WO2024096737A1 - Methods for determining the age of platelets - Google Patents

Abstract

The present invention provides an in vitro method for determining the physiological age of blood platelets in a test sample wherein said physiological age is expressed in days and/or hours. The method comprises providing a test sample of blood platelets, determining the level of expression of one or more of tubulin, VWF, SPARC, CD63 and PF4 in one or more of individual platelets within said test sample, and determining, based on the level of expression determined, the physiological age of the blood platelets in said test sample. Preferably, the method comprises the using a machine learning data processing model.

Description

Title: Methods for determining the age of platelets

FIELD OF THE INVENTION

The present invention relates to diagnostic methods for the in vitro analysis of blood, in particular blood platelets, preferably using immunological staining techniques. The present disclosure provides an in vitro method for determining the physiological age of blood platelets in a sample.

BACKGROUND OF THE INVENTION

Platelets are small, anucleated cells with their primary physiological role to repair vascular damage (hemostasis) and initiate thrombus formation in response to vascular injury. Subjects with a low platelet count have an increased risk of bleeding. Platelet transfusion is a routinely used lifesaving procedure to control or prevent bleeding in patients with low platelet count or other platelet dysfunction.

For platelet transfusion, donor blood from healthy donors is used. This donor blood is often stored before it is used in patients. However, platelet quality and function deteriorates significantly during storage due to so-called storage lesion. Much research has been performed to measure platelet characteristics over time during storage as the platelets age. It has been shown that during ageing, platelets lose surface receptors and intracellular proteins. Also, the shape of the platelet and the regulation of platelet granular content appears essential in maintaining a healthy platelet function. Yet, none of these characteristics can be used in routine clinical practice as the correlation with exact age is lacking. Using prior art methods, only differences between a very young and very old platelet can be seen, but establishing the physiological age of platelets in a sample remains difficult. Currently no in vitro tests are available to determine physiological platelet age.

Platelet concentrates (PCs) are generally stored for 5-7 days at room temperature due to risk of infection. New techniques are introduced to prolong platelet storage time (such as pathogen inactivation techniques), but the effect of such techniques and the effect of prolonging the storage time on platelet function and/or age are currently unknown as no tests are available to determine apparent platelet age. Platelets may age quicker with these techniques and therefore be of lesser quality. Prevention or treatment of bleeding may be diminished or be less effective when using such PCs. Therefore, there is a need for tests to accurately determine or establish the physiological age of platelets in samples such as PCs used in transfusion medicine. Such tests may also be used to determine the effect of storage (conditions) on PCs used for platelet transfusion.

In patients with thrombocytopenia, such as immune thrombocytopenia (ITP), the low platelet count may be caused by a decreased platelet production or by an increased platelet clearance. Treatment of ITP patients is based on either stimulation of platelet production (e.g. by administering thrombopoietin receptor agonist (TPO- RA)) or by decreasing platelet clearance (e.g. by administering glucocorticoids and/or Rituximab). Presently, no evidence is available which treatment should be preferred. Patients with an increased platelet clearance will have relatively young platelets and may for example benefit from Rituximab, whereas patients with decreased platelet production will have relatively old platelets and would benefit more from TPO-RA. Providing such patients with the incorrect treatment may lead to major bleeding complications. Therefore, knowledge of platelet age in these patients will lead to better treatment options. In addition, knowledge of platelet age in these patients would provide a new method for diagnosing the underlying cause of thrombocytopenia. The present disclosure demonstrates that using confocal and super-resolution microscopy loss of platelet granule content during aging can be quantified. In this study Artificial intelligence techniques (AIT) are used to derive physiological platelet age from imaging data.

The availability of microscopy techniques, in particular superresolution microscopy techniques, allows imaging of platelets in more detail. Using specific age-related expression markers in combination with superresolution microscopy the present inventors were able to observe subtle differences in platelets during storage. For example the expression of tubulin and von Willebrand Factor (VWF) changes over time. Analysis of expression of a combination of markers using platelet images now allows accurate determination of the age of platelets. The use of artificial intelligence greatly improves the accuracy of the method of determining the age of platelets from platelet images using expression markers.

The aim of the present invention is to provide a method for determining the physiological platelet age, in particular from images of platelets acquired by confocal or super-resolution microscopy. For example, stored in plasma-rich-plasma (PRP) and/or in platelet concentrates (PCs). Preferably, the method comprises the use of Artificial intelligence techniques-based algorithms.

The invention provides methods for determining the quality of platelet concentrates used for transfusion based on platelet age, and methods for determining the platelet function in vivo. The invention provides methods for determining efficacy of thrombocytopenia treatments involving compounds that increase platelet production or decrease platelet clearance by using in vitro analysis methods on platelets of thrombocytopenia patients prior to, during, or after such treatment. The invention also provides methods for diagnosing thrombocytopenia as being caused by a decreased platelet production or by an increased platelet clearance. The invention also provides improved methods of treating patients suffering from thrombocytopenia involving monitoring of platelet age and/or function using in vitro analysis methods of the invention, optionally in combination with therapeutic or surgical intervention as described herein.

SUMMARY OF THE INVENTION

The present invention now provides an in vitro method for determining the physiological age of blood platelets in a test sample wherein said physiological age is expressed in days and/or hours, the method comprising: a) providing a test sample of blood platelets; b) determining the level of expression of one or more of tubulin, VWF, SPARC, CD63 and PF4 in one or more of individual platelets within said test sample, c) determining, based on the level of expression determined under b), the physiological age of the blood platelets in said test sample.

In preferred embodiments of a method of the invention step c) comprises comparing the level of expression determined under b) with the level of expression in a reference sample of blood platelets of known physiological age, and determining the physiological age of the blood platelets in said test sample based on said comparison. Preferably, the level of expression is compared using a machine learning data processing model.

In preferred embodiments of methods of the invention as described herein, the step of determining the level of expression of one or more of tubulin, VWF, SPARC, CD63 and PF4 comprises the use of immunological staining techniques. Immunological staining, or immunostaining, techniques include the use of an antibody-based method to detect the tubulin, VWF, SPARC, CD63 or PF4 protein in a platelet. In methods of this invention, determining the level of expression by immunostaining preferably includes the use of fluorescent antibodies in combination with microscopy or flow cytometry to quantify the level of expression of said protein(s). Preferably, microscopy is used, more preferably in combination with image analysis of individually stained platelets. Preferably, microscopy is confocal or, more preferably, super-resolution microscopy. In most preferred embodiments of aspects of the invention the level of expression is determined by immunostaining of at least tubulin, VWF and SPARC proteins in said platelets or by immunostaining of at least tubulin, CD63 and PF4 proteins in said platelets, preferably in combination with image analysis of fluorescence intensities measured from said immunostaining, preferably of individually stained platelets. Preferably, the image analysis further comprises collection of data on the morphology of the platelet(s), preferably the (discoid) shape of said platelet(s) and the presence of granules in said platelet(s). Preferably, staining for tubulin is used as a marker for the shape of the platelet(s).

