck dissection is performed prophylactically in all NO cases, 70% of the cases will be subjected to an unnecessary radical surgery with all its costs & morbidity. Several clinico-pathological parameters have been used to try and predict the true 'N ' status in a neck that is 'NO ' clinically and on imaging. These include the tumour size, its depth, the grade of the tumour and the presence or absence of perineural invasion. Neither singly, nor in combination has these factors allowed a high enough predictability for a ' wait & watch ' policy to be pursued safely.

If the neck is "NO" clinically and on imaging, a surgeon is faced with the dilemma on whether to dissect the neck at this stage.
Roepman, Wessels et al. in their publication (Nat Genet. 2005 Feb;37(2):182-6. Epub 2005 Jan 9) and PCT Publication W02006/085746 disclosed a microarray-based study on the European population and reported a gene signature with the predictive strength of 102 differentially expressed genes (those which are up-regulated or down-regulated) for lymph node metastases from 82 primary SCCs of the oral cavity and oropharynx). Significantly, an overall predictive accuracy of 86% was achieved compared with a clinical staging accuracy of 68%. Ropemann et al. (Cancer Res. 2006 Feb 15; 66(4):2361-6) further reported several subsets of an 825-gene panel that are capable of predicting lymph node metastasis in head and neck cancer with equal robustness. A further refinement was reported by Roepman et al. (Cancer Res. 2006 Feb 15;66(4):2361-6) on the predictive genes that enriched two over represented functional categories viz, binding to the extracellular matrix and protease activity for the degradation of the extracellular matrix, supporting the fact that invasion of the tumor cells in the surrounding tissues involve the non-tumor cells in the tumor microenvironment. The initial 102 gene signature developed on a in-house 'customized' array platform and was subsequently subjected to a commercial whole gene expression array for platform transition, followed by the analysis on a dedicated diagnostic array for multicentric validation. Van Hoof SR et al. (Carcinoma Journal of Clinical Oncology 2012 30:33, 4104-4110) reported an analysis that showed an overall accuracy of 72% for the whole validation cohort and an 89% NPV upon combining with clinical assessment.
The 825 genes or subset thereof generated from the HNSCC tumors of patient population in the Netherlands disclosed in the PCT publication W02006/085746 suffer from severe limitations and is not applicable to patients in the Asian subcontinent due to their distinct population specific differences in the genetic architecture of the HNSCC tumors. The gene list could classify only at 72% accuracy for the validation cohort that reaches to 89% upon combining with clinical information. Further, the said gene list is unlikely to successfully classify the HNSCC samples originating from the Asian subcontinent as a major percentage of these population cohorts have a history of exposure to smokeless tobacco, as depicted by the current set of patient volunteers whose tissues were subjected to transcriptomics analysis (Table 1).
Smoking and alcohol consumption are the foremost risk factors for HNSCC. However, the difference in prevalence in different population is attributable to the indigenous habit of chewing a mixture of tobacco, areca nut, lime betel leaf and spices in variety of combinations as

