Methods for aiding diagnosis of oesophageal conditions

Abstract

The invention relates to a method of collecting information useful in aiding diagnosis of an oesophageal disease in a subject comprising: providing a sample from said subject, wherein said sample comprises cells collected from the surface of the subject's oesophagus; contacting said sample with an aqueous fluid; taking an aliquot of said aqueous fluid; assaying said aliquot of aqueous fluid to determine the presence of one or more soluble biomarker(s). The invention provides sets of biomarkers for different applications. For Barrett's Oesophagus (BE), preferred biomarkers are TFF3, REG4, FABP2, MEP1A, CXCL8; and SPINK4. For Eosinophilic Oesophagitis (EOE) preferred biomarkers are C-kit, Eotaxin-3, IL-5, IL-13, IL-33, IL-25; and Eotaxin-2. The invention also provides uses, apparatus and kits.

Description

[0001] P11112GBW0

[0002] METHODS FOR AIDING DIAGNOSIS OF OESOPHAGEAL CONDITIONS

[0003] BACKGROUND TO THE INVENTION

[0004] The invention relates to improvements in methods for diagnosing, or methods for aiding the diagnosis of, oesophageal conditions such as Barrett’s oesophagus (‘BO’ or ‘BE’) and / or eosinophilic esophagitis (EOE).

[0005] Historically, detection of Barrett’s Oesophagus required biopsy. This involved a skilled operator using an endoscope. The endoscope is introduced into the oesophagus of the patient. This usually requires sedation of the patient. The skilled operator would then visually inspect the surface of the oesophagus. The skilled operator would then use their judgement to visually select areas of the oesophagus for biopsy, and / or would attempt a randomised sampling protocol to collect biopsy material from different areas of the oesophagus. This was extremely time consuming, required complex clinically trained operators, involved risks to the patient such as side-effects from sedation and / or the risk of oesophageal spasm, and represented a considerable burden on the healthcare provider.

[0006] In order to address those problems, a capsule sponge sampling technique was developed. In this approach, a capsule sponge (sometimes referred to as “cytosponge” ™) is attached to a retrievable device such as a string. The patient then swallows the capsule sponge. Upon reaching the stomach, the capsule retaining the sponge in a compressed state dissolves, and the sponge expands to its full size. The sponge is then withdrawn through the oesophageal lumen, which has the effect of sampling cells from the oesophagus during its transit through the oesophageal lumen. The sponge and the collected cells were then placed into a preservative liquid. A preservative liquid / sponge / container was then transported to a laboratory for analysis. The fluid / cell mixture is then centrifuged to collect the cells in a pellet. This cell pellet is then resuspended and treated to form a clot. These clotted cells are then formalin- fixed and paraffin embedded (FFPE). This FFPE cell preparation is then sectioned. Those sections are then mounted on slides. Those slide mounted sections are then stained with the relevant detection reagents (e.g. antibodies) in order to visualise the proteins (markers) of interest. The visual output of this procedure is then assessed by a suitably trained clinical operator, and a judgement is made whether or not the patient has particular condition(s) such as Barrett’s Oesophagus. P11112GBW0

[0007] This process is extremely time consuming. The process requires different steps to be taken in different clinical / laboratory settings. The method requires a high degree of skill and training on the part of the operator. The process is laborious and time consuming. The process requires handling of hazardous chemicals such as formalin. The methodology requires judgment which can be subjective and lead to difficulties in reproducibility between different assessment centres / different operators.

[0008] WO / 2014 / 128460 discloses a method of aiding detection of a surface abnormality in the oesophagus of a subject. The method involves collection of cells from the oesophagus using a capsule sponge, and then pelleting and resuspending those cells in plasma, followed by thrombin treatment to form a clot, and sectioning said clot for microscopic examination. This is very labour intensive, and relies on skilled operators making inspections of cells via immunohistochemistry and producing subjective judgements as to the condition of those cells to aid diagnosis.

[0009] US8,709,736B2 discloses use of TFF3 in the diagnosis and detection of Barrett's esophagus using non-invasive, non-endoscopic methods. The method involves collection of cells from the oesophagus using a capsule sponge, and then pelleting and resuspending those cells in plasma, followed by thrombin treatment to form a clot, formalin-fixing the clotted cells, paraffin embedding them and staining the cells with H&E and / or TFF3 antibodies. This is a laborious method with time-consuming steps and human operators needed to carry out the analysis via immunohistochemistry including subjective interpretation of antibody staining patterns.

[0010] WO 2018 / 039422 discloses a ‘non-invasive screening test’ for detection of Barrett's esophagus (BE). Amongst other things, this document describes collection of endoscopic brushings and testing for expression of MLIC2 and / or CDX2 via RT-PCR followed by qPCR. This document mentions immunoassays, but does not provide any examples nor any data from any immunoassays. This document is focussed on nucleic acid based testing.

[0011] The present invention seeks to overcome problem(s) associated with the prior art.

[0012] SUMMARY OF THE INVENTION

[0013] The prior art methods focussed on analysis of the cells harvested from the oesophageal surface. It is those cells which were examined for the abnormalities P11112GBW0 characteristic of conditions such as Barrett’s Oesophagus. The prior art has been focussed on the targeted collection and preservation of those cells, and the correct processing and handling of those cells, in order to obtain the best possible clinical information.

[0014] By contrast, the present inventors have surprisingly discovered that certain molecules such as proteins present in the surrounding collection fluid could in fact provide clinically robust and useful information. Molecules which can contribute this information which were unexpectedly discovered in the surrounding preservative fluid can include glycoproteins (e.g. shed from the surface of the cells of interest) and / or secreted proteins which are introduced into the surrounding preservative fluid by the collected cells. It was of particular surprise to the inventors that the preservative fluid - without cells - could be used to provide such valuable clinical information as described herein.

[0015] To place this in context, it is important to note that the current method (outlined above in the background section) currently involves throwing away the preservative liquid. The preservative liquid is literally discarded accordingly to current techniques.

[0016] However, the inventors now teach to do the opposite to the prior art method i.e. to exclude the cells from the liquid, and then take the “cell free” liquid and perform analyses on that in order to provide clinically useful information to aid diagnosis.

[0017] In this context, the complex process used to select and refine a panel of biomarkers effective in providing information to aid diagnosis of BE or EOE was surprisingly founded on using this cell free fluid as a starting point. This itself was a significant decision departing from the prior art approaches.

[0018] It is an advantage of this method that the cells collected can be used for a different analysis. It is an advantage of the invention that no cells at all are needed in the method for collecting information useful for aiding diagnosis. It is an advantage of the invention that the cells may be used for the sequencing of nucleic acids such as DNA. It is an advantage of the invention that molecules which may not survive the prior art extraction - pelleting - FFPE - sectioning process, and / or molecules which may not co-locate to the cells of interest (e.g. if they were secreted from those cells rather than being immobilised with said cells during the prior art process) can now be interrogated to provide clinically useful information according to the present invention. P11112GBW0

[0019] Currently, if DNA sequence analysis is desired, cells have to be extracted from the FFPE sample by physically cutting them out, those recovered cells then have to be processed via more complicated routes (because FFPE treatment can lead to more difficult nucleic acid extraction) and typically yield poorer quality nucleic acid sequence information (due to lower nucleic acid sample quality following FFPE processing and subsequent nucleic acid extraction). Thus it is an advantage of the invention that easier and superior quality nucleic acid information can be extracted from the cells, since the cells are no longer needed for the method in the invention.

[0020] It is an advantage of the invention that a better quality cell assay may be used (if desired) because options are kept open when the cells are not required to be formalin fixed / paraffin embedded. It is an advantage of the invention that it requires less labour, requires less cost, and is safer (omitting (for example) handling of Formalin).

[0021] All of these advantages flow from the important advance, which departs from the prior art, which is to focus on the soluble biomarkers in the fluid contacted with the cells.

[0022] Thus in one embodiment the invention relates to a method of collecting information useful in aiding diagnosis of an oesophageal disease in a subject comprising: providing a sample from said subject, wherein said sample comprises cells collected from the surface of the subject's oesophagus; contacting said sample with an aqueous fluid; taking an aliquot of said aqueous fluid; assaying said aliquot of aqueous fluid to determine the presence of one or more soluble biomarker(s).

[0023] Suitably said one or more soluble biomarker(s) comprise the soluble protein biomarker(s):

[0024] (i) TFF3

[0025] (ii) REG4

[0026] (iii) FABP2

[0027] (iv) MEP1A

[0028] (v) CXCL8; and

[0029] (vi) SPINK4 and wherein presence of each of said soluble protein biomarkers (i) to (vi) in said aliquot is indicative of said subject having Barrett’s Oesophagus. Suitably said method P11112GBW0 has an ALIROC value of at least 0.93. More suitably said method has an ALIROC value of at least 0.93 at a sensitivity of 71 % and specificity of 100%. We refer to the table below:

[0030] In one embodiment, Gal-3 may also be assayed. An advantage of this is that by assaying Gal-3 in addition to the 6 markers noted above, the AUG may be moved from 0.93 to 0.95.

[0031] Thus, suitably said one or more soluble biomarker(s) comprise the soluble protein biomarker(s):

[0032] (i) TFF3

[0033] (ii) REG4

[0034] (iii) FABP2

[0035] (iv) MEP1A

[0036] (v) CXCL8

[0037] (vi) SPINK4; and

[0038] (vii) Gal-3 and wherein presence of each of said soluble protein biomarkers (i) to (vii) in said aliquot is indicative of said subject having Barrett’s Oesophagus. Suitably said method has an ALIROC value of at least 0.95. More suitably said method has an ALIROC value of at least 0.95 at a sensitivity of 71% and specificity of 100%.

[0039] Suitably said one or more soluble biomarker(s) comprise the soluble protein biomarker(s):

[0040] (i) TFF3

[0041] (ii) REG4

[0042] (iii) FABP2 P11112GBW0

[0043] (iv) MEP1A

[0044] (v) CXCL8

[0045] (vi) SPINK4

[0046] (vii) Gal-3

[0047] (viii) MUC2

[0048] (ix) GPA33; and

[0049] (x) ITLN1 and wherein presence of each of said soluble protein biomarkers (i) to (x) in said aliquot is indicative of said subject having Barrett’s Oesophagus.

[0050] In another embodiment the invention relates to a method as described above wherein said one or more soluble biomarker(s) comprise the soluble protein biomarker(s):

[0051] 1) C-kit

[0052] 2) Eotaxin-3

[0053] 3) IL-5

[0054] 4) IL-13

[0055] 5) IL-33; and

[0056] 6) IL-25 and wherein presence of each of said soluble protein biomarkers (1) to (6) is indicative of said subject having Eosinophilic Oesophagitis.

[0057] In another embodiment the invention relates to a method as described above wherein said one or more soluble biomarker(s) comprise the soluble protein biomarker(s):

[0058] 1) C-kit

[0059] 2) Eotaxin-3

[0060] 3) IL-5

[0061] 4) IL-13

[0062] 5) IL-33

[0063] 6) IL-25; and

[0064] 7) Eotaxin-2 and wherein presence of each of said soluble protein biomarkers (1) to (7) is indicative of said subject having Eosinophilic Oesophagitis.

[0065] In another embodiment the invention relates to a method as described above wherein said one or more soluble biomarker(s) comprise the soluble protein biomarker(s): P11112GBW0

[0066] 1) C-kit

[0067] 2) Eotaxin-3

[0068] 3) IL-5

[0069] 4) IL-13

[0070] 5) IL-33

[0071] 6) IL-25

[0072] 7) Mast cell tryptase

[0073] 8) CPA3

[0074] 9) ECP; and

[0075] 10) Periostin and wherein presence of each of said soluble protein biomarkers (1) to (10) is indicative of said subject having Eosinophilic Oesophagitis.

[0076] In another embodiment the invention relates to a method as described above wherein said one or more soluble biomarker(s) comprise the soluble protein biomarker(s):

[0077] 1) C-kit

[0078] 2) Eotaxin-3

[0079] 3) IL-5

[0080] 4) IL-13

[0081] 5) IL-33

[0082] 6) IL-25

[0083] 7) Eotaxin-2

[0084] 8) Mast cell tryptase

[0085] 9) CPA3

[0086] 10) ECP; and

[0087] 11) Periostin and wherein presence of each of said soluble protein biomarkers (1) to (11) is indicative of said subject having Eosinophilic Oesophagitis.

[0088] In another embodiment the invention relates to a method as described above wherein said one or more soluble biomarker(s) comprise the soluble protein biomarker(s):

[0089] 1) C-kit

[0090] 2) Eotaxin-3

[0091] 3) IL-5

[0092] 4) IL-13

[0093] 5) IL-33 P11112GBW0

[0094] 6) IL-25

[0095] 7) Eotaxin-2 (CCL24)

[0096] 8) Mast cell tryptase

[0097] 9) CPA3

[0098] 10) ECP

[0099] 11) Periostin

[0100] 12) CLC / GAL-10; and

[0101] 13) EDN (eosinophil derived neurotoxin) and wherein presence of each of said soluble protein biomarkers (1) to (13) is indicative of said subject having Eosinophilic Oesophagitis.

[0102] In another embodiment the invention relates to a method as described above wherein said one or more soluble biomarker(s) comprise:

[0103] (a) a specific binding partner of Cytisus scoparius (CSA) lectin, and wherein presence of said specific binding partner of Cytisus scoparius (CSA) lectin is indicative of said sample being adequate to support collection of information useful in aiding diagnosis of an oesophageal disease. Suitably said method has an ALIROC value of at least 0.95. More suitably said method has an ALIROC value of at least 0.95 with sensitivity of 87.96 and specificity of 94.44.