In preferred embodiments of methods of the invention as described herein, the step of determining the level of expression of one or more of tubulin, VWF, SPARC, CD63 and PF4 comprises determining the level of expression of two or more of tubulin, VWF, SPARC, CD63 and PF4, including tubulin and one of VWF, SPARC, CD63 and PF4, more preferably, three or more of tubulin, VWF, SPARC, CD63 and PF4. Three or more of tubulin, VWF, SPARC, CD63 and PF4 may include any combination of these proteins, including tubulin, VWF and SPARC; tubulin, VWF and CD63; tubulin, VWF and PF4; tubulin, SPARC and CD63; tubulin, SPARC and PF4; tubulin, CD63 and PF4; VWF, SPARC and CD63; VWF, SPARC and PF4; VWF, CD63 and PF4; SPARC, CD63 and PF4. Highly preferred embodiments of the step of determining the level of expression of one or more of tubulin, VWF, SPARC, CD63 and PF4 comprise determining the level of expression of tubulin, VWF and SPARC. Alternatively, highly preferred embodiments of the step of determining the level of expression of one or more of tubulin, VWF, SPARC, CD63 and PF4 comprises determining the level of expression of tubulin, CD63 and PF4. Tubulin is preferably alpha-tubulin.

In preferred embodiments of methods of the invention as described herein, the step of determining the level of expression (of one or more of tubulin, VWF, SPARC, CD63 and PF4) in one or more of individual platelets within said platelet sample, includes determining the level of expression in a multitude of individual platelets, preferably at least 10, 20, 30, 40, 50, 60, 70, 80. 90, 100, 200, 300, 400, 500, 600, 700, 800, 900, 1000, 2000, 3000, 4000, 5000, 6000, 7000, 8000, 9000, 10000 or more platelets within said platelet sample.

A test or reference sample of blood platelets may be a blood sample, a platelet-rich plasma (PRP) sample or platelet concentrate (PC) sample.

In preferred embodiments of methods of the present invention, the step of determining the physiological age of the platelets comprises comparing the level of expression (of one or more of tubulin, VWF, SPARC, CD63 and PF4 in one or more of individual platelets within said platelet sample) with the level of expression in reference platelets of known physiological age, and determining the physiological age of the platelets within said platelet sample based on said comparison.

In preferred embodiments of methods of the present invention, the step of determining the physiological age of the platelets comprises the use of a trained machine learning data processing model.

In preferred embodiments of methods of the present invention, the step of determining the physiological age of the platelets by comparing the level of expression (of one or more of tubulin, VWF, SPARC, CD63 and PF4 in one or more of individual platelets within said platelet sample) with the level of expression in reference platelets of known physiological age, comprises the use of a trained machine learning data processing model. A trained machine learning data processing model is preferably a machine learning data processing model for automatically associating the level of expression (of one or more of tubulin, VWF, SPARC, CD63 and PF4 in one or more of individual platelets within said platelet sample) with a physiological age of said one or more platelets, such as to enable providing the age of the one or more platelets as or at an output of the machine learning data processing model.

Preferred embodiments described above may be used in combination and are also preferred embodiments in other aspects of this invention.

In a further aspect, the present invention provides a method of diagnosing or typing an individual suffering from thrombocytopenia, the method comprising: determining the physiological age of platelets in a blood sample from said patient using the in vitro method for determining the physiological age of platelets in a blood sample as described above, and typing the patient based on the physiological age of the platelets determined as:

(a) a patient suffering from thrombocytopenia with increased platelet clearance, or

(b) as a patient suffering from thrombocytopenia with decreased platelet production.

In a preferred embodiment is provided the method of typing an individual suffering from thrombocytopenia as disclosed herein, further comprising determining a treatment regime.

In a further embodiment is provided a method of treating a patient suffering from thrombocytopenia, comprising the step of typing the patient by a method as disclosed herein, and treating the patient with therapeutic compounds aimed at decreasing platelet clearance or increasing platelet production, depending on the result of the typing method. In a further embodiment is provided a treatment regime aimed at decreasing platelet clearance for the treatment of an individual typed as (a) a patient suffering from thrombocytopenia with increased platelet clearance according to the method as disclosed herein.

In a further embodiment is provided a treatment regime aimed at increasing platelet clearance for the treatment of an individual typed as (b) a patient suffering from thrombocytopenia with decreased platelet production according to the method as disclosed herein.

In a further embodiment is provided a method of training a machine learning data processing model for determining the age of platelets in a blood sample, the method comprising: a) receiving, by the machine learning data processing model, a training data set comprising training data indicative of a first level of expression of one or more of tubulin, VWF, SPARC, CD63 and PF4 in a representative number of first individual platelets within a platelet sample, wherein preferably the first level of expression is determined from the platelet sample using immunological staining techniques in combination with super-resolution microscopy and image analysis; b) receiving, by the machine learning data processing model, a ground truth data set comprising ground truth data indicative of the physiological age of the first platelets, wherein the physiological age of the first platelets is determined using the method as disclosed herein; c) training the machine learning data processing model based on the training data received in step a) and the ground truth data received in step b) for enabling, after completion of the training period, the step of automatically associating measurement data indicative of a second level of expression of one or more of tubulin, VWF, SPARC, CD63 and PF4 in at least one second individual platelet with an physiological age of the at least one second platelet, such as to enable providing the age of the at least one second platelet at an output of the machine learning data processing model. In a further embodiment is provided a system for determining the age of platelets in a blood sample, the system comprising: one or more processors for receiving measurement data indicative of a level of expression of one or more of tubulin, VWF, SPARC, CD63 and PF4 in at least one individual platelet within a platelet sample obtained from a blood sample; and a memory storing a trained machine learning model, wherein the trained machine learning model is configured to output a classification of the at least one individual platelet based on the measurement data; and wherein the one or more processors are configured to determine, data indicative of the age of the at least one individual platelet based on the output classification.

In a further embodiment is provided an in vitro method for determining the physiological age of platelets in a blood sample, the method comprising: a) providing a sample of platelets from the blood sample; b) determining the level of expression of one or more of tubulin, VWF, SPARC, CD63 and PF4 in one or more of individual platelets within said platelet sample, c) comparing the level of expression determined under b) with the level of expression in reference platelets of known physiological age.

Preferably, wherein the level of expression is determiner from two or more, or three or more of tubulin, VWF, SPARC, CD63 and PF4.

The present invention further provides a method of determining a treatment regime, in particular in an individual suffering from thrombocytopenia, the method comprising determining the physiological age of the platelets in a sample derived from a subject using a method of the present invention, and typing the patient based on the physiological age of the platelets as a patient suffering from thrombocytopenia with increased platelet clearance (characterized by platelets having an age in the range of 1-24 hours, or 1-12 hours), or as a patient suffering from thrombocytopenia with decreased platelet production (characterized by platelets having an age in the range of 1-10 days, preferably 5-7 days).