described by Tabor, Brakenhoff et al. in 2001. This is also supported by the fact that there is very little resemblance with list generated from the data obtained using European cohorts versus this work based on Indian patients (Fig. 7).
There is therefore an unmet need to provide biomarkers that predict the aggressive subset amongst early squamous cancer of tongue, biomarkers that can predict the aggressive subset amongst early squamous cancer of the buccal mucosa within the population in the Asian subcontinent and an analytical process using micro-array based gene expression signature to distinguish potentially metastatic from non-metastatic HNSCC tumors to aid in the decision making by a Surgeon whether to undertake a prophylactic radical neck dissection or whether to wait and watch with regular follow up.
OBJECTS OF THE INVENTION
The main object of the invention is to provide a chip, a kit , biomarkers and an in-vitro method for head & neck cancer prognosis.
Another major objective of the invention is to provide biomarkers to aid the surgeon in making evidence based decision, whether to dissect or not to dissect the neck during surgery for an early (T1,/T2, NO) Squamous cancer of the oral cavity thus minimising the risk of under-treatment and avoiding the morbidities of over-treatment.
A major objective of the invention is to provide biomarkers to identify the aggressive subset amongst early HNSCC (T1, T2, NO) so as to assist the head and neck oncologist in prognostication as also in planning appropriate and optimum treatment upfront.
Another object of the invention is to provide biomarkers that predict the aggressive subset amongst early squamous cancer of tongue (T1 /T2) within the population in the Asian subcontinent.
Another object of the invention is to provide biomarkers that can predict the aggressive subset amongst early squamous cancer of the buccal mucosa (T1/T2) within the population in the Asian subcontinent.
Yet another object of the invention is to provide an analytical process using micro-array based gene expression data to classify metastatic and non-metastatic HNSCC tumors that can be used by a surgeon to decide on whether to dissect the neck of a patient.

Yet another object of the invention is to provide a kit for the determination of head-and-neck tumor, comprising a microarray which includes one or more polynucleotides comprising a nucleic acid sequence complementary to the sequence of one or more mRNAs selected from the group arrived at using the said analytical process, capable of measuring the expression level of the mRNA in tumor biopsy materials or body fluids like saliva and blood, so as to yield a comparative expression profile between sample and control to determine presence of tumor and its metastatic potential.
Yet another object of the invention is to provide the said kit for the determination of head-and-neck tumor, including a primer set capable of amplifying the sequence of one or more mRNA in tumor biopsy materials or body fluids like saliva and blood, selected from the group arrived at using the analytical process and a fluorescent probe comprising a polynucleotide consisting of a nucleic acid sequence complementary to the mRNA sequence or a part thereof.
SUMMARY OF THE INVENTION
Accordingly, the present invention relates to a chip for detection of head & neck cancer prognosis consisting of cDNA sequence complimentary to the nucleotide sequence having Seqidno. 1-957.
In an embodiment of the invention it provides a method for microarray-based gene-expression assay conducted on a microarray chip for quantifying the expression levels of signature biomarkers in a tumour biopsy material, such that the said method comprises steps of:
(a) preparing total RNA from a tumour sample (T);
(b) purifying the said RNA from samples (T) converting them to respective cDNA;

(c) characterized in converting the said cDNA of sample (T) into cRNA and labelling that cRNA with a fluorescent dye (selected from colours Red, Green, Blue, yellow) to obtain a flurorescent-labelled cRNA (fT);
(d) purifying the fluorescent labelled cRNAs (fT) from free, unreacted dye;
(e) hybridising the said fT as obtained in step (d) with oligonucleotide probes on the said microarray chip;
(f) washing the said chip after hybridization in order to remove excess, unbound fT, followed by processing and scanning using an array reader;
(g) the scanner generates fluorescent coloured image of the resultant spots due to hybridization of fT to the probes on the Chip.
(h) fluorescent spots for the colour assigned to fT is normalized and corrected against the background and non-specific (undesired) hybridization.

