WO2013064789A1 - Method for classifying samples of tissues that are fixed and embedded in paraffin - Google Patents

Abstract

The invention relates to a method for classifying a sample of tissue that is fixed and embedded in paraffin, which comprises the step of determining whether said sample to be tested, referred to as sample "j", is of high enough quality to enable a study of gene expression by quantifying the RNA thereof, said method including the step of comparing the level of expression of n reference genes in said sample "j" with respect to the control expression level of said n reference genes.

Description

 Method for classifying fixed and paraffin-embedded tissue samples

The invention relates to a method of classifying a tissue sample fixed and embedded in paraffin, before a transcriptional evaluation of biomarkers.

BACKGROUND OF THE INVENTION

 Determining levels of gene expression in tissues is critical to accurately and individually diagnosing pathologies, and is increasingly used to establish the treatment strategy. Pharmacogenomic methods can also identify patients who may or may not respond to treatment.

 The standard tissue source for transcriptional evaluation of biomarkers is traditionally fresh frozen tissue. However, formalin-fixed, paraffin embedded tissue (or FFPE for "formalin-fixed, paraffin-embedded" tissue) is by far the most abundant source of tumors, and is even the only material available for tumors. primary such as melanomas (Ravo et al, 2008, Lab Invest, 88: 430-440).

 The use of archived tissue would also be valuable for conducting retrospective studies. Transcriptional analyzes on these fixed and paraffin embedded tissues are possible and have been conducted to validate biomarker signatures associated with clinical features, survival rates, or therapeutic responses (Coudry et al, 2007, J. Mol. Diag 9: 70-79, Cronin et al., 2004, Am J Pathol, 164: 35-42). Unfortunately, these analyzes are not conducted routinely because of the deterioration of the RNA in these stored tissues. Indeed, the low quality of the extracted RNAs reduces the reliability of the biomarker quantifications (Mittempergher et al., 201 1, PLoS One 6: e17163).

 There is therefore a need for a method for selecting fixed and paraffin-embedded tissue samples that can be used for transcriptional analyzes. Summary of the invention

The invention addresses this need by providing a method of classifying a fixed and paraffin-embedded tissue sample, wherein it is determined whether said test sample, designated Sample "j", is of sufficient quality for a study. of the gene expression by quantification of its RNA, said method comprising comparing the level of expression of n reference genes in said sample "j" relative to the level of expression control of said n reference genes. The combination of the reference genes and the entire number n depends on the type of tissue to be qualified.

 Samples with a level of expression of the reference genes comparable to the level of expression control, namely the level of expression of these same genes in the control samples, are selected, and are particularly useful for performing reliable transcriptional studies.

In a preferred embodiment, the method comprises:

 determining the level of expression of said n reference genes in the sample "j" to be tested, and

 the determination of the deviation

où n est le nombre de gènes de références, μί et ai sont respectivement la moyenne et la déviation standard pour l'expression d'un gène « i » de référence établie au préalable ou évaluée sur un jeu d'échantillons contrôles de bonne qualité ; xij est le niveau d'expression dudit gène i évaluée dans l'échantillon « j »,

where n is the number of reference genes, μί and ai are respectively the mean and the standard deviation for the expression of a reference "i" gene previously established or evaluated on a set of good quality control samples; xij is the level of expression of said i gene evaluated in the sample "j",

the value of the deviation indicating whether sample "j" is of sufficient quality for a gene expression study or transcriptomic analysis.

Preferably, the tissue sample is a sample of tumor tissue, preferably a melanoma, more preferably n greater than or equal to 4, and said n reference genes comprise at least the β-actin, HPRT, TBP genes. and TRFC. Legend of figures:

 Figure 1 represents a "Boxplot" of mean mRNA levels of 19 genes of interest and 10 reference or "housekeeping" genes in all frozen and FFPE samples (formalin-fixed and included). in paraffin) analyzed.

Figure 2A shows a study of the correlation between frozen samples and FFPE samples at the median level of each gene, evaluated in all patients. The adjusted Pearson correlation coefficient was 0.88 (p <0.0001).