BRIEF DESCRIPTION OF THE DRAWINGS

Figure 1 provides a schematic overview of a training workflow in accordance with the invention. After storage in platelet-rich-plasma (PRP) platelets were stained with VWF, SPARC or alpha-tubulin and a convolutional network was trained with the images as input. The resulting model after training for 100 training cycles could predict the correct platelet age with over 95% accuracy (blue line: set that was trained on; orange line: separate validation set).

Figure 2 provides the results of training of the prediction model for platelet age of platelet concentrates stored up to 10 days. First panel shows training on all time points, resulting in overfitting seen by the discrepancy between training (blue) and validation (orange) set. Not including day 9 and 10, as shown in the middle panel, shows a high accuracy in categorization (predicting platelet age). The right panel shows that the model is still able to distinguish young (1,3,6 days stored) from old (10 days stored) platelets (blue line set that was trained on; orange line separates validation set).

DETAILED DESCRIPTION OF THE DISCLOSED EMBODIMENTS Definitions

As used herein, "to comprise" and its conjugations is used in its non-limiting sense to mean that items following the word are included, but items not specifically mentioned are not excluded. In addition the verb “to consist” may be replaced by “to consist essentially of’ meaning that a compound or adjunct compound as defined herein may comprise additional component(s) than the ones specifically identified, said additional component(s) not altering the unique characteristic of the invention. The articles “a” and “an” are used herein to refer to one or to more than one (i.e., to at least one) of the grammatical object of the article. By way of example, “an element” means one element or more than one element.

As used herein, the terms "treatment," "treat," and "treating" refer to reversing, alleviating, delaying the onset of, or inhibiting the progress of a disease or disorder, or one or more symptoms thereof, as described herein. In some embodiments, treatment may be administered after one or more symptoms have developed. In other embodiments, treatment may be administered in the absence of symptoms. For example, treatment may be administered to a susceptible individual prior to the onset of symptoms (e.g., in light of a history of symptoms and/or in light of genetic or other susceptibility factors). Treatment may also be continued after symptoms have resolved, for example to prevent or delay their recurrence.

The term “immunostaining”, as used herein, refers to the use of polyclonal or monoclonal antibodies (or fragments thereof) as specific markers for detecting the (quantitative) presence of the tubulin, VWF, SPARC, CD63 or PF4 in or on platelets as described herein and as well known in the art. Platelets may be fixated, for instance using 1% formaldehyde, are then mounted on a microscope slide, permeabilized, and stained. A primary antibody targeting one of the indicated proteins may be used in combination with a secondary antibody, using methods well known in the art. The label that is used to detect the binding of the antibody/antibodies is preferably a fluorescent label, and detection of fluorescence is preferably quantitative.

The present disclosure, inter alia, provides an in vitro method for determining the physiological age of platelets in a blood sample wherein the method comprises providing a sample of platelets from a blood sample, determining the level of expression of one or more of tubulin, VWF, SPARC, CD63 and PF4 in the platelets in said sample and comparing said level of expression to the level of expression in reference platelets. Preferably, the level of expression is compared using a machine learning data processing model. In some embodiments the method can be used for typing a patient suffering from thrombocytopenia, optionally followed by choosing the appropriate treatment.

The present disclosure further provides a method of training a machine learning data processing model for determining the age of platelets in a blood sample. Furthermore, the disclosure provides a system for determining the age of platelets in a blood sample.

Platelets, also called thrombocytes, are a component of blood whose function (along with the coagulation factors) is to react to bleeding from blood vessel injury by clumping, thereby initiating a blood clot. Platelets have no cell nucleus. They are fragments of cytoplasm that are derived from the megakaryocytes of the bone marrow or lung, which then enter the circulation. Mature megakaryocytes produce platelets via intermediate cytoplasmic extensions known as proplatelets. Subsequently, the platelets are released into the bloodstream. Platelets have a short lifespan. The lifespan of platelets is 7 to 10 days in circulation.

Platelets play a role in many pathophysiological processes, including homeostasis, inflammation, infection, vascular integrity and metastasis. However, their role in haemostasis and thrombosis remains one of the most important. Upon damage to the vascular endothelium, subendothelial matrix proteins are exposed, which provides docking sites for platelet receptors. Binding of the platelet receptors initiates a complex process to promote platelet activation, and ultimately forming a platelet plug to close the damaged endothelium.

Platelets have a typical discoid shape. The cytoskeleton of the platelet contributes to maintaining the typical discoid shape of the platelets. Furthermore, platelets comprise different types of granules and the exocytosis of these granules is central to the function of the platelet. The three major granule types are dense granules, ct-granules and lysosomes. Platelet granules are formed in megakaryocytes, prior to transport into platelets. Granules remain stored in circulating platelets until platelet activation triggers the exocytosis of their contents. ct- Granules are unique to platelets and are the most abundant of the platelet granules, numbering 50—80 per platelet and account for about 10% of platelet volume. They contain mainly proteins, both membrane- associated receptors (for example, ctIIbB3 and P-selectin) and soluble cargo (for example, platelet factor 4 [PF4] and fibrinogen). Dense granules (also known as 8-granules) are the second most abundant platelet granules, with 3—8 per platelet. Dense granules mainly contain bioactive amines (for example, serotonin and histamine), adenine nucleotides, polyphosphates, and pyrophosphates as well as high concentrations of cations, particularly calcium.

A normal platelet count ranges from 150,000 to 450,000 platelets per microliter of blood. Having more than 450,000 platelets is a condition called thrombocytosis; having less than 150,000 is known as thrombocytopenia. The platelet number can be tested for example with a blood test called a complete blood count (CBC).

Thrombocytosis is a medical condition having too many platelets. There are two types of thrombocytosis: primary (or essential) thrombocytosis and secondary thrombocytosis. Primary thrombocytosis is caused by abnormal cells in the bone marrow resulting in an increase in platelets. Secondary thrombocytosis can be caused by a disease such as anemia, cancer, inflammation or infection. Thrombocytosis can cause symptoms such as blood clots in for example the arms and legs. These blood cloths can also lead to heart attacks and stroke.

The medical term for having too few platelets is thrombocytopenia. Thrombocytopenia can have different causes for example, medication, such as chemotherapy, an inherited condition, certain types of cancer, such as leukemia or lymphoma, kidney infection or dysfunction or alcohol abuse. These causes lead to either decreased platelet production and/or increased platelet clearance. Both result in lower platelet numbers.