(i) the normalized values of the each fluorescent spot from the sample as obtained in (h) are matched with the pre-determined Hyperplane approximation as template (also called classifier template) to obtain the classified NO and N+ samples. In another embodiment of the invention it provides a method wherein the microarray chip has a biomarker specific oligonucleotide probe on the surface of a glass slide.
In yet another embodiment of the invention provides a method, wherein the tumour is selected from a group consisting of HNSCC, at NO, N+, Tl, T2, T3, T4 stage of tongue and buccal mucosa.
In yet another embodiment of the invention provides a method wherein the cRNA as in step 'c' is a biomarker for detection of head & neck cancer prognosis having Seq id no. 1-957.
In yet another embodiment of the invention provides a kit for head & neck cancer prognosis consisting of:
I. a panel of cDNA sequence complimentary to the nucleotide sequence having Seq id no. 1-957 on a chip,
II. suitable reagents capable of detecting singly or a combination of the cDNA;
III. instruction manual for using the kit.
wherein, up-regulation (induction) of genes having nucleotide sequence of seq id no. 1-555 and concomitant down-regulation (suppression) of genes having nucleotide sequence of seq id no. 556-957 yielding a comparative expression profile between sample and control to determine presence of tumour and its node metastatic potential.
In yet another embodiment the invention provides an in-silico method for analysing the micro-array data obtained by using the chip comprising steps:
I. normalising the intensity output from the iScan for all Tumour and Normal
samples (n=443;) using Robust Spline Normalization in Lumi, Bioconductor package method;
II. removing the Outlier samples (n=21) from the total sample by using K-means
clustering method;
III. randomly selecting from all samples 50 Tumour samples as validation set and generating robust gene list from remaining samples (n=372) consisting of Tumour samples (n=236) and Normal samples (n=136);
IV. using the normalized intensity values of the two groups of samples (n=372 and n=236) for differential expression analysis using the Bioconductor packages limma and lumi and generating a list of 2980 differentially expressed genes that

classify the tumour and normal samples with adjusted p value < 0.05 and fold change > 1.5 or <-1.5;
V. generating the robust feature list that would classify the NO and N+ tumour
samples using the 236 Tumour samples with 2980 genes wherein the tumour samples (n=236) are separated randomly in two parts; one part containing — 90% of the samples (n=216) and the other part containing —10% of the samples (n=20) and denoting them as "Sample 1" and "Sample 2", respectively;
VI. applying a Recursive Feature Elimination process on Sample 1 and identifying
top 300 features (genes) as important ones;
VII. predicting Sample 1 and Sample 2 separately by choosing top "i" transcripts
(genes) from this list varying i=l,2,...,300, and finding the feature lists with the highest prediction accuracy in both sample sets;
VIII. repeating Steps V to Step VII 200 times to remove the sample bias and
collecting all 200 feature lists so obtained;
IX. preparing a frequency distribution of genes found in all 200 feature lists;
X. characterized in preparing a list of 957 genes from the above 200 feature lists
such that each of these 957 transcripts or genes appears at least once in the 200 feature lists thus making it the largest list of which successive lists are subsets of this 957-gene list;
XI. predicting 50 validation samples set with the gene that appeared highest
number of times and calculating prediction accuracy followed by prediction with two transcripts having highest and next highest frequency and repeating the procedure by increasing the number of transcripts.
In yet another embodiment the invention provides a method for screening therapeutic agents for head-and-neck tumour, comprising the steps:
i.administering a test substance to a non-human animal suffering
from artificially induced head-and-neck tumour; ii.applying the sample from step (i) onto the chip; iii.comparing the expression level to the expression level of the mRNA (or corresponding cDNA) with control wherein test substance is not administered.

We Claim:

1.A chip for detection of head & neck cancer prognosis consisting of cDNA sequence complimentary to the nucleotide sequence having Seq id no. 1-957.
2. The method for microarray-based gene-expression assay conducted on a microarray chip as claimed in claim lfor quantifying the expression levels of signature biomarkers in a tumour biopsy material, such that the said method comprises steps of:

(a) preparing total RNA from a tumour sample (T);
(b) purifying the said RNA from samples (T) converting them to respective cDNA;