[0104] In another embodiment the invention relates to a method as described above wherein said one or more soluble biomarker(s) comprise:

[0105] (a) a specific binding partner of Cytisus scoparius (CSA) lectin; and

[0106] (b) a specific binding partner of Peanut Agglutinin (PNA) lectin, and wherein presence of each of said specific binding partners (a) and (b) is indicative of said sample being adequate to support collection of information useful in aiding diagnosis of an oesophageal disease. Suitably said method has an ALIROC value of at least 0.96. More suitably said method has an ALIROC value of at least 0.96 with sensitivity of 87.96 and specificity of 94.44.

[0107] Suitably presence of said one or more soluble biomarker(s) is determined as presence at an elevated level in said aliquot by comparison to a reference value.

[0108] In another embodiment the invention relates to an apparatus or system which is P11112GBW0

[0109] (a) specifically adapted to analyse an oesophagal sample from a subject, wherein said analysis comprises:

[0110] (b) taking an aliquot of aqueous fluid previously contacted with oesophageal cells obtained from said subject; assaying said aliquot of aqueous fluid to determine the presence of one or more soluble biomarker(s) according to any one of: Table 1 or Table 2; said apparatus or system comprising an output module, wherein if presence of said markers are detected, then said output module indicates an increased likelihood of an oesophageal disease for said subject, wherein said oesophageal disease is selected from the group consisting of Barrett’s Oesophagus and Eosiniphilic Oesophagitis. Suitably when presence of markers of Table 1 are detected, more suitably detected at elevated levels compared to a reference aliquot, increased likelihood of Barrett’s Oesophagus is indicated. Suitably when presence of markers of Table 2 are detected, more suitably detected at elevated levels compared to a reference aliquot, increased likelihood of Eosiniphilic Oesophagitis is indicated.

[0111] In another embodiment the invention relates to use for applications relating to aiding detection of an oesophageal disease in the oesophagus of a subject, wherein said oesophageal disease is selected from the group consisting of Barrett’s Oesophagus and Eosiniphilic Oesophagitis, of a material which recognises, binds to or has affinity for certain soluble biomarkers, wherein the soluble biomarkers are as defined in any one of: Table 1 (detection of Barrett’s Oesophagus) or Table 2 (detection of Eosiniphilic Oesophagitis). In another embodiment the invention relates to use as described above of a combination of materials, each of which respectively recognises, binds to or has affinity for one or more of said soluble biomarkers.

[0112] In another embodiment the invention relates to an assay device for use in aiding detection of an oesophageal disease in the oesophagus of a subject, wherein said oesophageal disease is selected from the group consisting of Barrett’s Oesophagus and Eosiniphilic Oesophagitis, which comprises a solid substrate having a location containing a material, which recognises, binds to or has affinity for certain soluble biomarkers, wherein the soluble biomarkers are as defined in any one of: Table 1 (detection of Barrett’s Oesophagus) or Table 2 (detection of Eosiniphilic Oesophagitis).

[0113] DETAILED DESCRIPTION OF THE INVENTION P11112GBW0

[0114] The term ‘aliquot’ has its normal meaning in the art. For example, according to the U.S. National Institute of Health (NIH), “An aliquot pertains to a portion of the whole, especially a sample taken for biological analysis or other treatment.” Thus, in the context of the present invention, when a volume of fluid has been produced by contacting that fluid with the sample of cells, an ‘aliquot’ of said fluid is a volume of said fluid which is less than the whole volume of said fluid. The term ‘aliquot’ is used herein to refer to a portion of that fluid in order to promote ease of understanding, especially when the term ‘sample’ is often used to refer to the sample of cells obtained from the subject’s oesophageal surface. The biomarkers described herein are assayed in the aliquot of fluid. The biomarkers described herein are not assayed on the cells obtained from the subject.

[0115] In one embodiment a specimen may be taken from the aqueous fluid sample, and assaying presence of one or more soluble biomarker(s), such as soluble protein biomarkers, may be performed on said specimen.

[0116] The phrase “specific binding partner” has its natural meaning. This is most often used herein when discussing the use of lectins as biomarkers. It is not detection of the lectin(s) themselves which is important in the invention, it is the detection of the target(s) of those lectin(s) i.e. detection of molecule(s) specifically recognised by the particular lectin(s) which are mentioned herein. In case any further guidance is required, nonspecific binding (if any) should be discounted in the normal manner, for example by determination of ‘background’ signal when the lectin of interest is presented with a non-specific target and subtraction of that background signal from any assay of the lectin associating with its specific binding partner.

[0117] The term ‘contacting’ has its natural meaning. The purpose of this step is to allow soluble biomarker(s), (such as soluble protein(s)), from the cells in the sample to enter the aqueous fluid. Thus, suitably ‘contacting’ is for a time sufficient to allow this to happen. For example, suitably the sample of cells is contacted with the aqueous fluid for at least 30 seconds, suitably at least 60 seconds, suitably for at least 5 mins, suitably for at least 10 mins, suitably for at least 30 mins, suitably for at least 60 mins or more. Suitably ‘contacting’ with may mean incubating with.

[0118] In principle there is no minimum time, provided that soluble biomarkers present can enter the aqueous fluid. Cut-off time for degradation is 30 days after sample collection - thus suitably contacting or incubation is 30 days or less. Suitably optimal contacting or incubation time is 24-48 hours after collection. P11112GBW0

[0119] Suitably the method is non-invasive. An example of an invasive method is one which requires collection of biopsy material such as pinch biopsy collection from the subject. Suitably collection of biopsy is omitted or excluded by the present invention.

[0120] Suitably the method is non-endoscopic.

[0121] Suitably the method is a method of screening for an oesophageal disease. Suitably screening may be screening one subject or a number of subjects, such as population screening.

[0122] The oesophageal disease may comprise Barrett’s oesophagus. The oesophageal disease may comprise Eosinophilic Oesophagitis.

[0123] The method may be a method of aiding detection of a surface abnormality in the oesophagus of a subject. In this embodiment suitably detection of abnormal levels of the marker(s) infers that the subject has an increased likelihood of a surface abnormality in the oesophagus.

[0124] In one embodiment, detection of a specific binding partner of CSA and / or a specific binding partner of PNA in said fluid indicates adequacy of sample for collection of information for aiding diagnosis of BE or EOE, most suitably for aiding diagnosis of BE.

[0125] A soluble biomarker may be a soluble protein.

[0126] A soluble biomarker may be a soluble glycoprotein.

[0127] A soluble biomarker may be a specific binding partner of a lectin such as a specific binding partner of Cytisus scoparius (CSA) lectin.

[0128] A soluble biomarker may be a specific binding partner of a lectin such as a specific binding partner of Peanut Agglutinin (PNA) lectin.

[0129] Lectins recognise and bind particular sugar sequences in carbohydrates. Those carbohydrates may be independent molecules such as carbohydrates free in solution, or may be part of a larger molecule such as a glycoprotein. Suitably the specific binding partner of a lectin is a glycoprotein such as a soluble glycoprotein.

[0130] In one embodiment is described a method of collecting information useful in determining adequacy of an oesophageal sample comprising the assay of protein biomarkers as disclosed herein. P11112GBW0

[0131] In one embodiment is described use of the molecules set out in any one of Table 1 , Table 2 or Table 3 as biomarkers for detection of soluble protein(s) for aiding diagnosis of an oesophageal disease in a subject.

[0132] SAMPLE COLLECTION

[0133] Suitably sample collection is unbiased.

[0134] Suitably the sample is pan- esophageal. Suitably the sample comprises cells from the whole length of the oesophagus, such as from the stomach to the oro-pharynx.

[0135] Suitably the sample may comprise oesophageal brushings or surface cells. Oesophagal brushings may be obtained using an endoscope or by other means; suitably when the sample comprises oesophagal brushings they are obtained by non- endoscopic means.

[0136] More suitably the sample is obtained by non-targeted sample collection such as sampling the entire surface of the oesophagus rather than only targeting areas of suspected lesions (Barrett’s). Suitably the sample does not comprise an endoscopic biopsy.

[0137] Suitably the sample may comprise cells sampled from the entire oesophagal lumen. Suitably the sample may comprise both oesophagal and non-oesophagal cells. Suitably the sample may comprise oesophagal cells together with gastric cardia cells. Most suitably, the sample may comprise cells collected using a capsule sponge type sampling technique.

[0138] Especially suitable sampling techniques are described in the examples section.

[0139] Examples of suitable samples include oesophagal brushings (whether endoscopically or non-endoscopically obtained), samples obtained via balloon cytology, samples obtained via capsule sponge sampling. Most suitably, a sample comprises cells obtained via capsule sponge sampling.

[0140] A non-endoscopic capsule sponge device which has been approved by the Medical Health Regulatory Agency (Ref no: CI / 2007 / 0053) in the UK may be used for sample collection. The device consists of a polyurethane sponge, contained within a gelatin capsule, which is attached to a string. The polyurethane sponge is an abrasive material capable of collecting cells from the surface of the oesophagus. Suitably this should include being sufficiently abrasive to ensure sampling of any columnar cells P11112GBW0 which are a characteristic of Barrett’s oesophagus. The capsule is swallowed and dissolves within the stomach after 3-5 minutes. The sponge can then be retrieved by pulling on the string. This mode of sample collection is particularly suitable for use in the present invention.

[0141] In particular, the sampling is efficient because it is not directed to a particular site within the oesophagus but instead the sample of cells is taken across the entire surface of the oesophagus. This has the advantage of avoiding more invasive sampling techniques such as biopsy collection techniques which penetrate below the surface of the oesophagus. This has the advantage of providing unbiased sampling, because the cells are collected from the whole lumen of the oesophagus, and are therefore not biased by operator choice of particular areas for collection.

[0142] Typical sampling devices which may find application in obtaining a cell sample for use in the present invention include Cytosponge™ from MedTronic Limited, Building 9 Croxley Park, Watford, Herts, WD18 8WW, U.K., or EndoSign™ from Cyted Limited, 2 Falcon Road, Hinchingbrooke Business Park, Huntingdon, Cambridgeshire, PE29 6FG, U.K.

[0143] Suitably the cells are non-FFPE cells.

[0144] Proteins and glycoproteins are known to be secreted into the mucus layer of oesophagus with potential to provide vital information on oesophageal health. In one embodiment sampling of the surface of the oesophagus includes the mucous layer of the oesophagus in addition to the surface epithelial cells, which could be exclusively tested after collecting the capsule sponge sample including from proteins that are present in the mucous that leach into the aqueous liquid (such as preservative fluid).

[0145] AQUEOUS FLUID

[0146] The aqueous fluid may be any suitable medium for maintenance and / or preservation of the collected cell sample.

[0147] Suitably the aqueous fluid is a buffer suitable for preservation of mammalian cells.

[0148] Suitably the aqueous fluid is a buffer suitable for preservation of soluble biomarkers such as soluble proteins and / or carbohydrates such as glycoproteins.

[0149] Suitably the aqueous fluid is a preservative fluid.

[0150] The preservative fluid may be a preservative buffer.

[0151] The preservative fluid may be of any suitable composition, for example the preservative buffer may be RNAIater. More suitably the preservative buffer may be SurePath™ (BD P11112GBW0

[0152] SurePath™ Preservative Fluid) which is an ethanol-based fixative, which also has a small percentage of formaldehyde (0.2%).

[0153] The composition of BD SurePath™ Preservative Fluid is known in the art, see for example Fremont-Smith M, Marino J, Griffin B, Spencer L, Bolick D. Comparison of the Surepath™ liquid-based Papanicolaou smear with the conventional Papanicolaou smear in a multisite direct-to-vial study. Cancer. 102:269-279.

[0154] In summary, SurePath preservative fluid suitably comprises, or is composed of: 21.7% ethanol, 1.2% methanol, 1.1% isopropanol, 0.2% formaldehyde, buffered saline solution (balance suitably H2O). Additional components may be present.

[0155] Suitably BD SurePath™ Preservative Fluid is as supplied by Becton, Dickinson and Company of 1 Becton Drive Franklin Lakes, NJ 07417-1880, USA.

[0156] Suitably the aqueous fluid may comprise protease inhibitors.

[0157] Suitably the aqueous fluid may comprise inhibitors of microorganisms such as bactericide(s) and / or fungicide(s).

[0158] Suitably the aqueous fluid is sterile. Sterile has its natural meaning i.e. free from bacteria or other living microorganisms.

[0159] Suitably the aqueous fluid is not PBS (phosphate buffered saline). PBS is a combination of water and salts. It is a non-toxic solution used in many laboratories. PBS is typically composed of sodium chloride, potassium chloride, potassium dihydrogen phosphate, and disodium hydrogen phosphate, normally in concentrations that match the human body (isotonic). PBS is used in biological research to maintain a constant pH, typically around 7.4, for cell and tissue work. We provide comparative data showing the advantages of using Surepath compared to disadvantages of using PBS - we refer to the examples section for further details.

[0160] Suitably no washing steps are used. Suitably the aliquot of aqueous fluid is an aliquot of aqueous fluid into which the cells were placed directly. ‘Directly’ means immediately following collection. “Immediately following collection” means without any wash step(s) I rinse step(s). For example, when the cells are collected using a capsule sponge, suitably said retrieved capsule sponge is placed immediately (directly) into the aqueous fluid. Suitably said aqueous fluid is Surepath™. Suitably the sponge / cells are not placed into any intermediate fluid. Suitably the sponge / cells are not rinsed nor washed before placing into the aqueous fluid. The reason is that the inventors realised that they can assay proteins in the fluid, for example proteins secreted into the fluid. P11112GBW0

[0161] Washing or rinsing would remove or dilute those proteins, or require them to be further replenished by ongoing secretion. Better results are obtained when no wash steps or rinse steps are used. Suitably wash steps are excluded from the method of the invention. Suitably rinse steps are excluded from the method of the invention.