Symptoms of thrombocytopenia are for example easy bruising and frequent bleedings. Increased bleeding risk can be life threatening and platelet transfusion is a routinely used lifesaving procedure to control or prevent bleeding in patients with thrombocytopenia or platelet dysfunction. For platelet transfusion donor blood of healthy donors is used. After donation the donor blood is stored for some time before it is used for a patient. It is known that during storage the condition of the blood platelets deteriorates. Therefore it is of great value to be able to test physiological age of platelets in the donor blood.

Currently it is difficult to predict or determine the physiological age of the platelets in a sample as no in vitro tests are currently available. Knowing the age of platelets is important in the field of transfusion medicine as well as for the treatment of patients with thrombocytopenia.

The present disclosure provides an in vitro method for determining the physiological age of platelets in a blood sample, the method comprising: a) providing a sample of platelets from the blood sample; b) determining the level of expression of one or more of tubulin, VWF, SPARC, CD63 and PF4 in one or more of individual platelets within said platelet sample, c) comparing the level of expression determined under b) with the level of expression in reference platelets of known physiological age, d) determining from said comparison the physiological age of the platelets in said blood sample.

The method of the disclosure is an “in vitro” method. The term “in vitro” refers to a process performed or taking place in a test tube, culture dish or elsewhere outside a living organisms. A process performed in a living organisms is called “in vivo”. In some embodiments the method is a diagnostic method, preferably an in vitro diagnostic method. A diagnostic method is a method concerned with the diagnosis of illness or other problems.

Age of the platelet

In aspects of this invention, the age of the platelets may be determined as of the moment since they were obtained from a donor. Methods of this invention can also be used to determine the age of platelets since their release in the bloodstream.

The physiological age of a platelet may thus be expressed as the age of the platelet since the moment it was obtained from the donor. Alternatively, the age of a platelet may be expressed as the age of a platelet since the origin of the platelet from the megakaryocytes followed by its release into the bloodstream.

The methods of the invention are performed to determine the physiological age of the blood platelets. The physiological age of the blood platelets represents the condition of the platelets. It relates to the function of the platelets, which is mainly determined by the ability of the platelet to aggregate. The ability to aggregate is dependent on the morphology of the platelet. Important factors for the morphology of the platelet are the typical discoid shape as well as the presence of the granules in the platelet, in particular the ct-granules.

In embodiments of this invention, the physiological age of a blood platelet is expressed in days and/or hours. Preferably the physiological age of the platelets is in hours. In preferred embodiments, a method of the invention is performed to determine the exact physiological age of an individual platelet, wherein the term exact means the age in hours.

The condition of the platelets can be affected by many factors such as storage and stress conditions. These factors can positively or negatively affect the aging of the platelets. Therefore the physiological age of the platelet can differ from the chronological age of the platelets.

Stored platelets have chronological age, namely the time from taking the sample from the subject, or the time from origin and release in the bloodstream. One of skill in the art will understand that proper reference samples can be used to calibrate the expression levels of markers measured using the methods of this invention to either the age from taking the sample from the subject, or to the time from their release in the blood stream.

Shelf life of platelet concentrates is limited to 5-7 days due to loss of platelet function during storage. Prolonged platelet storage leads to a decrease in platelet functionality known as the platelet storage lesion. The functionality of platelets is reduced upon storage as has been shown by decreased response to agonists like ADP, collagen, ristocetin and PARI activation peptides. During storage the morphology of the platelets changes and the typical discoid shape is lost. The platelets also become activated, which results in the release of ct-granules. In addition, levels of membrane proteins such as GPIbct and GPV decline during storage.

Improvements in storage conditions may increase the platelet quality. For example, adding additives may prolong the platelet function for stored samples containing platelets. The aging of these platelets will be slowed down and the function of the platelets is preserved for a longer time. Thus, the physiological age of these platelets is lower than the chronological age.

Young platelets are in general larger and are more able to activate. Older platelets have released their alfa granules and are smaller. Younger platelets also might be more able to activate and play an important role in thrombus formation. The role in thrombosis is unclear but platelet age might also play a role in this. Storage of blood plasma, Storage conditions and limited storage time Platelet samples for use in this invention may comprise (whole) blood samples, platelet-rich plasma (PRP) or platelet concentrate. Platelet concentrates (PCs) can be made either by a single donor or a pool of different donors. PRP may be prepared as described in the Examples below. PCs may for instance be obtained from Sanquin (Amsterdam, The Netherlands). Methods of the present invention enable the prediction or determination of platelet age of platelets stored as platelet concentrates at room temperature. Knowing platelet age in a platelet concentrate is important to improve the quality of the PCs. New techniques are investigated to improve storage time, for example pathogen inactivation techniques.

Platelets also have an age in vivo in the blood circulation. As indicated, the average life span of platelets in the circulation is between 7 and 10 days. The mixture of platelets in the bloodstream (and hence in a sample taken therefrom) is heterogenous, comprising platelets of different age. Some platelets have just been released from megakaryocytes while other platelets are reaching the end of their lifespan.

In methods of this invention, both a physiological age or a chronological age may be determined. The term “age of a platelet”, as used herein, may refer both to its physiological age, as well as to its chronological age. The age of platelets in an in vitro sample or in vivo may follow a Gaussian distribution. In methods of this invention, the expression “age of a platelet” refers to the age of the platelet relative to known reference values, and based on the expression level of proteins as described herein. Such reference values may for instance be provided in the form of the limits specified by two standard deviations of the age distribution of all (or a representative number of) platelets in a sample, with the age of individual platelets in the sample being in the range of values lying between these limits. Alternatively, the reference values may for instance be provided in the form of standard values (of marker expression vs. platelet age) for an individual or for a group of individuals that are healthy, or that suffer from a platelet disorder as described herein.

In methods of this invention, the age of a platelet is determined for each platelet individually based on quantitative image analysis of fluorescence intensities measured from one or more markers as described herein. The measure of the age of platelets in the blood sample determined may be expressed as the mean for all platelets measured.

In some embodiments the method is used to determine the physiological age of individual platelets in a sample. In some embodiments, the method is used to determine the average physiological age of platelets in a sample. In some embodiment, the method is used to determine the distribution of the physiological age of platelets in a blood sample.

The method of the disclosure provides a sample of platelets from a blood sample. Blood contains many different components including blood cells. The blood cells are mainly red blood cells, white blood cells and platelets. A normal platelet count ranges from 150,000 to 450,000 platelets per microliter of blood, however, the platelets accounts for less than 1 % of the blood volume.

For a blood platelet test as described herein it is favourable to have a platelet sample, having a higher percentage of platelets. The skilled person known various techniques to enrich a blood sample for platelets or to isolate platelets from a blood sample. For example by centrifuging a blood sample without or with a low brake. The centrifugal forces cause a distribution of the various compounds. A fraction of the sample will be enriched for platelets. This fraction is called platelet-rich-plasma. This exemplary method is described in detail in the example section.