Figure 2B shows examples of patients with "good" or "bad" correlations between frozen samples and FFPE samples. Figure 3 shows correlations corrected FFPE samples / frozen samples for each individual. The correlation test tests the hypothesis "the correlation is not valid". The threshold represents the rejection threshold (alpha = 5%) according to the Bonferroni multiple test. The individuals to be distributed are those who are below the threshold.

Detailed description of the invention

 The inventors now propose a simple and inexpensive method to identify and select exploitable paraffin-embedded fixed samples of sufficient quality to allow quantification of biomarkers at their RNA level.

Definitions

 By "tissue sample fixed and embedded in paraffin" is meant any tissue biopsy, preferably human or animal, fixed by any method known to those skilled in the art, and included in paraffin. A conventional method of attachment includes formalin or ethanol fixation. The samples generally comprise at least 10, at least 100, or at least 1000, at least 5000, different RNA molecules.

 A sample of "sufficient quality" or "good quality" means that the RNA in the sample is not substantially degraded or remains quantifiable, allowing a study of gene expression or transcriptome, or suitable for serve in an amplification reaction, in particular by quantitative RT-PCR. Conversely, a sample of "poor quality" or "poor quality" is a sample whose RNA is too degraded to be suitable for gene expression or transcriptome study, or is unsuitable for use in a patient. amplification reaction, in particular by quantitative RT-PCR.

By "level of expression control" is meant a level of expression previously determined or evaluated on a set of good quality control samples, also called reference samples.

 By "good quality control sample set" is meant paraffin-embedded frozen or fixed tissue samples having expression levels of the reference gene RNAs correlated with the expression levels of the RNAs extracted from the frozen tissues. The control samples are of the same type as the samples to be tested (that is to say that they come for example from the same type of organ, presenting the same pathology).

"Reference genes" means ubiquitous or "housekeeping" genes, the expression of which does not substantially vary, in particular as a function of the age, gender, or condition of the patient , or tissue (healthy or tumoral for example). Tissue samples

 The tissue sample to be tested may be any tissue sample whose gene expression level is desired or transcriptomic analysis.

It may be a healthy tissue, or a tissue of a patient with a particular condition, or in a particular condition. In a preferred embodiment, it is a sample of tumor tissue.

 In particular, it may be a melanoma sample.

 It may also be a sample of a breast, lung or kidney tumor tissue, for example, or a tumor sample of cancer of the colon, prostate, gastric, pancreas, cervix uterine, ovarian, liver, bladder, thyroid, brain, or hematological malignancy (the tissue sample is then a bone marrow biopsy).

Reference genes:

 The reference genes may vary depending on the type of tissue to be tested.

The number of reference genes is at least 2, preferably at least 3, more preferably at least 4.

 In a particular embodiment, the number of reference genes is greater than or equal to 3, preferably greater than or equal to 4.3, and the said n reference genes comprise at least three genes among the lactin genes, preferably actin genes. β, HPRT (gene coding for hypoxanthine phosphoribosyltransferase), TBP (gene encoding the TATA-BOX binding protein) and TFRC (gene encoding the transferrin receptor). This is particularly suitable in the case where the tissue sample is a melanoma sample. It may be a primary melanoma, or with cutaneous metastases, or with metastases in the lymph nodes. Preferably, the method is implemented by determining the expression levels of these 4 genes only.

Quantification of transcripts of reference genes, and classification of the sample:

The method of the invention aims to classify a tissue sample without having to analyze the expression of other genes than the only reference genes. At the end of the process, the sample is classified as "good quality" or "poor quality". If the sample is classified as "good quality", then it can be used for analysis of the expression of any other gene of interest, or for transcriptome analysis.

The method of the invention comprises determining the level of gene expression, namely the quantification of mRNAs, reference genes. The amplified mRNA fragments do not need to be full-length: amplicons ranging in size from about 70 to about 100 base pairs are sufficient.

This quantification can be carried out by any method known to those skilled in the art, in particular by quantitative RT-PCR (qRT-PCR).

The preliminary step is the extraction of the RNA and its purification, by any known method, for example by the method of Chomcynsky and Sacchi, 1987, Anal. Biochem 162: 156-159, or by improved methods (see for example WO01 / 46402). Extraction involves the removal of paraffin, for example using xylenes as solvents or by incubation at 65-75 ° C in lysis buffer. The extraction of the RNA can then be carried out by precipitation or by chromatography for example. Control steps, for example by gel electrophoresis, may be included.