[0162] ALIQUOT

[0163] Suitably the aliquot of aqueous fluid is an aliquot of a supernatant fluid taken after settling or centrifugation of the cells to form a cell pellet.

[0164] Suitably the aliquot is cell free. The aliquot is considered ‘cell-free’ if there are no cells present, or if the number of cells present is ‘de minimis’ so that any reasonable person skilled in the art would consider it to be cell free.

[0165] Suitably the aliquot is acellular.

[0166] Suitably the aliquot may be filtered to remove cells.

[0167] In one embodiment suitably the aqueous fluid and / or the aliquot may be treated to lyse any residual cells.

[0168] Suitably the aliquot is sterile.

[0169] SUBJECT

[0170] The invention may be used to monitor subjects known to have BE.

[0171] The invention may be used to screen subjects not known to have BE.

[0172] The invention may be used to screen subjects suspected of having BE.

[0173] In one embodiment suitably the subject does not present with BE.

[0174] Suitably the subject does not present with BE lesions.

[0175] Suitably the subject is mammalian, most suitably human.

[0176] LATERAL FLOW TEST

[0177] In one embodiment, suitably the sample is analysed using a lateral flow test.

[0178] In one embodiment, detection of biomarkers is carried out with simple fluorescencebased LFA or screen printed electrodes for portable sensing.

[0179] The invention finds application in aiding detection of signs of disease (via the supernatant biomarkers).

[0180] EOSINOPHILIC ESOPHAGITIS

[0181] This condition may be characterised by detection of Eosinophils and / or mast cells. In one embodiment, this condition is characterised by detection of Eosinophils. In one P11112GBW0 embodiment, this condition is characterised by detection of mast cells. In one embodiment, this condition is characterised by detection of Eosinophils and mast cells in the same sample. MARKERS

[0182] As mentioned above, the inventors’ surprising selection of soluble biomarkers in the method of the invention represents a step change and a departure from the thinking in the prior art.

[0183] Suitably the biomarkers are soluble biomarkers. Suitably the biomarkers are soluble protein biomarkers.

[0184] Suitably the biomarkers are secretory proteins.

[0185] Suitably the biomarkers are secreted proteins.

[0186] Suitably the biomarkers are free in solution.

[0187] Suitably the biomarkers are not attached to any cell. Suitably the biomarkers are shed from the cell(s).

[0188] Suitably the biomarkers are detached from the cell(s).

[0189] Suitably for sample adequacy embodiments the biomarkers are specific binding partner(s) of the lectins set out below (Table 3). Suitably said specific binding partner(s) may be, or may comprise, carbohydrates. Suitably said specific binding partner(s) may be, or may comprise, glycoproteins.

[0190] Suitably the following markers are assayed: P11112GBW0 P11112GBW0 P11112GBW0 P11112GBW0 P11112GBW0 P11112GBW0 P11112GBW0 P11112GBW0 P11112GBW0

[0191] Unless otherwise apparent, accession numbers are for GenBank (GenBank, National Center for Biotechnology Information, National Library of Medicine, 38A, 8N805, 8600 Rockville Pike, Bethesda, MD 20894, USA. The Genbank accession numbers are provided with reference to the database as of the filing date of this application i.e. 10 Oct 2024. In case any further assistance is needed, the database release is Genetic Sequence Data Bank, Release number 261: 15 June 2024.

[0192] In another embodiment the invention relates to a method as described above wherein the soluble biomarker(s) comprise, or consist of, the group of soluble protein biomarker(s) in Table 1, and wherein presence of each of said soluble protein biomarkers in said aliquot is indicative of said subject having Barrett’s Oesophagus.

[0193] In another embodiment the invention relates to a method as described above wherein the soluble biomarker(s) comprise, or consist of, the group of soluble protein biomarker(s) in Table 2, and wherein presence of each of said soluble protein biomarkers is indicative of said subject having Eosinophilic Oesophagitis.

[0194] In another embodiment the invention relates to a method as described above wherein said one or more soluble biomarker(s) comprise, or consist of, the specific binding partner(s) of the lectin(s) as in Table 3, and wherein presence of each of said one or more specific binding partners is indicative of said sample being adequate to support collection of information useful in aiding diagnosis of an oesophageal disease.

[0195] Also described is a kit comprising reagents for determining the level of each of the markers in Table 1 , or each of the markers in Table 2, or each of the markers in Table 3 in a biological sample, such as an aliquot of aqueous fluid.

[0196] The examples section and / or the attached figures / drawings set out alternate marker combinations which may find application in the invention. Especially suitable marker combinations are as in Tables 1 to 3 above.

[0197] SEQUENCE HOMOLOGY / IDENTITY

[0198] Although sequence homology can also be considered in terms of functional similarity (i.e., amino acid residues having similar chemical properties / functions), in the context P11112GBW0 of the present document it is preferred to express homology in terms of sequence identity.

[0199] Sequence comparisons can be conducted by eye or, more usually, with the aid of readily available sequence comparison programs. These publicly and commercially available computer programs can calculate percent homology (such as percent identity) between two or more sequences.

[0200] Percent identity may be calculated over contiguous sequences, i.e. , one sequence is aligned with the other sequence and each amino acid in one sequence is directly compared with the corresponding amino acid in the other sequence, one residue at a time. This is called an “ungapped” alignment. Typically, such ungapped alignments are performed only over a relatively short number of residues (for example less than 50 contiguous amino acids). For comparison over longer sequences, gap scoring is used to produce an optimal alignment to accurately reflect identity levels in related sequences having insertion(s) or deletion(s) relative to one another. A suitable computer program for carrying out such an alignment is the GCG Wisconsin Bestfit package (University of Wisconsin, U.S.A; Devereux et al., 1984, Nucleic Acids Research 12:387). Examples of other software than can perform sequence comparisons include, but are not limited to, the BLAST package, FASTA (Altschul et al., 1990, J. Mol. Biol. 215:403-410) and the GENEWORKS suite of comparison tools.

[0201] In the context of the present document, a homologous amino acid sequence is taken to include an amino acid sequence which is at least 40, 50, 60, 70, 80 or 90% identical. Most suitably a polypeptide having at least 90% sequence identity to the biomarker of interest will be taken as indicative of the presence of that biomarker; more suitably a polypeptide which is 95% or more suitably 98% identical at the amino acid level will be taken to indicate presence of that biomarker. Suitably said comparison is made over at least the length of the polypeptide or fragment which is being assayed to determine the presence or absence of the biomarker of interest. Most suitably the comparison is made across the full length of the polypeptide of interest. The same considerations apply to nucleic acid nucleotide sequences.

[0202] DETECTION

[0203] Any suitable method of detection of biomarkers may be used.

[0204] Any suitable method of determination of presence of biomarkers may be used. Suitably the sample or aliquot is assayed to measure the concentration of biomarker in the aqueous fluid. The sample and / or the aliquot may need to be diluted to ensure that the concentration is within the appropriate range for accurate determination depending P11112GBW0 on the method of assay / measurement selected. Suitably dilution is made using a buffer appropriate for the assay / measurement method being used. For example the diluent may be water. For example the diluent may be preservative fluid as used at the point of sample collection. For example the diluent may be any buffer which is compatible with the biomarker(s) being assayed and which is compatible with the assay method used.

[0205] Suitably ELISA may be used.

[0206] Suitably ELLA may be used.

[0207] Suitably commercially available detection platform(s) may be used, for example a BioTechne™ Ella™ assay, such as the Simple Plex™ or multiplex automated ELISA system from BioTechne™ / ProteinSimple™ (614 McKinley Place NE, Minneapolis, MN 55413 USA), using either their ‘off-the-shelf’ cartridges or bespoke cartridges depending on the biomarkers being assayed. Bespoke cartridges may be purchased by ordinary commercial arrangements. For example, in one embodiment suitably GAL- 3 may be assayed using the same multiplex cartridge as CXCL8, TFF3 & FABP2. SPCLC-PS-009050 is the catalogue number from Biotechne / ProteinSimple - it is a custom cartridge. Suitably each of the markers for a particular indication such as BE may be assayed on a single multiplex cartridge. Further details are below including in Tables 1 - 3.

[0208] In one embodiment, the markers are assayed by ELISA test. In one embodiment, the markers are assayed by traditional antibody sandwich based secondary antibody detection methods. In one embodiment, the markers may be assayed using an automated approach such as Bio-Techne’s ELLA system (ibid.). In this embodiment, reagents for detection of alternative markers may be added into the same single cartridge for analysis, meaning that no additional process steps are required in order to read out additional markers from the same sample.

[0209] The method of the invention may be started from an appropriate sample, for example a container comprising a capsule sponge previously used to sample cells from the oesophageal lumen of the subject and preservative fluid. The preservative fluid is then withdrawn from the container and centrifuged to pellet the cells. The supernatant is then withdrawn from the centrifuge container, and is diluted to an appropriate concentration for ELISA detection and / or commercially available detection systems such as the Bio-techne cartridge system. P11112GBW0

[0210] In one embodiment, the centrifugation step to remove the cells might be omitted. In this embodiment, the sample may simply be the preservative fluid which may comprise cells which have been washed off the capsule sponge during transit in the preservative fluid container.

[0211] In either embodiment, the key departure from the prior art is focus on assaying what is in the fluid by (for example) ELISA. In contrast, prior art techniques have been based on assaying the actual cells collected.

[0212] MARKER DETECTION THRESHOLDS

[0213] Suitably the presence or absence of the marker is scored when assaying a sample. Suitably presence or absence may be a binary score such as a 1 or a 0 for “present” or “not present”. Suitably a marker may be scored as present if it is detected at a level which is elevated compared to the level found in a reference sample. Suitably, each marker level which is determined for a sample from a patient is compared to the level of that marker and a reference sample from a control subject, “level” means concentration of the marker in the sample or aliquot. Concentrations which are ‘not detected’ (e.g. none present, or below the threshold for detection) are scored as ‘absent’ or ‘O’.

[0214] The control subject may be a population of control subjects. The control subject may be a single control subject.

[0215] In one embodiment, a reference sample / reference aliquot from a control subject is assayed alongside each sample from subject of interest. In one embodiment, a previously determined reference value (e.g. concentration of the relevant biomarker(s)) may be used, and the value of a marker (e.g. concentration of the relevant biomarker(s)) determined for the sample / aliquot from a subject of interest is compared to that reference value.

[0216] Suitably presence of said one or more soluble biomarker(s) is determined as presence at an elevated level in said aliquot by comparison to a reference value.

[0217] In one embodiment said reference value is a previously determined reference value. In one embodiment said reference value is determined contemporaneously by assay of a reference aliquot in parallel to assay of the aliquot. P11112GBW0

[0218] Thresholds for the absolute concentration of a marker found in a particular sample may be statistically determined. For example, one particular marker might be scored as “elevated” if it is present at a level of 1 ,2X or greater than the concentration of that marker found in a reference sample from a control subject. The “threshold” at which a marker is determined to be present / absent or elevated / normal in a sample from the subject of interest may be varied using appropriate statistical methods. In one embodiment, suitably random forest model processing is used to determine the threshold(s) for the marker(s) of interest.

[0219] Depending on the application to which the invention is put, it may be that the skilled operator might vary the thresholds selected for example to enhance specificity, or for example to enhance sensitivity. In one embodiment, suitably the thresholds are chosen to optimise the specificity and sensitivity of the method overall.

[0220] By way of example, 1.2x is the perfect starting point - the average TFF3 supernatant is 1.2x higher in BE compared to heartburn controls, while other markers e.g. REG4 show a different increase such as a 5.9x increase.

[0221] A suitable control subject is a subject having heartburn. A most suitable control subject is a subject having heartburn, but not having oesophageal disease.

[0222] In one embodiment the invention finds application in methods for aiding diagnosis of Barrett’s Oesophagus.

[0223] In one embodiment, the invention find application in methods for determining sample adequacy for samples collected from the oesophageal lumen of a subject.

[0224] In one embodiment, the invention may be applied to methods for aiding detection of Eosinophilic Esophagitis (EOE).

[0225] SAMPLE ADEQUACY

[0226] Whether the sampling device such as capsule sponge has gone all the way into the stomach or not is referred to as adequacy of the sample. When the capsule sponge does not reach the stomach this can lead to a false negative results, for example it might have missed Barrett’s at the distal end of the oesophagus or it may have sampled some Barrett’s cells in a long segment but missed a small patch of dysplasia or cancer at the gastro-oesophageal junction. This is a clinically highly important P11112GBW0 distinction. Thus, embodiments of the invention addressing sample adequacy are useful in overcoming these technical problems.

[0227] When methods of the invention are used to aid diagnosis of a particular condition such as Barrett’s Oesophagus and / or Eosinophilic Esophagitis, normal standards of clinical statistical confidence levels should be employed. For example, the AUG (more specifically the ALIROC) will typically be approximately 0.95 or higher. This is an exemplary figure - lower AUG values may still be useful in the invention.

[0228] However, when applying the methods of the invention to sample adequacy applications, a much more flexible approach may be taken to AUG (ALIROC) values. This is because the only downside of an inadequate sample “slipping through” the test is that it is later discovered that the sample needs to be repeated. In contrast, when using the methods of the invention to assist in a clinical diagnosis, the possible downside of a sample “slipping through” is far greater - for example a false negative sample might risk missing a patient who actually has a condition in need of treatment. Therefore, when applying the invention to a test of sample adequacy, the skilled operator may choose to set the sensitivity / specificity levels to be tailored to their desired ‘cutoff’.