In some embodiments, the level of expression of the markers is determined a representative number of platelets. Analysing a representative number of platelets gives a good representation of the population of platelets in the sample. The term “representative number of platelets”, as used herein, may refer to at least a 100 platelets, more preferably at least 200 platelets, more preferably at least 300, 400, 500, 600, 700, 800, 900, or at least a 1000 platelets.

Using image analysis, it is preferred that at least 5-20, preferably around 10 microscopic images are analysed and that each microscopic image holds about 100 platelets, preferably resulting in the analysis of around 1.000 or more platelets.

The disclosure provides determining the expression levels of one or more markers selected from tubulin, VWF, SPARC, CD63 and PF4. These markers are expressed in and/or on the platelets. Methods of the present invention and other aspects described herein may be performed using blood samples or platelets of mammalian subjects, preferably primates, more preferably human subjects.

Expression levels can be determined by various techniques known by the person skilled in the art. For example the expression levels can be determined by immunological staining techniques. For example, immunohistochemistry by detecting expressed proteins with antibodies labelled with different detectable probes (e.g., Alexa Fluor®, Oregon Green® or Pacific Blue®; horseradish peroxidase (HRP) and alkaline phosphatase (AP)). Antibodies for immunological staining techniques as well as Alexa fluorophore antibodies for the method disclosed herein, are commercially available.

In one embodiment of the in vitro method the expression level is determined using immunological staining techniques. In preferred embodiments, expression level is determined by using Alexa fluorophores.

Detecting the immunological staining can be performed using various techniques known to the person skilled in the art. In some embodiments, the immunological staining is detected by microscopy, preferably by confocal microscopy. In preferred embodiments by super- resolution microscopy. In one embodiment, the expression level is determined using immunological staining techniques in combination with super-resolution microscopy and image analysis. Alternatively, in embodiments, the expression level is determined using immunological staining techniques in combination with flow cytometry.

One of the markers of which the expression level can be determined is tubulin. Tubulin is a superfamily of globular proteins. The tubulin superfamily contains six species, including: alpha-(ct), beta-(B), gamma-(y), delta-(8), epsilon-(c), zeta-© and q-tubulins. The tubulins ct, 6 and y-tubulin are ubiquitous in eukaryotic cells and are most abundant in human platelets, in particular u-tubulin and 6-tubulin. Particularly, the known isoforms for u-tubulin in platelets are: ulA (TUBA1A), ulB (TUBA1B), ctlC (TUBA1C), u3C (TUBA3C), u4A (TUBA4A) and u8 (TUBA8). The human 6-tubulin known isoforms in platelets are: 61 (TUBB1), 62A (TUBB2A), 62B (TUBB2B), 62C (TUBB2C), 63 (TUBB3), 64 (TUBB4), 65 (TUBB), 66 (TUBB6) and 68 (TUBB8). Among them, 61 -tubulin is the major 6-tubulin isoform. In some embodiments, the tubulin marker is u-tubulin (alpha-tubulin). Suitable antibodies to determine the level of expression of tubulin are provided in the Examples below. One such an antibody targets UniProt accession P68363.

Dimers of u-tubulin and 6-tubulin polymerize to form microtubules. Microtubules are one of the major components of the cytoskeleton of the cell. The microtubules function in many processes in the cell including structural support, intracellular transport and DNA segregation. In platelets, microtubules form a subcortical ring, the so-called marginal band, which confers the typical platelet discoid shape. Resting platelets have this typical discoid shape. The marginal band or marginal band is formed by bundles of 3-24 microtubules and is also named the microtubule coil. The microtubules are also involved in the changes in platelet morphology upon activation. Preferably, staining for tubulin, preferably u-tubulin, is used as a marker for the shape of the platelets.

Von Willebrand factor (VWF) is a blood glycoprotein and is involved in hemostasis including platelet adhesion. Defects or deficiency of VWF leads to bleeding tendency and is especially known from von Willebrand disease.

VWF is synthesized by megakaryocytes and endothelial cells and is stored in the Weibel-Palade bodies of endothelial cells and in the ct-granules of platelets. VWF is also present in subendothelial connective tissue. In the platelets VWF is primarily stored in the alpha granules. VWF is released from the endothelial cells and platelets upon cell activation together with an VWF propeptide. The propeptide is also called the von Willebrand factor antigen II. The primary function of VWF is platelet adhesion to damaged vascular endothelium. VWF can bind to various other proteins, including factor VIII, platelet receptors, collagen, in particular fibrillar collagen type I and III. Suitable antibodies to determine the level of expression of VWF are provided in the Examples below. One such an antibody targets UniProt accession P04275.

Secreted protein, acidic and rich in cysteine (SPARC), is a multifunctional matricellular glycoprotein. SPARC is also known as osteonectin (ON) or basement-membrane protein 40 (BM-40). Human platelets contain and secrete SPARC.

SPARC is a glycoprotein found primarily in the matrix of bone and in blood platelets in vivo. It is known as an acidic extracellular matrix glycoprotein that plays a role in bone mineralization. Furthermore, SPARC plays a role in cell-matrix interactions as well as collagen binding. Suitable antibodies to determine the level of expression of SPARC are provided in the Examples below. One such an antibody targets UniProt accession P09486.

Cluster of differentiation 63 (CD63) was first detected as a marker of platelet activation, which is increased on the surface of platelets after granule release. It is a membrane protein with four transmembrane domains and is a member of the transmembrane 4 superfamily, which is also knowns as the tetraspanin family. CD63 is mainly associated with membranes of intracellular vesicles. In cell biology, CD63 is often used as a marker for multivesicular bodies, as well as for extracellular vesicles released from either the multivesicular body or the plasma membrane. A suitable antibody to determine the level of expression of CD63 is provided in the Examples below. A suitable antibody targets UniProt accession P08962.

Platelet factor 4 (PF4) is a small cytokine belonging to the CXC chemokine family and is also known as CXCL4. In platelets, PF4 is present in ct-granules and is released from these granules upon platelet activation. Release of PF4 promotes blood coagulation. PF4 has a high affinity for binding to heparin. The major physiologic role of PF4 appears to be neutralization of heparin-like molecules on the endothelial surface of blood vessels, thereby inhibiting local antithrombin activity and promoting coagulation. Suitable antibodies to determine the level of expression of PF4 are provided in the Examples below. One such an antibody targets UniProt accession P02776.

Fibrinogen may include the alpha chain Uniprot P02671, or beta chain Uniprot P02675.

It will be understood that antibodies or fragments thereof against proteins that are mutational variants of the proteins indicated above may also be employed in methods of this invention. Hence, the term one or more of tubulin, VWF, SPARC, CD63, PF4 and fibrinogen includes reference to mutational variants of these proteins.