 Quantitative RT-PCR can use different techniques of intercalators or fluorescent probes. One usual technique uses a DNA-specific fluorophore of the SyBr Green type, and the increase in fluorescence is then measured at the end of each elongation step. According to another usual technique, it is also possible to use an internal fluorescent probe specific for the amplified product. Several types of probe formats are available. For example, the Roche Light Cycler can be used with two hybridization probes, which hybridize "head-to-tail" on their target sequence, one of the two probes carrying a fluorescein at its end. In a preferred embodiment, hydrolysis probes or TaqMan®, which utilize the exonuclease activity of Taq polymerase, are used to hydrolyze a fluorescent probe attached to its target. It is also possible to use "molecular beacons" (Stratagene) which are hairpin hybridization probes, the loop of the probe being complementary to the targeted sequence.

The method of the invention makes it possible to select the samples in which the integrity of the RNAs allows a transcriptional analysis. The samples are selected to express reference genes at a comparable level comparable to the level of expression of these same genes in the control samples.

 The inventors have thus developed a statistical method capable of identifying the samples whose expression levels are correlated with those of frozen samples, by quantifying the transcripts of only a few reference genes.

 This approach was based on the assumption that the level of expression of these reference genes should be generally stable between good quality samples.

où n est le nombre de gènes de références, μί et σί sont respectivement la moyenne et la déviation standard pour l'expression d'un gène « i » de référence établies au préalable ou évaluées sur ledit jeu d'échantillons contrôles de bonne qualité ; xij est le niveau d'expression dudit gène i évaluée dans l'échantillon « j », A deviation of the expression from the good quality control samples can be calculated as follows:

where n is the number of reference genes, μί and σί are respectively the mean and the standard deviation for the expression of a reference "i" gene previously established or evaluated on said set of good quality control samples; xij is the level of expression of said i gene evaluated in the sample "j",

the value of the deviation indicating whether sample "j" is of sufficient quality for a study of gene expression.

 More precisely, the significance of the deviation can be tested by a classical Chi square test, with n degrees of freedom:

If the p value is nonsignificant, we conclude that there is no deviation, and the sample is classified as "good quality".

 If the p value is significant, it is concluded that there is a substantial deviation, and the sample is classified as "poor quality".

In a particular embodiment, the test can be performed without using new quality control samples, the mu and sigma reference values having been previously established. Thus, the reference values μ and σ, representing the means / standard deviations of each reference gene are

 μ = 8.15 and σ = 1.86 for the β-actin gene;

 μ = -15.06 and σ = 2.03 for the HPRT gene;

 μ = -2.55 and σ = 1 .41 for the TBP gene

 μ = -14.34 and σ = 1.91 for the TRFC gene.

Samples classified as "good quality" can be used for analysis of the expression of any other gene of interest, or for transcriptome analysis. In these subsequent analyzes, preferably by RT-PCR, internal control may still be desirable, but normalization with other reference genes is no longer necessary.

The figures and examples illustrate the invention without limiting its scope. Examples

Example 1 Selection of Melanoma FFPE Tissues for a Transcriptional Analysis

MATERIALS AND METHODS patients

 The inventors studied tissue samples collected from 2000 to 2005, frozen or formalin-fixed and paraffin-embedded (FFPE), for the same tumor sections, from 25 patients with primary melanomas (7 patients), melanomas with cutaneous metastases (4 patients), and melanomas with metastases in the lymphatic nuclei (14 patients).

 Sample, RNA extraction, and reverse transcription

 All frozen fresh melanoma tissues were divided, half frozen, and one formalin fixed and paraffin embedded. More than 90% of the tissue is composed of tumor cells. Five sections of 10μηι of frozen samples and ten sections of 10μηι of FFPE blocks were extracted with RNA.