[0229] For example, in carrying out capsule sponge style testing in the prior art, there is currently approximately a 8-10% “misprocess rate”. This means a rate of inadequate sample of approximately 8-10%. In one scenario, the skilled operator could select threshold values giving a specificity / sensitivity level which would be expected to detect all of those 8-10% inadequate samples but may risk also misreporting some samples as inadequate which were in fact adequate. The possible downside of this is repeating sampling on a patient when that was not strictly necessary. However, it may be that a small number of potentially unnecessary repeat sampling events is an “overhead” which is set against the strong advantageous effect of eliminating 10% misprocess samples - wasting the labour and resource costs of processing those inadequate samples in order to not achieve a clinically useful result.

[0230] In another implementation, the skilled operator might choose the threshold values so as to lower the risk of any unnecessary resampling, but that might ultimately falsely read out some samples as “adequate” which, when processed, might then be determined as inadequate and require recall of the patient. P11112GBW0

[0231] It is for the skilled operator to decide for their particular application whether it is economically / clinically more desirable to choose the threshold values to err on the side of resampling a proportion of patients who might have already provided an adequate sample, or whether it is more desirable to err on the side of minimising repeat sampling at the risk of needing to recall a certain proportion of patients when it is later discovered that unfortunately their sample had been inadequate. Selecting the specificity / sensitivity thresholds according to the desired stringency of testing / balance of economic / clinical priorities for determining sample adequacy is a choice well within the capabilities of the skilled person given the guidance provided herein.

[0232] In cases where the Capsule Sponge device does not reach the stomach, there is a risk of ‘insufficient sampling’ of the distal esophagus. Consequently, samples with fewer than five columnar groups are categorised as low confidence or inadequate samples (Figure 2). The BEST2 trial reported an overall sensitivity of 79.9% in a per protocol analysis that included inadequate samples. However, when inadequate samples were excluded, the sensitivity remained consistently high across the participating centers, ranging from 91% to 98%. In the subsequent BEST3 trial, patients with a low confidence result were invited for a repeat test to improve sensitivity and increase confidence in the obtained results. Using the present invention, samples can be tested for adequacy straightaway and this directly improves the methods and clinical information collected.

[0233] Exemplary different variations in sensitivity / specificity for the adequacy test are as follows:

[0234] P11112GBW0 Test

[0235] Sensitivity 0-98

[0236] Specificity 8.72

[0237] 0D cut off > 0.05

[0238] CSA Test

[0239] Sensitivity 0.96

[0240] Path Specificity 8.83 report

[0241] 00 cut off > 0.16

[0242] Sensitivity 0-94

[0243] Specificity 6.94

[0244] In one embodiment the invention relates to an apparatus or system which is

[0245] (a) specifically adapted to analyse an oesophagal sample from a subject, wherein said analysis comprises:

[0246] (b) taking an aliquot of aqueous fluid previously contacted with oesophageal cells obtained from said subject; assaying said aliquot of aqueous fluid to determine the presence of one or more soluble biomarker(s) according to Table 3; said apparatus or system comprising an output module, wherein if presence of said markers are detected, then said output module indicates that said sample is adequate to support collection of information useful in aiding diagnosis of an oesophageal disease.

[0247] In one embodiment the invention relates to use for determining if a sample is adequate to support collection of information useful in aiding diagnosis of an oesophageal disease in the oesophagus of a subject, wherein said oesophageal disease is selected from the group consisting of Barrett’s Oesophagus and Eosiniphilic Oesophagitis, of a material which recognises, binds to or has affinity for certain soluble biomarkers, wherein the soluble biomarkers are as defined in Table 3. P11112GBW0

[0248] In one embodiment the invention relates to use as described above of a combination of materials, each of which respectively recognises, binds to or has affinity for one or more of said soluble biomarkers.

[0249] In one embodiment the invention relates to an assay device for use in determining if a sample is adequate to support collection of information useful in aiding diagnosis of an oesophageal disease in the oesophagus of a subject, wherein said oesophageal disease is selected from the group consisting of Barrett’s Oesophagus and Eosiniphilic Oesophagitis, which comprises a solid substrate having a location containing a material, which recognises, binds to or has affinity for certain soluble biomarkers, wherein the soluble biomarkers are as defined in Table 3.

[0250] In one embodiment, when applying the invention to determination of sample adequacy, the lectin PNA may be used. PNA detects glycoproteins such as 1-3 N-acetyl- galactosamine (Gaip3GalNAc epitopes).

[0251] In one embodiment, when applying the invention to determination of sample adequacy, the lectin CSA may be used. CSA detects glycoproteins such as Terminal GalNac glycans.

[0252] In one embodiment, a single lectin reagent may be used such as CSA. Using CSA as the sole reagent has the advantage of saving labour / costs, and still advantageously provides an AUC of approximately 0.95 on its own.

[0253] In one embodiment, a combination of lectin reagents may be used such as the combination of CSA and PNA. In this embodiment, advantageously the AUC is increased to 0.96.

[0254] It should be noted that the use of lectins themselves (i.e. the selection of lectins as a detection reagent) is itself a cryptic choice which was arrived by the inventors after detailed insight arising from the study of their discoveries. The conventional or prior art approach is to use antibodies as detection reagents. However, lectins have advantages such as cost advantages and simplicity of handling compared to antibody reagents. Moreover, there is a large pool of naturally occurring lectins which may be biochemically well understood candidates for possible detection reagents once the decision to use those molecules as detection reagents has been made. This choice was itself surprising to the inventors because it required a detailed analysis of the P11112GBW0 various possible glycoproteins which might have been targets for detection via the use of lectins as detection reagents. For example, the inventors began with a mass spectrometry approach to survey the molecules which they were able to find in the supernatant fluid being used as the sample. From this, the inventors then thoughtfully selected a panel of approximately 96 - 100 lectins as a start point. From here, the inventors carried out copious testing and analysis and were able to arrive at their teaching to use the particular subsets of lectins, or single lectin as disclosed herein.

[0255] Regardless of whether the invention is applied to detection of Barrett’s Oesophagus using protein / antibody assays, or whether the invention is applied to determination of sample adequacy using glycoprotein / lectin detection assays, the common technical approach which departs from the prior art is the choice to use an aliquot of the supernatant fluid as the sample for analysis / detection, which is in contrast to the prior art which is focussed on using the actual collected cells. It was surprising to the inventors to be able to find the informative protein / glycoprotein markers in sufficient abundance in this supernatant fluid to provide a robust and useful detection platform.

[0256] FURTHER EMBODIMENTS

[0257] In one aspect the invention relates to a method comprising: providing a sample from said subject, wherein said sample comprises cells collected from the surface of the subject's oesophagus; contacting said sample with an aqueous fluid; taking an aliquot of said aqueous fluid; assaying said aliquot of aqueous fluid to determine the presence of one or more soluble biomarker(s).

[0258] Suitably said method is a method of collecting information useful in aiding diagnosis of an oesophageal disease in a subject. Other aspects of the method are disclosed herein which have the same technical features as noted above.

[0259] In another aspect, the invention relates to a method of collecting information useful for detecting oesophageal disease comprising carrying out the steps as described above. In another aspect, the invention relates to a method of diagnosis of oesophageal disease comprising carrying out the steps as described above.

[0260] In another aspect, the invention relates to a method of aiding diagnosis of oesophageal disease comprising carrying out the steps as described above. P11112GBW0

[0261] In one embodiment, collected cells could be lysed and the lysate might form part of the sample. An advantage of this embodiment is to maximise the marker protein such as TFF3 available for detection. For example, such proteins can be made (transcribed I translated) inside the cell but may not yet have been secreted by lysing those cells, and so signal might be maximised by lysing the cells. In this embodiment, suitably a higher sensitivity might be achieved.

[0262] It is noted that gastric cardia cells do not express TFF3.

[0263] It is possible in some circumstances that salivary cells can express a marker such as TFF3. In principle, this might mean that the presence of salivary cells in the sample could contribute an occasional false positive via detection of a marker such as TFF3 which had not originated from a condition such as Barrett’s Oesophagus but instead had originated from a contaminating salivary cell. However, in contrast to prior art methods, the present invention advantageously does not rely only on a single marker. Thus, in this situation, the selection of the panel of markers being analysed can eliminate such possible false positives. For example, addition of markers to the panel which are known not to be expressed by salivary cells provides an excellent internal improvement to the specificity of the test.

[0264] SELECTION METHODS

[0265] The inventors thoughtfully selected biomarker candidates from existing bulk and single cell transcriptomic data based on their differential expression in BE compared to normal squamous and gastric tissue with protein level confirmation using immunohistochemical staining. Enzyme-linked immunosorbent assays(ELISA) were used to shortlist proteins from capsule sponge supernatants. The inventors trained a gradient-boosted tree model (lightGBM) on secretory protein data from a retrospective(N=161) and prospective(N=130) training cohort and independently tested it in an external prospective cohort(N=132). This resulted in a specific and small panel of soluble markers, such as secreted proteins, for detection of BE as defined in the claims.

[0266] RNA profiling identified goblet cells as the source of SPINK4, FABP2, and REG4; localized MEP1A to foveolar cells, and mapped CXCL8 to intraepithelial lymphocytes. Using sponge supernatant circulating expression levels of six-proteins, we established a diagnostic signature which achieved an AUROC of 0.87 (95% Cl, 81-93) when tested in an independent prospective cohort. This rapid assay could identify BE among P11112GBW0 symptomatic reflux patients with a positive predictive value of 95.6%, at 71 % sensitivity (95% Cl, 59-81) and 97% specificity (95% Cl, 90-99). Thus the invention provides a highly specific protein panel that can distinguish BE from reflux patients. This has industrial application as a quantitative diagnostic tool suitable for large-scale deployment (e.g. population screening (for example screening of heartburn patients)). The inventors disclose a mechanistic link between secreted protein biomarkers and the cell lineages for Barrett’s metaplasia. We demonstrate that these Barrett’s-goblet cell markers have maintained expression in dysplastic and neoplastic disease.

[0267] In more detail, the inventors first identified proteins secreted by cell lineages specific to Barrett’s esophagus from bulk and single cell RNA sequencing data. Their fidelity as physiologically relevant biomarkers was confirmed by qPCR and immunohistochemical staining. Specimens from more than 420 capsule sponges from multiple independent cohorts with BE and heartburn controls from across the UK were then studied to develop and test a novel assay that is useful as a diagnostic tool for BE screening.

[0268] Regarding at least BE, it is highly surprising from our findings of proteins detectable in the aqueous fluid (e.g. cell-free Surepath™ buffer) is the finding of proteins, both membranous and secretory in nature, in abundance. We undertook strenuous and robust testing, mining multi-omic data generated from our deep investigations, to first ascertain their differential expression pattern in our tissue of interest (Barrett’s) and their relative absence in cells making up the majority of the capsule sponge sampling (normal squamous oesophageal and normal gastric cells). We vigorously tested candidate markers using diverse approaches including qPCR and IHC to check expression at mRNA and protein level, and shortlisted those showing the biggest differences between Barrett’s and normal cells, with scientific evidence across all levels.

[0269] Quantifying proteins in a sample without a cell lysis step is extremely challenging in any cell or tissue culture setting. Even when staining for proteins expressed in samples in situ, one must first permeabilise the cells to access the proteins inside the cell. We investigated our dataset carefully to identify proteins secretory in character, and hypothesised that the goblet cells which are the hallmark of Barrett’s histopathology, would be a strong source leaking secretory proteins into the buffer. Despite it being unlikely for such proteins to permeate outside of cells fixed in Surepath™, we have robustly and consistently detected the presence of our panel of proteins in sponge supernatant samples using multiple, independent, clinically well-defined cohorts of P11112GBW0 patient samples. This is itself going against expectations, as it would be expected that this would be too technically demanding to yield clinically useful information.

[0270] We tested multiple approaches, including ‘upconcentrating’ the sample to increase the concentration of protein, but through our comparisons, realised that the simplicity of letting secreted proteins release into the buffer naturally over time, which keeps them more stable at room temperature compared to simpler buffers such as PBS, gives our approach a clinically stronger standing. This is again contrary to expectations.

[0271] Regarding at least EOE, similar to BE the current gold standard to diagnose and monitor EoE remains endoscopy which is invasive and uncomfortable, particularly in the affected EoE population which affects children and younger people. Turning to test sponge supernatant for inflammatory mediators secreted during the pathogenesis of this allergic inflammatory condition, and our ability to detect these proteins was challenging and contrary to expectations, given that inflammatory cytokines are acutely released and we are capturing only a snapshot of disease progression.

[0272] However, through repeatedly and robustly testing our supernatant samples, we show the surprising ability to consistently detect these proteins, at differential levels, between both histologically active EoE, as well as in patients in remission.

[0273] We underwent deep intellectual thought, brought in our expertise from the neurogastroenterology (gut-brain axis) field to look at markers of neuro-immune interaction - which is an angle to approaching EoE that has never been used before this invention. This unique approach has allowed us to shortlist unlikely candidates for monitoring EoE, including mast cell tryptase and ECP.

[0274] FURTHER ADVANTAGES

[0275] Without wishing to be bound by theory, the inventors arrived at the inspiration for this invention via a serendipitous / accidental route. As explained above, the whole focus in prior art techniques was about the collection of appropriate cell samples, and the handling and processing of those cells to maximise the reliable information that can extracted from them. Thus, the thinking in the art was always focussed on the cells and their preservation, and other elements of the process (such as the surrounding fluid) were regarded as “waste”. The inventors, as part of their research, happened to be measuring other molecules such as cytokines in pursuit of the detection of Eosinophilic esophagitis. As part of this analysis, they unexpectedly noticed that they were able to detect trefoil factor 3 (TFF3). From this observation, the inventors had the inspiration to undertake a whole discovery program to explore this anomalous observation. As part of this exercise, the inventors looked into RNA sequencing data P11112GBW0 for Barrett’s Oesophagus, and made intellectual choices such as a selection of secretory proteins possibly expressed by Barrett’s Oesophageal cells to see if information could be extracted from these surprising observations, as well as extensive multi-pronged statistical analysis coupled to prospective and retrospective testing of their ideas and insights. The inventors had the insight that secretory proteins present in the acellular supernatant could be the basis for a simple, scalable assay that does not rely on cytological reporting to triage patients at high risk for Barrett’s esophagus (e.g. patients with heartburn). The test can be used to identify patients in need of further intervention e.g. endoscopy.