In some embodiments, the method comprises determining the level of expression of one or more of tubulin, VWF, SPARC, CD63 and PF4. In some preferred embodiments, the method comprises determining the level of expression of one or more of tubulin, VWF, SPARC, CD63 and PF4 or fibrinogen. In preferred embodiments, the method comprises determining the level of expression of two or more of tubulin, VWF, SPARC, CD63 and PF4. In preferred embodiments, the method comprises determining the level of expression of three or more of tubulin, VWF, SPARC, CD63 and PF4.

In some embodiments, the method comprises determining the level of expression of tubulin and one or more of VWF, SPARC, CD63 and PF4. In preferred embodiments, the method comprises determining the level of expression of tubulin and two or more of VWF, SPARC, CD63 and PF4. In some embodiments, the method comprises determining the level of expression of tubulin, VWF, and SPARC. In some embodiments, the method comprises determining the level of expression of tubulin, CD63 and PF4.

In one embodiment, the level of expression is compared with the level of expression in reference platelets of known physiological age, in order to determine on the basis of this comparison the physiological age of the platelets in said blood sample. The comparison may be made against one or more reference platelets of same or different age. For example, the differences between levels of expression of platelets of different ages can be used in order to assess the physiological age of the platelet or platelets under consideration.

In some embodiments, instead of performing a comparison against reference platelets, the physiological age of the platelets may also be determined using a trained machine learning data processing model. For example, the reference samples or reference platelets of known age may be used in order to provide a training set for training a convolutional neural network. After training, the trained machine learning data processing model will then be able to process, as input, the levels of expression of one or more platelets in any blood sample, and provide at the output of the data processing model a determined physiological age.

The level of expression is compared with the level of expression in reference platelets of a known age. Reference platelets are obtained from a reference biological sample. The reference sample is obtained from one or more normal (e.g. healthy) reference subjects. In addition, the storage time of the reference sample is exactly known. The storage conditions of the reference sample are the standard storage conditions of donor blood. The similarity or the differences between the test and reference sample indicate the physiological age of the platelets in the test sample.

The age of platelets in a reference sample may suitably be set by assigning an age of 0 hours or days to the platelets in a freshly prepared platelet sample from freshly sampled blood from a subject, wherein the moment of taking the blood sample is set at T=0. Age is then reflected as natural aging during storage for 1-8 hours or days, wherein reference samples may be taken at different time points (e.g., T=1 to T=8). Platelet characteristics as described herein as embodiment of the invention may then be compared to the first platelet sample (T=0) measured. Reference samples may be stored platelet samples of healthy donors. In machine learning methods of this invention sampled timepoint are preferably first classified and used as a training set, where after AIT are used to classify further samples through machine learning.

The disclosure further provides a method of typing an individual suffering from thrombocytopenia. An individual or patient suffering from thrombocytopenia has too few platelets in the bloodstream. A normal platelet count ranges from 150,000 to 450,000 platelets per microliter of blood. Having less than 150,000 is known as thrombocytopenia. The platelet number can be tested for example with a blood test called a complete blood count (CBC). These patients have a decreased platelet production and/or an increased platelet clearance, resulting a lower platelet number.

Decreased platelet productions results in a low platelet number, but the platelets in the blood stream have a normal physiological platelet age. This means that the condition of the platelets spectrum found in the bloodstream is comparable with that of a healthy individual. Increased platelet clearance results in a low platelet number, but the platelets in the blood stream are relatively young. Due to increased clearance the platelets have in average a shorter life span in the circulation, thus the “older” platelets will be much less present in the bloodstream. Thus, the spectrum of the age of the blood platelets in a blood sample obtained from a patient can distinguish between different causes of thrombocytopenia. A spectrum of platelets which are relatively young is causes by increased platelet clearance. A normal spectrum of platelet age is caused by decrease platelet production.

In one embodiment the disclosure provides a method of typing an individual suffering from thrombocytopenia, the method comprising:

- determining the physiological age of platelets in a blood sample from said patient using the method as disclosed herein, and

- typing the patient based on the physiological age of the platelets determined as

(a) a patient suffering from thrombocytopenia with increased platelet clearance (characterized by young platelets), or

(b) as a patient suffering from thrombocytopenia with decreased platelet production (characterized by older platelets).

In preferred embodiments, the method further comprises determining a treatment regime. The term “treatment regime” as disclosed herein, refers to choosing effective treatments for individual patients based on the results method as disclosed herein. The method types a patient as a patient suffering from thrombocytopenia with increased platelet clearance and/or decreased platelet production. These different causes are preferably treated differently. Thus, the treatment regime includes choosing a specific therapy or a combination of therapies for a patient.

In some embodiments the disclosure provides a treatment regime aimed at decreasing platelet clearance for the treatment of an individual typed as (a) a patient suffering from thrombocytopenia with increased platelet clearance according to the method as disclosed herein.

In some embodiments the disclosure provides a treatment regime aimed at increasing platelet clearance for the treatment of an individual typed as (b) a patient suffering from thrombocytopenia with decreased platelet production according to the method as disclosed herein.

Patients suffering from thrombocytopenia benefit from getting the right treatment. The cause of the disease is important for the choice of treatment. For patients having increased clearance of platelets a medicament inhibiting the clearance is a suitable treatment, while a patient having decreased platelet production will benefit more from a treatment stimulating the production of platelets.

In one embodiment the disclosure provides a method of treating a patient suffering from thrombocytopenia, comprising the step of typing the patient by a method as disclosed herein, and treating the patient with therapeutic compounds aimed at decreasing platelet clearance or increasing platelet production, depending on the result of the typing method.

For example, therapeutic compounds aimed at decreasing platelet clearance can inhibit the production of anti-platelet antibodies, reduce the number of immune cells (e.g., B-cells) and/or inhibit phagocytosis/degradation of platelets. Therapeutic compounds aimed at decreasing platelet clearance may include glucocorticoids, intravenous immunoglobulin (IVIg), anti-RhD immunoglobulins (anti-RhD), rituximab, Syk inhibitors, BTK inhibitors, FcRn blockers and blockers of co- stimulatory molecules (e.g., humanized anti-CD40L antibodies).

Glucocorticoids prevent increased platelet clearance by decreasing the production of antibodies against platelets. Suitable glucocorticoids include prednisone and dexamethasone. Other glucocorticoids, including, but not limited to, prednisolone, methylprednisolone, cortisone and hydrocortisone, are also suitable. Several mechanisms of action have been described for glucocorticoids in the treatment of thrombocytopenia, including suppressing production of antibodies against platelets, influencing T- and B-cell reactivity and affecting cytokines. Some recent studies also suggest that glucocorticoids improve platelet counts by increasing platelet production. However, the exact mechanism remains unknown. The platelet count rises within two to four weeks of glucocorticoid therapy. Glucocorticoids represent a first-line therapy for thrombocytopenia due to their low cost, high initial response rates and acceptable short-term tolerability.