The total RNA of archived FFPE block pairs and related frozen tumors were extracted with the method of Chomcynsky and Sacchi et al, supra, for frozen samples and using the Qiagen FFPE RNA extraction kit, after xylene treatment for FFPE samples, according to the manufacturer's protocol. Each sample was treated with DNase I to remove all traces of genomic DNA. Reverse transcription was conducted with Super-Script II (Invitrogen).

 Total RNA yield was determined using a NanoDrop spectrophotometer

ND-1000 (NanoDrop Tech, Wilmington, DE). RNA integrity was assessed by Agilent 2100 Bioanalyzer electrophoresis (Agilent Technologies) compared to standard reference RNA as previously described (Cronin et al, supra).

 Probes and primers for RT-PCR were designed, tested, and validated. Fluorogenic probes were doubly labeled, with 5-FAM as reporter, and TAMRA as quencher.

The sequences of the primers used are indicated below:

 For the actin β gene:

 Meaning primer CTGGCACCCAGCACAATG (SEQ ID NO: 1)

GCCGATCCACACGGAGTACT antisense primer (SEQ ID NO: 2)

For the HPRT1 gene:

Primer meaning TTATGGACAGGACTGAACGTCTTG (SEQ ID NO: 3)

GACACAGAGG G CTACAATGTG (SEQ ID NO: 4) antisense primer

For the TBP gene:

 Sense Primer CACgAACCACggCACTgATT (SEQ ID NO: 5)

Antisense primer TTTTCTTgCTgCCAgTCTggAC (SEQ ID NO: 6)

For the TFRC gene:

 Meaning primer CTTTGCCAGTTGGAGTGCTG (SEQ ID NO: 7)

ACGAAAGGTATCCCTCTAGCCAT antisense primer (SEQ ID NO: 8) Muliparametric quantization of transcripts

 Invasive biomarkers and angiogenesis / lymphangiogenesis: VEGF (soluble forms VEGF121 and VEGF165), VEGF receptors (VEGFR1 and VEGFR2a, PDGF-A, PDGF-B, PDGF receptors (PDGFR-alpha and beta), serine proteases (uPA PAI-1) and matrix metalloproteinase (MMP1, 2, 9, and 14), MMP inhibitors (TIMP1, 2), protease inducer EMMPRIN, transcriptional transcription factors angiogenesis / lymphangiogenesis HIF1a, PROX-1.

 Reference transcripts studied: HPRT, TBP, microglobin B2, GAPDH, ACTBP, GUS, PPIA, TFRC, 18S, 5S, RPLP1, RPLPO, RPL5, RPL19, actin-alpha, and actin-beta. normalization

 To compare the expression profiles between the samples, standardization based on the reference genes was conducted to correct the differences arising from the variability in the quality of the RNA and the total amount of RNA in each assay. The relative quantification of each transcript was inspired by the work of Cronin et al, supra.

Statistical analysis

 Statistical analyzes were conducted with R1 software (version 2.13.1). P-values are considered significant below the 5% mark after Bonferroni adjustment for multiple tests. RESULTS

Comparison of RNA expression patterns of FFPE melanoma tissues and in the frozen state:

 The expression of 25 genes involved in angiogenesis, lymphangiogenesis and tumor invasion was analyzed in malignant melanomas. For this, total RNA was prepared for 25 related pairs of frozen samples and FFPE. An RNA inspection by Agilent's Bioanalyzer showed a typical undegraded RNA profile in the frozen samples, whereas the FFPE extracts had a degraded RNA of about 150 and 50 bp, depending on the samples. .

Heterogeneity in the quality and amount of RNA extracted is known to be primarily caused by variations in the amount of tissue, the type of fixation and the delay in fixation after surgery. In addition, the efficiency of reverse transcription and PCR itself may represent an additional parameter of variability. Also, the qRT-PCR measurements of the genes of interest were normalized against a valid set of housekeeping genes. This validated game was determined by comparing 15 different "housekeeping" genes between tissues frozen and FFPE tissues, 10 genes having been retained as reference genes, for their stable expression (close means between frozen samples and FFPE).

Figure 1 represents a "Boxplot" of mean mRNA levels of 19 genes of interest and 10 "housekeeping" genes in all frozen and FFPE samples, showing comparable averages for most genes, but not all. For example, VEGFR-2 and PDGFR-beta show closely related means, whereas MMP1 and PDGFR-alpha are unrelated.