[0276] Very briefly, the initial biomarker discovery process involved a number of distinct methodologies. For lectin-based biomarkers immunofluorescence staining and lectin array technology was utilised to screen BE and HB samples to look for markers which were differentially expressed, or more highly expressed in GC when assessing adequacy. For protein-based markers, RNA-seq data was interrogated to produce a list of candidate proteins with differentially expressed mRNA samples across sample types and then validated. Validation involved, among other things, validation at protein level on immunohistochemistry staining using matched patient biopsies to confirm selective expression of protein candidates only in BE, but not gastric cardia or normal squamous epithelium tissue. The inventors then moved on to validate biomarker (such as secretory protein) concentrations with ELISAs. The invention is based upon the surprising findings from these experiments.

[0277] It is an advantage of the present invention that more markers may be easily assayed in a single step. Using an extra marker in prior art cell based staining methods is an additional and laborious step. Moreover, it can involve the selection of different range of secondary antibodies for detection and / or different reagents to provide nonoverlapping signals when assaying different markers in the same FFPE section under a microscope. By contrast, since the present invention involves detection of molecular markers from an aliquot of aqueous fluid, such as preservative fluid, additional markers may be assayed in parallel in a streamlined and efficient manner which adds little or no extra labour to the analysis.

[0278] We refer to WO 2018 / 039422. Only MUC2 and CDX2 are mentioned for detection at protein level. No data or support are provided in this document. The present invention has been tested and validated on a panel of protein markers (REG4, TFF3, SPINK4, MEP1A, CXCL8, FABP2) that can differentiate Barrett's from heartburn with great accuracy. P11112GBW0

[0279] By contrast in the present invention, suitably a capsule sponge is used to collect cells into a preservative buffer (most suitably Surepath™ ). This has the advantage that it can be shipped at room temperature to site(s) separate from the collection site, and still support the quantification of protein using immunoassays.

[0280] In WO 2018 / 039422, the main focus is on nucleic acids which require a laborious process to extract and measure via PCR. No protein / ELISA data seems to be included in the figures and the AU ROC curves are trained on the PCR.

[0281] By contrast, in the present invention, the cells are not washed; the collected cells (e.g. capsule sponge comprising collected cells) is placed directly into an alcohol preservative (most suitably Surepath™); the proteins taught to assay herein have typically been secreted into the solution without any wash step. These proteins could be from surface mucous, proteins still being secreted from live cells, or from cells that lyse on collection. This is important because the inventors considered extra steps to wash or lyse the cells but surprisingly designed a method in which these extra steps are not performed; it was surprising to the inventors that the proteins are present in the buffer with no steps to collect proteins. This advantageously simplifies the method and reduces the labour and cost.

[0282] In one embodiment suitably the marker(s) do not include one or more of; more suitably the marker(s) do not include any of: MUC2 (e.g. NM 002457.3), MUC5AC (e.g. NM 001304359.1), CDX2 (e.g. NM 001265.4), BMP4 (e.g. NM 001202.4) and villin (e.g. NM_004604).

[0283] Regarding CDX2, it is unsuitable to detect in supernatant as it is localised to the nucleus, so release would require breaking through not only the cell membrane, but the nuclear membrane too. It is not secreted and cannot diffuse into the extracellular environment.

[0284] Regarding BMP4, it is unsuitable for the invention as it is also increased in inflamed squamous epithelium, so would also give a high reading in conditions such as erosive esophagitis, resulting in a high rate of false positives in the population being tested. Regarding MUC2, it is unsuitable for the invention at least for BE - see Examples section.

[0285] Further particular and preferred aspects are set out in the accompanying independent and dependent claims. Features of the dependent claims may be combined with features of the independent claims as appropriate, and in combinations other than those explicitly set out in the claims. P11112GBW0

[0286] Where an apparatus feature is described as being operable to provide a function, it will be appreciated that this includes an apparatus feature which provides that function or which is adapted or configured to provide that function.

[0287] Some of the data presented in support of this patent application features staining of whole cells. However as disclosed herein, the methods of the invention are specifically focussed on detection from the surrounding fluid (and not the cells themselves as in the prior art). It should be noted that this supporting data is provided showing that the markers which have been selected by intellectual choices made by the inventors are indeed physically expressed and found in the cells of interest. Therefore, the cell staining data presented in this document are shown for their illustrative value validating that the markers are indeed associated with the cells of interest at source.

[0288] BRIEF DESCRIPTION OF THE DRAWINGS

[0289] Embodiments of the present invention will now be described further, with reference to the accompanying drawings, in which:

[0290] Figure 1 shows a schematic of the biological material sampled by the Capsule Sponge Figure 2 shows a schematic depicting adequate and inadequate sampling of the Capsule Sponge

[0291] Figure 3 shows Lectin CSA and PNA staining on biopsy sample

[0292] Figure 4 shows Lectin CSA staining on Capsule Sponge histopathology samples

[0293] Figure 5 shows a schematic depicting Direct Lectin ELISA configuration depicting the main steps involved

[0294] Figure 6 shows box and whisker plot of CSA optical density values in an inadequate vs adequate cohort and respective OD values. ROC curve of CSA QC marker cohort. Figure 7 shows repeat gastric QC cohort experiment to show robustness and repeatability.

[0295] Figure 8 shows 2x2 Tables demonstrating how altering test cut off values influences sensitivity and specificity

[0296] Figure 9A shows plots of secretory protein data from a test cohort; 9B shows a diagram and values; 9C shows ROC curves.

[0297] Figure 10 shows plots of secretory protein data from prospective validation.

[0298] Figure 11 shows a heatmap on a matched cohort of samples (underlying data as for Figure 10).

[0299] Figure 12 shows diagram summarising data.

[0300] Figure 13 shows bar charts of SHAP scores.

[0301] Figure 14 shows a schematic depicting sample adequacy workflows. P11112GBW0

[0302] Figure 15 shows plots and ROC curves.

[0303] Figure 16 shows photographs of stained cells and a plot.

[0304] Figure 17 shows a graph of performance of supernatant biomarker assay

[0305] Figure 18 shows a graph of ROC curves of biomarker assay for BE (6 marker panel)

[0306] Figure 19 shows a graph of ROC curves of biomarker assay for BE (6 marker panel)

[0307] Figure 20 shows plots of individual biomarker performances for BE.

[0308] Figure 21 shows a graph of Surepath™ vs PBS. Top line at start (red) = Surepath™ at RT. 2ndline from top at start (green) = Surepath™ at 4C. 3rdline from top at start (blue) = PBS at RT. 4thline from top at start (purple) = PBS at 4C.

[0309] Figure 22 shows bar charts.

[0310] Figure 23 shows plots: Shortlisted biomarkers from bulk RNA-seq of esophageal and gastric biopsies. (A) Bulk RNA-seq identifies 12 transcripts upregulated in NDBE versus NE and NG (ITLN1-CXCL8). Shortlisting prioritized low NE / NG signal, maintenance across NDBE and dysplasia, and large normal-to-pathology effect sizes. (B) qPCR validation of top candidates in an independent biopsy cohort (n=70) confirms selective overexpression of TFF3, REG4, and SPINK4 in NDBE, DBE (LGD / HGD), and IMC / EAC versus NE and NG; CXCL8 is elevated in cancer groups. qPCR data were Iog2-transformed; statistics by Kruskal-Wallis with Dunn’s comparisons. Abbreviations: NDBE, nondysplastic BE; NE, normal esophagus (squamous); NG, normal gastric;

[0311] LGD, low-grade dysplasia; HGD, high-grade dysplasia; IMC, intramucosal carcinoma; EAC, esophageal adenocarcinoma.

[0312] Figure 24 shows (A) expression plots and (B) photographs of IHC stained cells: Cell of origin of shortlisted secretory biomarkers. (A) Expression of the most promising 6 BE markers across Barrett’s oesophagus (BE) compared to normal squamous epithelium (NE), gastric cardia (NG), and normal duodenum (ND) in the scRNAseq dataset (Cohort 2). (B) Representative staining of esophageal biopsies and capsule sponge sample for the top three BE markers, including TFF3 staining on matched samples as a positive reference. GC: gastric cardia. Scale bar represents 50pm.

[0313] Figure 25 shows Cell-free biomarker performance as a panel in a test cohort. A) Concentration of TFF3, REG4, MEP1A, SPINK4, FABP2 and CXCL8 in capsule sponge supernatant in BE patients compared to heartburn controls in training cohort. Mann-Whitney II test indicated statistical significance in the quantified TFF3, REG4, MEP1A, SPINK4, FABP2, and CXCL8 values (p = 5.81X10’10, p = 4.98x10’12, p = 1.17 x10-2, p = 5.92 x10'7, p = 1.45 x 10'13, p = 1.94 x10'4) respectively. The concentrations for each biomarker were Iog10-transformed to normalise data distribution. Boxplots display the median, interquartile range, and whiskers extend to 1.5 times the IQR. B) Heatmap displaying z-score normalised expression values demonstrating mutual P11112GBW0 exclusivity of biomarkers in the prospective arm of the training cohort. C) Boxplots demonstrating TFF3 supernatant concentrations in heartburn samples compared to histology-negative BE cases and histology-positive BE cases, with significant increase in TFF3 supernatant in histology-deemed false negative cases. (D) Cross-validated receiver operating characteristic (ROC) curves for multiple classifiers trained on supernatant protein data. Mean cross-validated ALIROC values were: LightGBM 0.83, XGBoost 0.82, Random Forest 0.83, Logistic Regression 0.75, and Lasso 0.75. Figure 26 shows Secretory protein panel performance in an external prospective test cohort. A) Boxplots showing distribution of biomarker concentrations in capsule sponge supernatant in an independent prospective cohort. Mann- Whitney U test indicated statistical significance in the quantified TFF3, REG4, MEP1A, SPINK4, FABP2, and CXCL8 values (p = 1x10’2, p = 8.45x10’5, p = 6.9x10’1, p = 8.45x1 O’10, p= 1.9x10’15, p = 8.46x1 O'6respectively). B) AU ROC for machine learning algorithms tested on protein data from the external test cohort. C) Confusion matrix on LightGBM model from external test cohort. D) SHAP values for average impact of each biomarker on the ML algorithm’s performance.

[0314] Figure 27 shows bar charts of marker performance in EOE.

[0315] Figure 28 shows graphs of marker performance in BE. AUC for diagnostic model with 7-marker panel including Gal-3 with AUC 0.81 (Fig.28A) compared to 5-marker panel with AUC 0.81 excluding Gal-3 and TFF3 (Fig. 28B), and 6-marker panel with the 6 top markers (Fig 28C), AUC 0.92.

[0316] EXAMPLES

[0317] In these examples the aqueous fluid is BD SurePath Preservative Fluid.

[0318] Example 1 : A Novel Liquid Biomarker Gastric Adequacy Test for the Capsule Sponge

[0319] Here we identify and validate a novel lectin biomarker using the cell-free liquid fraction of the capsule sponge sample to classify samples which have reached the stomach (adequate) compared to those which did not reach the stomach (inadequate).

[0320] Methods:

[0321] Triplet endoscopic biopsy samples from n=5 patients, of Normal Esophagus (NE), Barrett’s Esophagus (BE) and Gastric Cardia (GC) were screened with 20 distinct fluorescein isothiocyanate conjugated lectins, each with a specific carbohydrate binding preference, using immunofluorescence staining. Lectins which demonstrated differential binding in favour of GC were selected for further testing to determine P11112GBW0 whether biomarkers could be detected in the liquid fraction of the capsule sponge sample. We developed a direct Enzyme Linked Lectin Assay (ELLA) using HRP conjugated lectin to enable retrospective validation of 126 liquid based patient samples (18 histopathologically inadequate and 108 adequate samples). Data underwent statistical analysis using the Mann-Whitney U test and AUROC analysis.

[0322] Results:

[0323] Immunofluorescence staining demonstrated that Cytisus scoparius (CSA) lectin and Peanut Agglutinin (PNA) lectin showed preferential binding to GC tissue samples, primarily in the apical mucin caps and foveolar cells of the columnar epithelium. After ELLA optimisation horseradish peroxidase conjugated CSA lectin was shown to successfully bind to the liquid fraction of the capsule sponge sample. Using the n=126 patient sample cohort lectin CSA demonstrated excellent classification between inadequate and adequate samples, AUROC 0.94 (0.88-1.0 95% Cl, p<0.0001) with sensitivity of 87.96 (80.49 - 92.83 95% Cl) and specificity of 94.44 (74.24 - 99.72 95% Cl) at the selected threshold, Mann Whitney U test (p<0.0001).

[0324] Conclusion:

[0325] We successfully demonstrate on a large retrospective cohort that lectin CSA can accurately classify the adequacy of capsule sponge samples, using the cell-free liquid based portion of the sample alone. This could reduce overall recall rate, optimise laboratory processing and reduce costs.

[0326] Example 2: Early Detection of Barrett’s Oesophagus Using a Secretory Protein Biomarker Panel from a Capsule Sponge

[0327] We demonstrate using the acellular portion of the sponge sampling to test for secretory protein biomarkers. These proteins are released from the cells collected by the sponge into the preservative fluid and we show that they can be used for a more rapid diagnostic that does not rely on cytological pathology reports.