General inhibition of the antibody production and thereby the production of antibodies against platelets can be achieved by using immunosuppressive and cytotoxic agents, such as azathioprine, cyclophosphamide, cyclosporine A, mycophenolate mofetil and vinca alkaloids. Glucocorticoids may be used in combination with these immunosuppressive and cytotoxic agents. Glucocorticoids may also be used in combination with danazol or dapsone.

Another possible therapy the use of intravenous immunoglobulin (IVIg). IVIg prevents platelet degradation by binding to the Fc receptor (FcR) and directly blocking its interaction with IgG, leading to the blockade of Fc-receptor mediated platelet clearance. Other mechanism of IVIg include, but are not limited to, for example, inhibition of T-cells and dendritic cells, inhibition of the complement cascade, neutralization of pathogenic autoantibodies by anti-idiotypic antibodies, neutralizing the effect of chemokines, pro-inflammatory cytokines, and by the regulation of antibody synthesis and B cells. IVIg provides more rapid onset of the action than glucocorticoids and is generally considered to be a safe therapy due to minimal side effects. IVIg are generally recommended for severly affected or bleeding patients not responding to glucocorticoid therapy. Examples of approved IVIgs include, e.g., Panzyga®, Nanogam®, Octagam®, Privigen® and IVIG-SN™. Anti-RhD immunoglobulin (anti-RhD) is composed of polyclonal IgG prepared from the plasma of human RhD-negative donors repeatedly immunized against the D antigen and is used to treat thrombocytopenia of RhD-positive patients. Major mechanism of anti-RhD is through competitive inhibition and blockade of the mononuclear phagocytic system (MPS). After administration the anti-D-coated red blood cell (RBC) complexes compete with antibody-opsonized platelets for FCYIIIA— mediated degradation by macrophages, resulting in preferential destruction of RBCs, therefore sparing antibody-opsonized platelets. Similarly as IVIg, anti-RhD induces a rapid increase of platelets. Example of anti-RhD is WinRho SDF ®. Novel recombinant anti-RhD antibody mixture, such as rozrolimupab, may also be used. Rozrolimupab is a mixture of 25 human monoclonal IgGl antibodies against RhD. Rozrolimupab has similar efficacy as plasma derived anti- RhD.

Rituximab is a chimeric monoclonal anti-CD20 antibody and primarily depletes B cells, resulting in a decreased titers of anti-platelet antibodies. Other mechanisms of action include the normalization of T cell ratios. Rituximab is typically used to treat chronic thrombocytopenia. Examples of rituximab include Rituxan®, Truxima®, Ruxience® and Riabni®.

Syk inhibitors bind to spleen tyrosine kinase (Syk) and impair the phagocytosis of platelets via inhibition of Syk. Example of Syk inhibitor is fostamatinib (e.g., Tavalisse®).

FcRn blockers bind to neonatal Fc receptor (FcRn). FcRn is expressed by many cell types and plays a major role in the maintenance of plasma IgG levels. Inhibition of FcRn may reduce the recycle of pathogenic IgG, leading to increased catabolism of IgG. Example of FcRn blocker is rozanolixizumab. Rozanolixizumab is a humanized anti- FcRn monoclonal antibody (mAb). Blockers of co- stimulatory molecules bind to co- stimulatory molecules, such as CD40 and CD154 (also called CD40 ligand or CD40L). CD40 is mainly expressed on the antigen presenting cell (APC) and B cell, and can stimulate the activation of B cells in humoral immunity by binding to CD 154 which mainly expresses on the surface of activated CD4+ T cells. CD 154 also expresses at the activated platelet surface. It has increased expression level in ITP patients and may drive the activation of autoreactive B lymphocytes in ITP. Blocking the co- stimulatory effect of CD40-CD154 can inhibit autoreactive T-cells and B-cells, leading to increased platelet levels. Examples of said blockers include humanized anti-CD40L antibodies, such as toralizumab and ruplizumab. Other mAh, such as anti-CD44 antibodies, anti-CD80 and anti-CD86 antibodies (e.g., CTLA4-Ig) may also be used.

Further, in order to reduce platelet clearance, the individual or patient may be subjected to splenectomy. As platelets are cleared in the spleen and in the liver, patients with young platelets (e.g. fast clearance) may benefit from splenectomy where patients with production problems may benefit less from such intervention.

Examples of therapeutic compounds aimed at increasing platelet production, also known as platelet- stimulating agents or platelet growth factors can stimulate megakaryocyte differentiation and/or maturation to increase platelet production. Typical example of platelet-stimulating agents are thrombopoietin receptor agonists (TPO-RAs). TPO-RAs bind to the TPO receptor (also known as c-MpI) and activate JAK-STAT signaling pathways leading to increased platelet production. Examples of TPO-RAs include, for example, non-peptide ligands and peptide ligands of the TPO receptor. Examples of non-peptide TPO-RAs are avatrombopag (e.g., Doptelet®), eltrombopag (e.g., Promacta®, Revolade®), lusutrombopag (e.g., Mulpleta®) and hetrombopag. Example of the peptide ligand is romiplostim (e.g., Nplate®), a Fc-peptide fusion protein (peptibody). Another platelet stimulating agent is oprelvekin (Neumega®), a recombinant interleukin- 11. Oprelvekin directly stimulates the proliferation of hematopoietic stem cells and megakaryocyte progenitor cells and induces megakaryocyte maturation, resulting in increased platelet production. Oprelvekin is typically used for the prevention of thrombocytopenia in patients with or at risk of thrombocytopenia after chemotherapy (such as, e.g., myelosuppressive chemotherapy).

Platelet stimulating agents may also include anti-VPACl antibodies. VPAC1 is a receptor expressed on megakaryocytes and platelets. It has two ligands: pituitary adenylyl cyclase-activating peptide (PACAP) and vasoactive intestinal peptide (VIP). PACAP can bind to VPAC1 to inhibit megakaryocyte growth and differentiation. Anti-VPACl antibody may stimulate megakaryocyte maturation by inhibiting the interaction between PACAP and VPAC1, leading to increased platelet production.

Atorvastatin may also act as an platelet stimulating agent. Atorvastatin is lipid-lowering compound commonly used for hypercholesterolemia and atherosclerosis. Atorvastatin has also been proven to quantitatively and functionally improve bone marrow endothelial progenitor cells (BM EPCs), involved in hematopoiesis, including megakaryocyte maturation.

In some embodiments the methods as described herein may be used for a forensic test. The method as disclosed herein can be used the determine the age of the platelets since they were withdrawn from an individual. For forensic purposes it can be of value knowing the age of the blood platelets and thereby the age of a blood sample, for example of blood stains at a site of investigation.

In accordance with a further aspect, the disclosure further provides a method of training a machine learning data processing model. The machine learning data processing model may for example include a convolutional neural network, comprising an input and output layer and one or more convolution layers. The convolution layers comprise kernels of which the individual elements (i.e. weights) are to be adapted during training, such as to yield a correct output when at the input a data set is offered.