 Figure 2A shows a study of the correlation between frozen samples and FFPE samples at the median level of each gene, evaluated in all patients, achieving a "very good" Pearson correlation coefficient of 0.88 (p. <0.0001).

 Figure 2B shows examples of patients with "good" or "bad" correlations between frozen samples and FFPE samples (Pearson correlation coefficients of 0.87 and 0.48 with p <0.0001 and p = 0.007 respectively).

Of the 25 samples analyzed, 21 were rated "good", 1 "average", and 3 "bad" (see Figure 3).

Identification of samples with good correlation of gene expression between frozen state and FFPE:

 The inventors then developed a statistical method capable of identifying melanoma samples with a good frozen correlation / FFPE, to rule out samples whose expression levels are not correlated, based on paraffin data. for reference genes only.

 This approach was based on the assumption that the expression level of these reference genes should be globally stable between good samples. The inventors then defined a Mean Mean Good Expression Profile (MGEP) as the average of the expression profiles of the "good" samples and the samples deviating significantly from the MGEP as bad profiles.

The deviation from the MGEP was calculated as follows:

 The deviation was measured and tested by a classical Chi square test, with n degrees of freedom:

If the p value is nonsignificant, we conclude that there is no deviation from the MGEP, and the sample is rated "good". If the p value is significant, it is concluded that there is a substantial deviation from the MGEP, and the sample is rated "bad".

Of the 21 good samples identified, 14 were used as "good quality control" samples, to construct the MGEP (the selection being based on a correlation p value of less than 10 ^"3 ). wrong samples were used for validation.

 The set of reference genes with the minimum average coefficient of variation on the quality control samples was chosen: it included the 4 genes actin, HPRT, TBP, TRFC. With this set of reference genes, the MGEP profile-based approach discriminated perfectly against the 21 samples of the 3 bad ones (100% prediction).

Claims

Claims. 1. A method of classifying a fixed and paraffin-embedded tissue sample, wherein it is determined whether said test sample, designated sample "j", is of sufficient quality for a study of gene expression by quantification of its RNA said method comprising comparing the level of expression of n reference genes in said sample "j" relative to the level of expression control of said n reference genes. The method of claim 1, wherein n is greater than or equal to The method of claim 2, wherein n is greater than or equal to 4. Method according to one of claims 1 to 3, said method comprising:  determining the level of expression of said n reference genes in the sample "j" to be tested, and  the determination of the deviation

where n is the number of reference genes, μ, and σ, are respectively the mean and the standard deviation for the expression of a reference "i" gene previously established or evaluated on a set of control samples of good quality; x, _j is the level of expression of said evaluated gene i in sample "j",  the value of the deviation indicating whether sample "j" is of sufficient quality for a gene expression study or transcriptomic analysis. The method of one of claims 1 to 4, wherein the tissue sample is a tumor tissue sample. The method of claim 5, wherein the tissue sample is a melanoma sample. The method of claim 6, wherein n is greater than or equal to 4, and said n reference genes comprise at least the Yactin, HPRT, TBP and TRFC genes. The method of claim 7, wherein n = 4 and said reference genes are the β-lactin, HPRT, TBP and TRFC genes. The method of claim 8, wherein the values μ and σ for each of said reference genes are  μ = 8.15 and σ = 1.86 for the β-actin gene;  μ = -15.06 and σ = 2.03 for the HPRT gene;  μ = -2.55 and σ = 1 .41 for the TBP gene  μ = -14.34 and σ = 1.91 for the TRFC gene. The method of claim 5, wherein the tissue sample is a sample of breast, kidney or lung tumor tissue, or colon, prostate, gastric, pancreatic, cervical cancer. uterus, ovaries, liver, bladder, thyroid, brain, or hematological malignancy. 1 1. Method according to one of claims 1 to 10, wherein the expression of the reference genes is quantified by quantitative RT-PCR. The method according to one of claims 1 to 11, wherein the fixed and paraffin-embedded tissue sample is a formalin fixed tissue sample. 13. Method according to one of claims 1 to 12, wherein the expression of the reference genes is quantified by amplification of fragments of 70 to 100bp.