[0328] Methods:

[0329] Biomarker candidates were selected from previously published bulk and single cell- RNA-sequencing (scRNA-seq) data based on their differential expression in BO compared to normal oesophagus and gastric cardia, and subsequent validation at protein level with immunohistochemical staining. Enzyme-linked immunosorbent assays (ELISA) were used to quantitatively detect trefoil factor-3 (TFF3) (Bio-Techne, #SPCKB-PS-000530, USA), regenerating family member 4 (REG4) (Biorbyt, orb180655, UK), Meprin A subunit alpha (MEP1a) (Abeam, ab26727), and serine peptidase inhibitor kazal type-4 (SPINK4) (Abbexa, abx528747) using manual sandwich-ELISA from capsule sponge supernatant samples. Six different classification P11112GBW0 algorithms were tested to distinguish between BO and heartburn controls based on the cell-free protein quantification of protein biomarkers either used singly or in combination to maximise performance.

[0330] Results :

[0331] Data from scRNA-seq analysis highlighted REG4, TFF3, SPINK4, and MEP1a as strong candidate biomarkers related to the characteristic BO goblet cells which are secretory in nature. TFF3 and REG4 were individually detected at a significantly higher concentration in BO sponge supernatants (n=56) compared to controls (n=53) (p = <0.0001, Mann-Whitney test). While the average TFF3 concentration for BO samples was 1.2 times higher (62.5ng / ml) compared to controls (33.1 ng / ml), the average concentration of REG4 in BO samples was 5.9 times higher (2.1 ng / ml) compared to heartburn (0.36 ng / ml). REG4 protein concentration positively correlated with Barrett’s segment length (p < 0.0001 , Spearmen’s correlation). A significant difference was also observed with SPINK4 in BO (n=34) compared to controls (n=38) (p = 0.01). Gaussian NB emerged as the best performing prediction model with a specificity of 0.94, sensitivity of 0.70 and area under the curve (AUG) of 0.86 when testing TFF3 and REG4 combined. Although MEP1a concentration could not significantly differentiate between BO and heartburn alone (p = ns), adding MEP1a as an additional marker to the TFF3 / REG4 prediction model reached a ALIROC of 0.85 (p <0.0001, logistic regression).

[0332] Conclusions:

[0333] A panel of secreted proteins in the Cytosponge supernatant is used as non-invasive biomarkers for earlier detection and risk stratification of patients with BO.

[0334] Example 3: Secretory Proteins for Early Detection of Barrett’s Oesophagus: Looking Beyond the Cells

[0335] We demonstrate using the acellular portion of the Cytosponge sampling to test for secretory protein biomarkers for a more rapid bedside diagnostic that does not rely on cytological pathology reports.

[0336] Enzyme-linked immunosorbent assays (ELISA) were used to detect TFF3 (Bio-Techne, #SPCKB-PS-000530, USA) on the automated Ella system, and regenerating family member 4 (REG4) (Biorbyt, orb180655, UK) using a manual sandwich-ELISA. REG4 was identified as a promising secretory biomarker candidate with bulk RNA- sequencing, and subsequent validation data from the lab at protein level with immunohistochemical staining. The candidates were measured in a pilot cohort of supernatant samples from 56 BO patients and 53 controls. TFF3 and REG4 were both detected at a significantly higher concentration in BO Cytosponge supernatants P11112GBW0 compared to controls (p = <0.0001, Mann- Whitney test). Six different classification algorithms were tested to distinguish between BO and healthy controls based on the cell-free protein quantification of TFF3 and REG4. Gaussian naive bayes emerged as the best performing prediction model with a specificity of 0.94, sensitivity of 0.70 and area under the curve (AUG) of 0.85.

[0337] Thus, secreted biomarkers in the Cytosponge supernatant can serve as non-invasive biomarkers for a rapid diagnostic test for earlier detection and risk stratification of patients with BO.

[0338] Example 4 - Sample Adequacy

[0339] Lectins were tested first using immunofluorescence (IF) of tissue biopsies to assess their tissue specific localisation. 10 different lectins with known binding sites for Galactose / GalNac or Lactose sugars, exhibited specific binding to either both GC and BE tissues, or just to GC tissues. IF staining of lectins OSA and PNA indicated a specific binding to GC tissue, primarily in the apical mucin caps and the foveolar cells of the columnar epithelium (Figure 3). The staining pattern also clearly indicates the secretory nature of these markers, likely to be due to mucous secretions observed in the GC / BE segment.

[0340] Out of these, lectin CSA was chosen to be evaluated further based on its enhanced specificity to gastric cardia tissues. IF staining in Capsule Sponge samples (n=5) confirmed this specificity even in the heterogenous sampling. Figure 4 depicts representative IF-stained images of lectin CSA. No CSA-specific FITC (green) staining is observed in squamous cells collected by the Capsule Sponge, whereas a strong and intense binding to gastric columnar cells were observed. These findings verified that the lectins bind to glycoproteins originating exclusively from the cells collected from gastric cardia.

[0341] Thus, secretory markers from these cells could also demonstrate an increased binding to lectin CSA, in comparison to those secreted from other cell types. To test this hypothesis a cohort of Capsule Sponge supernatant samples were used consisting of both adequate and inadequate sampling of the gastric columnar cells.

[0342] ELISA validation Once the cellular specificity of the lectin was established, their presence in the cell- free component of the Capsule Sponge was next evaluated. A cohort was assembled of patient Capsule Sponge supernatant samples (n=126) P11112GBW0 consisting of 108 adequate, and 18 inadequate samples (didn’t reach the stomach and no gastric cardia cells) based on their respective cellular pathology classifications.

[0343] A challenge with evaluating lectins in an ELISA format is that the standard sandwich format which can be used in case of other antibody-based markers are not readily available commercially for lectins. Due to this constraint, a new manual lectin ELISA or an ‘ELLA’ format was setup in the lab based on protocols published in literature. A schematic of the lectin ELISA mechanism is presented in Figure 5. Briefly, a direct ELISA format is employed where antigens are first attached to the well plate. The antigens constitute the Capsule Sponge supernatant containing glycoproteins. Next, a BSA washing step was carried out to limit non-specific interactions, followed by incubation with lectin ECA conjugated with the enzyme HRP. This allowed for detection of glycoproteins that are specific to lectin ECA in the well plate, based on standard TMB-based colour development, and analysis.

[0344] To quantify the lectin-based biomarker in the supernatant samples, the customised ELISA platform described above, incorporating an HRP-tagged Lectin CSA was performed. The quantification of lectin CSA binding yielded a remarkable distinction between adequate and inadequate samples (p<0.0001). Figure 6 shows the ROC curve with an AUC of 0.95 indicating the biomarker’s performance for accurately classifying inadequate cases.

[0345] The robustness of this biomarker was further demonstrated when a similar result was obtained when repeating the ELISA assay on the same cohort in a separate experiment (Figure 7). This stability in signal across ELISA plates further indicates the suitability of using a simple standardised threshold-based approach to make the classification.

[0346] For clinical use the cut-off can be tuned to maximise sensitivity or specificity depending on whether you want to catch all inadequates and repeat some tests unnecessarily (false positives included), or let a small number of inadequates proceed to full processing (false negatives included). The performance of this marker in terms of its sensitivity and specificity is represented based on different thresholds which could be used with this assay are shown below (Figure 8).

[0347] Additionally, within the cohort of inadequate samples there were 3 medium segment length BE cases. The CSA ELLA was able to classify these all as inadequate (at the >0.05 cut off), suggesting that the test result can also distinguish gastric cells from Barrett’s cells. This has great clinical importance to ensure that distal Barrett’s changes are not missed - the table below shows CSA ELLA Results for inadequate samples with significant known BE segments: P11112GBW0

[0348] This approach presents an opportunity to provide an assessment of the Capsule Sponge sample adequacy prior to the cellular sample processing. This could save considerable time, cost and effort compared to the current sample processing workflow and help to accurately identify cases requiring a repeat Capsule Sponge, thus improving the overall diagnostic accuracy.

[0349] Additionally, development of portable biosensors to detect these lectin-based biomarkers could further enable point-of-use monitoring of sample columnar adequacy which could further streamline the workflow for Capsule Sponge-based sample testing. Figure 14 shows a summary of workflows.

[0350] Retrospective Validation of Gastric quality control lectins was carried out.

[0351] When integrated into a simple univariate logistic regression model the AUC was 0.9522 with sensitivity of 87.96 and specificity of 94.44 at the optimal threshold. Lectin PNA also had a significant difference between medians of inadequate (0.067) and adequate samples (0.326) Mann-Whitney II test p<0.0001. Integration into a univariate logistic regression model resulted in an AUC of 0.801.

[0352] When the models were combined using a multiple logistic regression model there was a non-significant increase in AUROC from 0.95 to 0.96, suggesting that PNA may not add significant value as a combined test, but does increase technical confidence that lectins are able to function as gastric adequacy biomarkers. Figure 15 shows box and whisker plots for OD values of CSA and PNA for adequate and inadequate samples and integration into distinct logistic regression models. AUC was 0.9522 (0.88-1.0 95% Cl, p<0.0001) with sensitivity of 87.96 (80.49 - 92.83 95% Cl) and specificity of 94.44 (74.24 - 99.72 95% Cl) at the optimal threshold. Repeat retrospective Validation of Gastric quality control lectins gave AUC 0.94 (0.87-1 , 95%CI), P<0.0001.

[0353] PNA alone had an AUC of 0.81. When CSA + PNA were combined, the AUC went up from 0.95 (CSA alone) to 0.96 (CSA + PNA).

[0354] Example 5 - Barrett’s Oesophagus

[0355] Figure 9A shows secretory protein data from a test cohort for panel of 7 markers for aiding detection of BO / BE.

[0356] Figure 9B shows Secretory protein ELISA data- test cohort; Figure 9C shows Metaclassifier for all markers combined had AUC of 0.95. P11112GBW0

[0357] The AUG of all biomarkers together: 0.95

[0358] The AUG of TFF3 + REG4: 0.87

[0359] Figure 17 shows a graph highlighting performance of supernatant biomarker assay on samples that are missed by the current capsule sponge processing method of staining for TFF3 on the collected cells. In this example the biomarker was TFF3 - the secretory TFF3 in the supernatant picked up samples that were missed by IHC staining as shown in graph, and these samples had significantly higher acellular TFF3 than the heartburn controls.

[0360] Figure 10 shows Secretory protein ELISA data- prospective validation. Barrett’s n = 53, Heartburn n = 77 AUG 0.89.

[0361] Figure 11 shows Secretory protein ELISA data- prospective validation cohort; the same prospective data as Figure 10 are presented as a heatmap on a matched cohort of samples, whereby data has undergone Z-score normalisation which transforms the data to have a mean of 0 and SD of 1, ensuring different features in the dataset are on the same scale. Data currently being tested in Random Forest model.

[0362] Figure 12 shows summary of data. Figure 12 demonstrates that we have a 1:1 split between controls and disease cases, while in the prospective internal validation, the split is closer to 2:1 due to the nature of the real-population setting. However, since our numbers are relatively small, our model was over-fitted to the retrospective training set, so the shift in balance going to 2:1 controls : cases means the model performs less well in the prospective validation. Thus a third cohort is analysed, (the 'test' cohort on the right-hand side in Figure 12) to test our model in an evenly split, independent prospective cohort, using both previous cohorts to re-train the model. Sample numbers are prepared to get to the numbers quoted in the blue box of Figure 12.

[0363] We show the empirical supernatant biomarker list for Barrett’s: TFF3, REG4, FABP2, MEP1A, CXCL8, SPINK4. (6 markers).

[0364] Gal-3 was removed in the prospective cohort, causing a drop in AUG from 0.95 to 0.93, but can advantageously be included to make a set of 7 markers with an AUG of 0.95. Figure 13 presents data showing these.

[0365] We refer to the graphs of figure 18 which show the performance of the 6-marker panel for BE, with AUG 0.93 in the retrospective training cohort when GAL-3 was removed from the model.

[0366] We refer to the graphs of figure 19 which show the performance of the 6-marker model in the independent prospective cohort used for internal validation of the model, with AUG 0.89. P11112GBW0

[0367] We refer to the plots of figure 20 which show the performance of the individual markers with a standard Mann- Whitney II test for statistical comparison, in the same prospective validation cohort.

[0368] Example 6 - Eosinophilic Oesophagitis

[0369] We show proof-of-concept immunohistochemistry data on Cytosponge tissue sections to show capture of immune cells (stained with brown dye) in EoE.

[0370] We can also pick up immune cells on Barrett’s cases using the capsule sponge, and we have detected their inflammatory cytokines in the supernatant.

[0371] We disclose to detect secretory inflammatory proteins in EoE supernatants.

[0372] N = 21; Heartburn = 6; NDBE = 8; Dysplasia = 7; EoE = 2.

[0373] We refer to Figure 16.

[0374] Example 7 - Application of Invention to Aiding Diagnosis of Barrett’s Oesophagus

[0375] In this example we demonstrate a method of collecting information useful in aiding diagnosis of an oesophageal disease in a subject comprising: providing a sample from said subject, wherein said sample comprises cells collected from the surface of the subject's oesophagus; contacting said sample with an aqueous fluid.

[0376] In this example the aqueous fluid is SurePath™.

[0377] After contacting for 60 minutes, the sample is centrifuged at 2500 rpm, and remaining cells on the sponge are washed twice more and centrifuged twice more to ensure all cells have been collected.

[0378] The cell-free supernatant is then collected for secretory protein analyses.

[0379] (Optionally, the cell pellet is processed into FFPE blocks as per standard procedures, which is an advantage of the invention since prior art methods require use of the cells in detection.)