(c) characterized in converting the said cDNA of sample (T) into cRNA and labelling that cRNA with a fluorescent dye (selected from colours Red, Green, Blue, yellow) to obtain a flurorescent-labelled cRNA (fT);
(d) purifying the fluorescent labelled cRNAs (fT) from free, unreacted dye;
(e) hybridising the said fT as obtained in step (d) with oligonucleotide probes on the said microarray chip;
(f) washing the said chip after hybridization in order to remove excess, unbound fT, followed by processing and scanning using an array reader;
(g) the scanner generates fluorescent coloured image of the resultant spots due to hybridization of fT to the probes on the Chip.
(h) fluorescent spots for the colour assigned to fT is normalized and
corrected against the background and non-specific (undesired)
hybridization.
(i) the normalized values of the each fluorescent spot from the sample as
obtained in (h) are matched with the pre-determined Hyperplane
approximation as template (also called classifier template) to obtain the
classified NO and N+ samples.
3. The method as claimed in claim 2, wherein the microarray chip has a biomarker specific oligonucleotide probe on the surface of a glass slide.
4. The method as claimed in claim 2, wherein the tumour is selected from a group consisting of HNSCC, at NO, N+, Tl, T2, T3, T4 stage of tongue and buccal mucosa.
5. The method as claimed in claim 2, wherein the cRNA as claimed in step 'c' is a biomarker for detection of head & neck cancer prognosis having Seq id no. 1-957.
6. A kit for head & neck cancer prognosis consisting of:
I. a panel of cDNA sequence complimentary to the nucleotide sequence having Seq id no. 1-957 on a chip as claimed in claim 1,
II. suitable reagents capable of detecting singly or a combination of the cDNA;
III. instruction manual for using the kit.
wherein, up-regulation (induction) of genes having nucleotide sequence of seq id no. 1-555 and concomitant down-regulation (suppression) of genes having nucleotide

sequence of seq id no. 556-957 yielding a comparative expression profile between sample and control to determine presence of tumour and its node metastatic potential.
7. An in-silico method for analysing the micro-array data obtained by using the chip as claimed in claim 1 comprising steps:
XII. normalising the intensity output from the iScan for all Tumour and Normal
samples (n=443;) using Robust Spline Normalization in Lumi, Bioconductor package method;
XIII. removing the Outlier samples (n=21) from the total sample by using K-means clustering method;
XIV. randomly selecting from all samples 50 Tumour samples as validation set and generating robust gene list from remaining samples (n=372) consisting of Tumour samples (n=236) and Normal samples (n=136);
XV. using the normalized intensity values of the two groups of samples (n=372 and
n=236) for differential expression analysis using the Bioconductor packages limma and lumi and generating a list of 2980 differentially expressed genes that classify the tumour and normal samples with adjusted p value < 0.05 and fold change > 1.5 or <-1.5;
XVI. generating the robust feature list that would classify the NO and N+ tumour
samples using the 236 Tumour samples with 2980 genes wherein the tumour samples (n=236) are separated randomly in two parts; one part containing — 90% of the samples (n=216) and the other part containing —10% of the samples (n=20) and denoting them as "Sample 1" and "Sample 2", respectively;
XVII. applying a Recursive Feature Elimination process on Sample 1 and identifying
top 300 features (genes) as important ones;
XVIII. predicting Sample 1 and Sample 2 separately by choosing top "i" transcripts
(genes) from this list varying i=l,2,...,300, and finding the feature lists with the highest prediction accuracy in both sample sets;
XIX. repeating Steps V to Step VII 200 times to remove the sample bias and
collecting all 200 feature lists so obtained;
XX. preparing a frequency distribution of genes found in all 200 feature lists;
XXI. characterized in preparing a list of 957 genes from the above 200 feature lists
such that each of these 957 transcripts or genes appears at least once in the 200 feature lists thus making it the largest list of which successive lists are subsets of this 957-gene list;

XXII. predicting 50 validation samples set with the gene that appeared highest
number of times and calculating prediction accuracy followed by prediction with two transcripts having highest and next highest frequency and repeating the procedure by increasing the number of transcripts.
8. A method for screening therapeutic agents for head-and-neck tumour, comprising the steps:
iv.administering a test substance to a non-human animal suffering
from artificially induced head-and-neck tumour; v.applying the sample from step (i) onto the chip as claimed in claim
i;
vi.comparing the expression level to the expression level of the mRNA (or corresponding cDNA) with control wherein test substance is not administered.