[0380] An aliquot of the supernatant aqueous fluid is removed by pipette. P11112GBW0

[0381] Said aliquot of aqueous fluid is assayed to determine the presence of one or more soluble biomarker(s):

[0382] (i) TFF3 (ii) REG4

[0383] (iii) FABP2

[0384] (iv) MEP1A

[0385] (v) CXCL8; and

[0386] (vi) SPINK4 In this example, the assay is by multiplex assay comprising TFF3, FABP2 and CXCL8 is S PC KA- PS-009050 by Biotechne, REG4 assay (orb180655) by Biorbyt, SPINK4 assay (abx528747) by Abbexa, MEP1A (Ab267627) by Abeam.

[0387] Presence of each of said soluble protein biomarkers (i) to (vi) in said aliquot is indicative of said subject having Barrett’s Oesophagus.

[0388] In this example, the method successfully picked up Barrett’s Oesophagus in a 73 year- old male with a COM2 short-segment lesion, a 74 year-old male with a COM2 short segment lesion, a 52 year-old male with a COM3 short segment lesion, a 78 year-old female with a C1M3 segment lesion, a 67 year-old female with a C0M1 short-segment lesion who were missed on the current pipeline involving pathological assessment of TFF3.

[0389] Example 8 - Comparative Data

[0390] By comparison to prior art methods, the invention offers superior performance.

[0391] The data from example 7 are provided in comparison to prior art methods in the table below. P11112GBW0

[0392] In the above table: “Risk prediction with current test on cells” means prior art I known methods staining cells (i.e. immunohistochemistry with skilled operator interpretation). The comparative data in this table highlight that staining for TFF3 on the collected cells according to the prior art has missed these patients (‘negative’). However, detecting soluble biomarker in the aqueous fluid I supernatant (e.g. exemplary data here for secretory TFF3 - “Model's prediction based on supernatant biomarker (TFF3)”) has accurately indicated disease (‘BE’).

[0393] This exemplary data highlights the benefit of the invention that discordant Barrett's oesophagus cases can be picked up by the assay of soluble biomarkers (such as secretory TFF3 used to exemplify here - but with reference to the panel of markers discussed throughout). This is exemplary data - given that this marker in its secretory form adds most value to the meta-classifier model, the inventors disclose that this shows superior performance of using the aqueous fluid I supernatant (e.g. acellular portion of the sponge sample) compared to staining for the cells (as currently / previously done in the prior art).

[0394] In addition, we provide further data - secretory MEP1A data - for 2 of the cases shown above: P11112GBW0

[0395] This is further validation of the superior performance of the methods of the invention by comparison to the prior art as explained above.

[0396] Example 9 - Comparative Data

[0397] In this example we show advantages of using Surepath™ as the aqueous fluid. We present a comparison to Phosphate Buffered Saline (PBS) as the aqueous fluid.

[0398] We measured concentration of recombinant TFF3 protein in Surepath™ vs PBS over a period of 30 days. We found that Surepath stabilised protein, and stabilised it over a long period of time. We found that use of Surepath™ also enabled storage at room temperature, and with only a negligible reduction in protein concentration.

[0399] We refer to Figure 21. PBS is inferior for sample collection for protein-based analyses. Levels of degradation of protein in PBS mean that in patient samples, protein of interest would be degraded resulting in poor detection, or no detection.

[0400] Advantageously, using Surepath™ preserves protein and thereby enhances detection.

[0401] Example 10 - Exclusion of Markers from Panel

[0402] Arriving at the panel was challenging for the inventors. We refer to Figure 22.

[0403] Three markers which were originally included in our longer list of candidates based on their differential expression patterns which were later excluded as they did not show any difference in Barrett’s samples compared to heartburn samples.

[0404] For example, MLIC2 appeared to be a possible ideal candidate marker based on initial analysis and shortlisting. It is also a secretory protein. However, when tested via ELISA, no difference between heartburn and Barrett’s samples was detected (n=26). P11112GBW0

[0405] Further research revealed that MLIC2 is an intact, large gel forming mucin, with extensive glycosylation, making it poorly soluble and potentially more difficult to detect. Thus the intellectual decision was made to remove MLIC2 from the panel at least for BE.

[0406] When GAL-3 was tested in the machine learning algorithm, it had the poorest performance compared to markers that we include in the final panel of 6. It is noted that the diagnostic algorithm was tested on the combination including Gal-3 - but other markers were excluded earlier in the procedure as they showed inadequate differences between disease vs control.

[0407] Example 11 - Selection of Markers in Panel, Shortlisting and Evaluation

[0408] RNA-Sequencing and Biomarker Shortlisting

[0409] Study Cohorts

[0410] This study analysed data from 160 biopsy samples from publicly available bulk and single cell RNA-sequencing (scRNAseq) datasets published by our group from two independent cohorts (Secrier et al Nature Genetics 2016 48:10, Nowicki-Osuch et al 2021 Science 373(6556):760-767, Jammula et al 2020 Gastroenterology 158(6): 1682- 1697. e1). Differential expression analysis of each individual disease phenotype over all other phenotypes was performed on counts using the edgeR R package with a cut-off of a log fold change greater than 1.5 and adjusted p value less than 0.01 in NDBE samples compared to normal esophageal epithelium and the proximal stomach, with a limit of 0.8 being set for both. The selection criteria for secretory biomarker candidates to validate at protein level included an overexpression of each transcript in NDBE, with maintained expression in dysplastic BE and absence of expression in NE and NG tissue.

[0411] We used the scRNAseq dataset to map their likely candidate cell of origin to further inform the selection of BE markers. The edgeR R package was used to perform differential expression testing on raw UMI counts per cell type, patient, and condition while accounting for variation between patients, as detailed previously.

[0412] The top shortlisted transcripts were assessed with quantitative real-time PCR (qPCR) in an independent cohort of 70 fresh-frozen biopsies. Protein expression of the top biomarkers was assessed in formalin-fixed paraffin embedded (FFPE) biopsies and endoscopic mucosal resection (EMR) specimens obtained from patients undergoing routine surveillance for BE and endoscopic treatment for dysplasia who were enrolled onto the Biomarkers study at Addenbrooke’s Hospital in Cambridge, UK. Ethics approval was obtained (LREC 01 / 149), and the BEST2 study (10 / H0308171). P11112GBW0

[0413] The secretory protein assay was developed in a retrospective training cohort of capsule sponge samples obtained from BEST2 (10 / H0308171), and DELTA studies (20 / EE / 0141). An independent, prospective cohort (N= 130) was included in an additional cohort to provide a representative population. The assay was then independently validated in a test cohort (N=132). All prospective (training and test cohort) samples were collected as part of the BEST4 Surveillance (REC: 23 / WM / 0210) and BEST4 Screening studies as part of a translational exploratory endpoint (REC: 24 / WM / 0017). mRNA expression analysis in upper gastric and oesophageal biopsies

[0414] Haematoxylin and eosin (H&E) staining and pathology review was performed to estimate the lesion size and highest-grade present in each sample before mRNA extraction of the remaining frozen biopsy. RNA was extracted with QIAgen’s AHPrep kit, quantified with Qubit’s RNA HS kit and a total of 500ng of RNA was reversed-transcribed for each sample with the QuantiTect RT kit (QIAgen). The real-time quantitative PCR (qPCR) reaction for 10 transcripts (9 targets and 1 housekeeper) was performed with 2ul of a 1 :10 dilution of the resulting cDNA. Taqman assays (ThermoFisher Scientific, MA, USA) with the FAM-MGB dye were used for the qPCRs. Beta-actin (VIC-MGB, #Hs99999903_m1) was run in the same well as CXCL1. All reactions were run in triplicates in a QuantStudio cycler (ThermoFisher Scientific, MA, USA). Kruskal-Wallis tests, followed by Dunn’s multiple comparisons tests were used to determine statistically significant expression differences for each marker between clinical groups in GraphPad Prism (version 9.5.1).

[0415] Immunohistochemical analysis of protein targets

[0416] The EMR samples used for immunohistochemistry contained a range of relevant disease stages, including BE, dysplasia, and intramucosal cancer (IMG). All samples were sectioned at 3.5pm thickness and IHC staining was carried out in the BOND RX (Leica automated stainer) using optimised staining conditions. CD31 (#ab207090, Abeam, UK) was used to stain endothelial cells as a staining control on the EMR samples. Expert pathologist review was used to select areas of interest and for all staining interpretations. All images were obtained with the AxioScanZI slide scanner (Leica, Germany).

[0417] Immunoassay based evaluation of secretory markers for BE

[0418] All capsule sponge specimens were placed in a sample pot containing 30 ml SurePath™ cell fixative (BD, USA) according to standard processing. Upon receipt in the diagnostic laboratory (Cyted Health or Cambridge Tissue Bank) and between 1 hour and 1 day at RT, t a cell pellet was obtained from shaking cells off the sponge and centrifuging at 13,000 rpm for 10 mins to create a cell pellet. The supernatant was removed and aliquoted as the cell free fraction for subsequent analysis. This processing approach was P11112GBW0 not part of the standardised clinical protocol and applied in an exploratory manner, consistently yielding between 200-500ng total protein per sample. All supernatant samples were stored at -80°C within 30 days of sampling and thawed just before use. Enzyme-linked immunosorbent assays (ELISA) were used to quantify TFF3 (ab233617), regenerating family member 4 (REG4) (Biorbyt, orb180655, UK), Meprin A subunit alpha (MEP1a) (Abeam, ab26727), and serine peptidase inhibitor kazal type-4 (SPINK4) (Abbexa, abx528747) using a sandwich-based method from capsule sponge supernatant samples. Sample dilutions were determined for each assay and ranged from 1 :5 to 1 :50. A subset of protein markers (TFF3, CXCL8, and FABP2) were assessed using the Ella® microfluidic multiplex cartridges (ProteinSimple, Bio-Techne, Benelux, Belgium) following manufacturer’s instructions. All samples were run at 1 :10 dilution along with a high- and low-quality control.

[0419] Classification models for BE determination using capsule sponge supernatant markers

[0420] For analysis of the six-marker panel (TFF3, REG4, MEP1A, SPINK4, FABP2, CXCL8) to distinguish Barrett’s oesophagus from heartburn data lasso-logistic regression, random forest, XGBoost, and LightGBM — were implemented as scikit-learn pipelines with median imputation and z-score scaling, using class weighting to address minor class imbalances. Model building used 5-fold stratified cross-validation. Performance was assessed by ROC AUC, sensitivity, and specificity; probabilities were thresholded at 0.5 for sensitivity / specificity. After cross-validation, each model was refit on the full development cohort and evaluated on the prospective validation hold-out. We report AU ROC with bootstrap 95% confidence intervals and sensitivity / specificity with Wilson 95% intervals. Analyses were conducted in Python (Jupyter) with scikit-learn, XGBoost, and LightGBM.

[0421] Example 12 - Selection of Markers in Panel, Cell of Origin analysis, IHC analysis of expression in clinically relevant cell types

[0422] Identification of secretory biomarker candidates from transcriptomic profiling

[0423] Bulk RNA-seq differential expression analysis revealed 12 transcripts significantly upregulated in nondysplastic Barrett’s esophagus (NDBE) compared with adjacent tissues normal squamous esophagus (NE) and normal gastric mucosa (NG). These included established BE-associated transcripts such as CDX2. From the initial pool of 1 ,191 transcripts, candidates were shortlisted if they demonstrated (i) low expression in squamous and gastric tissues, (ii) maintained expression across NDBE through P11112GBW0 dysplasia, and (iii) marked expression differences between normal and pathological groups We refer to Figure 23A.

[0424] To validate transcriptomic findings, we performed qPCR analysis of six shortlisted transcripts in an independent cohort of 70 frozen biopsies. qPCR confirmed statistically significant overexpression of TFF3, REG4, and SPINK4 in NDBE, dysplastic BE (DBE), and intramucosal carci noma / early adenocarcinoma (IMC / EAC) relative to squamous esophagus and gastric mucosa. Expression of these targets was maintained across disease progression, with no significant differences between NDBE and DBE groups. In addition, TFF3, REG4, SPINK4, and CXCL8 transcripts were significantly upregulated in cancer groups compared with NE. We refer to Figure 23B.

[0425] Cell-of-origin exploration of BE biomarkers

[0426] To confirm the cell of origin of expression for the most differentially expressed 6 NDBE markers with biggest differences between NE / NG and NDBE, we interrogated a singlecell RNA-seq dataset to map their expression profiles. Normal duodenum (ND) was used as a positive control due to its high density of goblet cells. This lineage profiling was important to maximise the specificity of the secretory biomarkers to BE. FABP2 expression was detected in NG and normal duodenum (ND) enterocytes and goblet cells. REG4 and TFF3 were most transcribed in NDBE tissue, but not exclusively constrained to goblet cells, being present in enteroendocrine cells and enterocytes of NDBE and ND epithelium. Both TFF3 and REG4 were also highly expressed in the goblet cells of NDBE and ND, validating earlier studies. TFF3 is also part of an established diagnostic method used on the cellular fraction of the capsule sponge. In contrast, SPINK4 expression was exclusive to goblet cells in NDBE and ND. MEP1A was expressed across goblet cells and enterocytes in ND and NDBE epithelium at relatively lower levels than FABP2. CXCL8 was transcribed in macrophages in NE, NDBE, and ND tissues. We refer to Figure 24A.

[0427] Immunohistochemical validation of BE protein biomarkers in tissue samples

[0428] To confirm the BE cell lineage specific overexpression of TFF3, FABP2, REG4, SPINK4, CXCL8 and MEP1A at protein level, ten patients with known BE with matched NDBE, GC and NE biopsies were immunohistochemically stained. We refer to Figure 24A - IHC. TFF3, REG4 and SPINK4 were present in goblet cells, foveolar cells and enterocytes of BE tissue. In contrast, MEP1A stained mucus-secreting columnar cells lining BE glands (Figure 24B, NDBE biopsy). Importantly, NE and GC tissues were negative for all three markers, except for chief cells in GC specimens for TFF3 (Figure 24B). FABP2 was expressed in foveolar cells in the glandular BE epithelium, while CXCL8 was P11112GBW0 expressed by immune cells infiltrating the superficial mucosa. Subsequent IHC staining of capsule sponge sections (n = 10) confirmed BE-specific staining in goblet cell positive glands.

[0429] Example 13 - Construction of a Prediction Model with a Secretory Protein Biomarker Panel in Capsule Sponge Supernatant

[0430] To assess the diagnostic performance of the shortlisted proteins, we measured the expression levels of the six candidate biomarkers present within the capsule sponge supernatant using multiplex immunoassays. The average cell-free TFF3 value for BE samples (64.5 ng / ml) was 1.85 times greater than the average TFF3 value observed in heartburn controls (34.8 ng / ml), (p = 5.81x1 O'10) - we refer to Figure 25. Similarly, REG4 was significantly higher in BE compared to heartburn (p = 4.98x1 O'12), although 55% (93 / 169) of BE cases had undetectable REG4 concentration using the standard ELISA . Assessment of MEP1A and SPINK4 revealed a modest significant increase in MEP1A in BE compared to heartburn supernatant (p = 1.17 x10'2) but provided complementary information to REG4. Conversely, SPINK4 levels were significantly increased in BE compared to heartburn (p = 5.92 x10'7) but there was strong co-linearity with REG4. Of note, REG4 and SPINK4 concentration in BE supernatant significantly correlated with segment length (M) (r= 0.354, p = 3.96 x1 O'6and r= 0.234, p = 0.009, respectively) when comparing short segments (0-3cm) to medium length segments (4-6cm) (REG4; p = 0.02, SPINK4; p = 0.003), and long segments (7-10cm) (REG4; p = 0.0001 , SPINK4; p = 0.002), respectively.

[0431] As certain markers from our shortlisted candidates were available as target analytes on the Ella multiplex assay platform, FABP2 and CXCL8 were assessed as supernatant markers for BE in a multiplex assay simultaneously with TFF3. CXCL8 was moderately increased in BE (p =1.94 x10'4). FABP2 was found to be significantly higher in BE compared to heartburn controls (p = 1.45 x 10'13) (Figure 25). It is noteworthy that there is mutual exclusivity between REG4, MEP1A, SPINK4 / FABP2, while TFF3 is expressed to some degree in all samples where other biomarkers are also present, and SPINK4 and FABP2 showed concordant expression across BE cases - Figure 25B. There was no correlation between expression levels of TFF3, FABP2, MEP1A or CXCL8 with segment length (r = 0.09; p = 0.256, r = -0.137; p = 0.112, r = 0.114; p = 0.178, and r = 0.118; p = 0.172, respectively).

[0432] Machine-learning classifiers were trained on the development cohort (retrospective plus prospective) using a standardised pipeline (median imputation, z-score scaling, class weighting). Models included logistic regression, lasso-logistic regression, random forest, XGBoost, and LightGBM. Performance was estimated by 5-fold stratified cross- P11112GBW0 validation, with all preprocessing fit exclusively within each training fold to prevent information leakage. Discrimination and operating characteristics were assessed using the six-marker panel (TFF3, REG4, MEP1A, SPINK4, FABP2, CXCL8). LightGBM demonstrated the best overall performance. On held-out folds, LightGBM achieved AUG 0.83± 0.031 , with 69% sensitivity and 86% specificity at a pre-specified probability threshold of 0.5, demonstrating robust test-set performance within the development cohort (Figure 25D). Of note, the supernatant TFF3 assay alone could significantly differentiate between IHC-negative BE and true negative heartburn controls (Figure 25C).

[0433] Example 14: Validation of a Prediction Model With 6-Supernatant Biomarker Panel on an Independent Prospective Cohort

[0434] To independently test the diagnostic accuracy and performance of our sponge supernatant panel, we evaluated the model’s robustness in an independent prospective cohort. Among the 132 sponge supernatant samples included in the external test cohort, we detected substantial validation of our training dataset, where CXCL8, FABP2, REG4, SPINK4 and TFF3 maintained the ability to distinguish BE patients from reflux controls, with an ALIROC of 0.87 (95% Cl, 81-93) and corresponding sensitivity and specificity values of 71% (95% Cl, 59-81) and 97% (95% Cl, 90-99), respectively (Figure 26A-B). The high specificity means that only 2 / 60 heartburn controls were falsely assigned to a BE label, resulting in a positive predictive value of 95.9% (Figure 26C).

[0435] Although the concentration of REG4 was low, we observed that of the six proteins that informed our diagnostic algorithm, REG4 had the highest mean SHAP value, implying the most impact on the model output (Figure 26D).

[0436] Finally, we conducted a stepwise feature drop test to assess the impact of reducing biomarkers on model performance. In the external validation cohort, the full six-marker panel achieved AUG 0.87; performance declined when individual markers were removed.

[0437] Summary

[0438] In this proof-of-concept study in 132 patients in the independent test set, our novel capsule sponge supernatant secretory protein panel has a positive predictive value of 95.9%. This shows the advantage of a reduced false-positive rate as well as an easier workflow. Additionally, the cost and time saved by using an automated, liquid based P11112GBW0 multiplexed assay of the invention renders the capsule sponge more suitable to larger scale e.g. population screening.

[0439] The invention provides many advantages. Biomarker discovery and shortlisting was conducted in a well-defined, large transcriptomic cohort with complete endoscopy followup data. While our secretory protein assay development comprised of a multi-centre retrospective cohort enriched with longer segment BE cases, we combined this cohort with an independent prospective, real-world, unenriched cohort across a range of BE segment lengths to train the ML model. The performance of our model remained consistent in an external prospective validation cohort. Moreover, the BE cases in the prospective cohorts are derived from an ongoing platform trial of screening and surveillance, with longitudinal follow-up data.

[0440] Example 15 - Eosinophilic Oesophagitis

[0441] We refer to Figure 27. Heartburn n = 42, EoE n = 46

[0442] We show data on the EOE biomarkers. The final panel is as in the claims.

[0443] The final panel was reduced to include strongest biomarkers for EoE. For example, G- CSF, CD163 and MCP1 were excluded.

[0444] Example 16 - Comparative Data

[0445] We refer to Figure 28A. We show plots comparing performance of 7 vs 6 marker panels for the BE test showing superior performance for 6 marker panel for BE. We refer to Figure 28B. We show plots demonstrating that 6 marker panel is better than 5 marker panel in BE: plot shows reduced performance with the 5-marker panel. We refer to Figure 28C. We show plots demonstrating beneficial performance of the 6-marker panel.

[0446] This demonstrates that the selection of the 6 marker panel as in the claims is unexpectedly superior to either a smaller panel and - even more surprisingly - also superior to a larger panel. These data illustrate that the 6 marker panel for BE as in the claims has special qualities beyond the expectations of the inventors.

[0447] Although illustrative embodiments of the invention have been disclosed in detail herein, with reference to the accompanying drawings, it is understood that the invention is not limited to the precise embodiment and that various changes and modifications can be effected therein by one skilled in the art without departing from the scope of the invention as defined by the appended claims and their equivalents.

Claims

P11112GBW0CLAIMS1. A method of collecting information useful in aiding diagnosis of an oesophageal disease in a subject comprising: providing a sample from said subject, wherein said sample comprises cells collected from the surface of the subject's oesophagus; contacting said sample with an aqueous fluid; taking an aliquot of said aqueous fluid; assaying said aliquot of aqueous fluid to determine the presence of one or more soluble biomarker(s), wherein said one or more soluble biomarker(s) comprise the soluble protein biomarker(s):(i) TFF3(ii) REG4(iii) FABP2(iv) MEP1A(v) CXCL8; and(vi) SPINK4 and wherein presence of each of said soluble protein biomarkers (i) to (vi) in said aliquot is indicative of said subject having Barrett’s Oesophagus.

2. A method according to claim 1 wherein said one or more soluble biomarker(s) comprise the soluble protein biomarker(s):(i) TFF3(ii) REG4(iii) FABP2(iv) MEP1A(v) CXCL8(vi) SPINK4; and(vii) Gal-3 and wherein presence of each of said soluble protein biomarkers (i) to (vii) in said aliquot is indicative of said subject having Barrett’s Oesophagus.

3. A method according to claim 1 wherein said method has an AU ROC value of at least 0.93.P11112GBW04. A method according to claim 2 wherein said method has an AU ROC value of at least 0.95.

5. An apparatus or system which is(a) specifically adapted to analyse an oesophagal sample from a subject, wherein said analysis comprises:(b) taking an aliquot of aqueous fluid previously contacted with oesophageal cells obtained from said subject; assaying said aliquot of aqueous fluid to determine the presence of one or more soluble biomarker(s) according to any one of claims 1 to 2; said apparatus or system comprising an output module, wherein if presence of said markers are detected, then said output module indicates an increased likelihood of an oesophageal disease for said subject, wherein said oesophageal disease is Barrett’s Oesophagus.

6. Use for applications relating to aiding detection of an oesophageal disease in the oesophagus of a subject, wherein said oesophageal disease is Barrett’s Oesophagus, of a material which recognises, binds to or has affinity for certain soluble biomarkers, wherein the soluble biomarkers are as defined in any one of claims 1 to 2.

7. Use according to claim 6 of a combination of materials, each of which respectively recognises, binds to or has affinity for one or more of said soluble biomarkers.

8. An assay device for use in aiding detection of an oesophageal disease in the oesophagus of a subject, wherein said oesophageal disease is Barrett’s Oesophagus, which comprises a solid substrate having a location containing a material, which recognises, binds to or has affinity for certain soluble biomarkers, wherein the soluble biomarkers are as defined in any one of claim 1 or claim 2.

9. A method of collecting information useful in aiding diagnosis of an oesophageal disease in a subject comprising: providing a sample from said subject, wherein said sample comprises cells collected from the surface of the subject's oesophagus; contacting said sample with an aqueous fluid; taking an aliquot of said aqueous fluid; assaying said aliquot of aqueous fluid to determine the presence of one or more soluble biomarker(s),P11112GBW0 wherein said one or more soluble biomarker(s) comprise the soluble protein biomarker(s):1) C-kit2) Eotaxin-33) IL-54) IL-135) IL-336) IL-25; and7) Eotaxin-2 and wherein presence of each of said soluble protein biomarkers (1) to (7) is indicative of said subject having Eosinophilic Oesophagitis.

10. A method according to claim 9 wherein said one or more soluble biomarker(s) comprise the soluble protein biomarker(s):1) C-kit2) Eotaxin-33) IL-54) IL-135) IL-336) IL-257) Eotaxin-28) Mast cell tryptase9) CPA310) ECP; and11) Periostin and wherein presence of each of said soluble protein biomarkers (1) to (11) is indicative of said subject having Eosinophilic Oesophagitis.

11. A method of collecting information useful in aiding diagnosis of an oesophageal disease in a subject comprising: providing a sample from said subject, wherein said sample comprises cells collected from the surface of the subject's oesophagus; contacting said sample with an aqueous fluid; taking an aliquot of said aqueous fluid; assaying said aliquot of aqueous fluid to determine the presence of one or more soluble biomarker(s),P11112GBW0 wherein said one or more soluble biomarker(s) comprise:(b) a specific binding partner of Cytisus scoparius (CSA) lectin, and wherein presence of said specific binding partner of Cytisus scoparius (CSA) lectin is indicative of said sample being adequate to support collection of information useful in aiding diagnosis of an oesophageal disease.

12. A method according to claim 11 wherein said one or more soluble biomarker(s) comprise:(a) a specific binding partner of Cytisus scoparius (CSA) lectin; and(b) a specific binding partner of Peanut Agglutinin (PNA) lectin, and wherein presence of each of said specific binding partners (a) and (b) is indicative of said sample being adequate to support collection of information useful in aiding diagnosis of an oesophageal disease.

13. A method according to claim 11 wherein said method has an AU ROC value of at least 0.95.

14. A method according to claim 12 wherein said method has an AU ROC value of at least 0.96.

15. A method according to any preceding claim wherein presence of said one or more soluble biomarker(s) is determined as presence at an elevated level in said aliquot by comparison to a reference value.

16. A method according to any preceding claim wherein said aqueous fluid comprises SurePath™ Preservative Fluid.

17. An apparatus or system which is(a) specifically adapted to analyse an oesophagal sample from a subject, wherein said analysis comprises:(b) taking an aliquot of aqueous fluid previously contacted with oesophageal cells obtained from said subject; assaying said aliquot of aqueous fluid to determine the presence of one or more soluble biomarker(s) according to any one of claim 9 or claim 10; said apparatus or system comprising an output module,P11112GBW0 wherein if presence of said markers are detected, then said output module indicates an increased likelihood of an oesophageal disease for said subject, wherein said oesophageal disease is Eosiniphilic Oesophagitis.

18. Use for applications relating to aiding detection of an oesophageal disease in the oesophagus of a subject, wherein said oesophageal disease is Eosiniphilic Oesophagitis, of a material which recognises, binds to or has affinity for certain soluble biomarkers, wherein the soluble biomarkers are as defined in any one of claim 9 or claim 10.

19. Use according to claim 18 of a combination of materials, each of which respectively recognises, binds to or has affinity for one or more of said soluble biomarkers.

20. An assay device for use in aiding detection of an oesophageal disease in the oesophagus of a subject, wherein said oesophageal disease is Eosiniphilic Oesophagitis, which comprises a solid substrate having a location containing a material, which recognises, binds to or has affinity for certain soluble biomarkers, wherein the soluble biomarkers are as defined in any one of claim 9 or claim 10.

Images