In the present disclosure, taking the example of a convolutional neural network (CNN) as machine learning data processing model, the training may be performed as follows. Preferably, a supervised learning method is applied, which is characterized by having a desired output for each given set of input signals. In order to train the CNN, a training set is therefore needed of which the correct outcome of the method to be performed is known. Thus, to train the CNN one may start with a reference blood sample for which the method of the present invention has already been carried out (e.g. manually by comparison and determination) and of which the age of one or more platelets in the reference blood sample is known. The determined outcome, thus the data about the (exact) physiological age of the blood platelets, may be used as ground truth data set. The training data set includes data on the levels of expression of the one or more platelets from the reference sample. The task in training the CNN is to adapt the weights of each kernel of each convolution layer such that eventually, providing the training data set at the input will yield the ground truth data set at the output. If the training has been performed most optimal, comparing the output of the CNN with the ground truth data set will thus yield no or minimal differences for any arbitrary reference sample. Such differences are reflected by the error function, and hence it will be the error function that needs to be minimized eventually during training.

Various adaption strategies are known to the skilled person and may be applied in order to converge the CNN efficiently during training. For example, a gradient descent method may be applied in order to minimize the error function by iteratively adapting the weights of the kernels such as to yield a reduction of the error function with each iteration. The iterations are continued until the reductions become too small or no reduction can be found anymore. Alternative training methods may likewise be applied. This may be done for a CNN with any arbitrary number of layers or kernel sizes. The present disclosure is not limited in terms of numbers of layers or kernel sizes. Furthermore, same can be done with different types of machine learning data processing models, such as a regular artificial neural network consisting of layers with weights and biases for converting neurons, rather than kernels.

Preferably, a large training data set is used in order to train the CNN, however any size of training set will do as long as the results of the trained CNN are satisfactory.

In another aspect, the present invention provides a system for determining the age of platelets in a blood sample. Such a system comprises one or more processors for receiving measurement data indicative of a level of expression of one or more of tubulin, VWF, SPARC, CD63 and PF4 in at least one individual platelet within a platelet sample obtained from a blood sample. The system further may include a memory for storing a trained machine learning model, e.g. such as the trained CNN described above. The trained machine learning model is configured to output a classification of the at least one individual platelet based on the measurement data. The one or more processors are configured to determine data indicative of the physiological age of the at least one individual platelet based on the output classification.

All embodiments described for different aspects of the present invention may be combined and such combinations of disclosed embodiments are themselves an embodiment of aspects of the present invention. All patents and literature references cited in the present specification are hereby incorporated by reference in their entirety. The invention is further explained in the following examples. These examples do not limit the scope of the invention, but merely serve to clarify the invention.

EXAMPLES

In support of the invention described above, the following two examples are provided. Example 1 relates to platelets in PRP, isolated from whole blood drawn from healthy donors. These platelets in PRP were stored in a tube on the bench for 8 hours. Using the model described herein it provide possible to predict whether platelets were stored for 0, 4 or 8 hours (Figure 1). Example 2 describes an experiment wherein platelet concentrates (PC) which are used in routine clinical practice are analysed. Three different routine single- donor PC were prepared by Sanquin and were stored for 10 days. At the timepoints indicated a sample of each PC was taken and the samples were pooled to a single sample. This sample was then analyzed. IN Example 2, platelet storage could be predicted accurately for days 1, 2, 3, 6, 7, 8, 9 or 10 and it was possible to distinguish young (platelets stored <6 days) from old platelets (platelets stored for 10 days) (Figure 2)

Antibodies used

For staining of VWF, Rabbit Polyclonal antibody DAKO A0082 and Sanquin CLB-Rag20 (Stel et al. 1984. Blood.63(6): 1408-1415) were used. For staining of alpha-tubulin, mouse IgG2b (Abeam, ab56676) and IgGl (Sigma, DMla) were used. For staining of SPARC, Mouse IgGl (SantaCruz, sc-73472, AON-5031) was used. For staining of CD63, CLB-granl2 M1544 (Sanquin) was used, and for staining of PF4 MAB7591 (R&D Systems, Inc. Minneapolis, USA) was used. For staining of Fibrinogeen, DAKO A0080 was used. Example 1 Platelets in PRP stored for 8 hours

Methods

Platelets were stored in platelet-rich plasma (PRP) on the bench for 8 hours at room temperature. Platelets were isolated and stained with antibodies for alpha-tubulin, von Willebrand Factor (VWF) and SPARC or alpha-tubulin and imaged at T=0, 4 and 8 hours using confocal microscopy. A neural network was trained consisting of multiple convolutional layers to classify age (time after donation) of platelets that were stored in either PRP or PC.

Results

During storage in PRP as well as of PC for each time point 1443 62gm x 62gm x 4gm 3D images were used, containing 50 to 150 platelets per image, for training. When staining with markers VWF, SPARC or alpha-tubulin alone as training set we could already classify mean platelet age up to an accuracy of 78%, 85%, 88% respectively. When combining all three markers a maximum accuracy of 95% for platelets stored in PRP up to 8 hours was reached (Figure 1). The additional effect of combining these three markers highlights that the model used information in the images synergistically. For PC, an accuracy of 90% was reached for classifying mean platelet age of platelets stored up to 8 days. The model could not be trained to discriminate between days 9 and 10 resulting in an overfitted model. However the model was able to distinguish day between day 1, 3, 6 and 10 (Figure 2). As negative control a model trained on a randomized sample was made. This control model was not able to assign a category correctly, indicating an actual difference in platelet characteristics during storage was detected. By using visualization of the weight-gradient (e.g. where in the images the model assigns the most weight) we additionally confirmed that the model used separate individual platelets to characterize the age of the complete platelet sample. Conclusion

Combining confocal platelet images with a convolutional neural network, we trained a model that could synergistically integrate different stainings and information of individual platelets to predict with >95% accuracy platelet age in vitro as well as in PC up to 10 days of storage. Additionally our model could distinguish “young” platelets (stored 1, 3 or 6 days) from “old” platelets (stored 10 days). The trained model can predict with >95% accuracy the storage time of a platelet in vitro and in platelet concentrates up to 10 days of storage. Knowing the exact age of platelets, and thereby its function, may have major consequences in the field of platelet transfusion medicine and in the treatment of patients with thrombocytopenia.

Example 2 Experiment using routine platelet concentrates stored according routine clinical practice

Material and methods

(Essentially as described in Example 1)

PRP Platelet Isolation and Storage

All steps were carried out at room temperature (RT) unless otherwise stated. Whole blood was drawn from consenting healthy donors in citrate tubes. Washed platelets were prepared by methods known in the art. Whole blood was centrifuged for 20 minutes at 120 x g with low acceleration (max. 5) and low brake (max. 3) to generate platelet-rich plasma (PRP). PRP was stored for 0, 4 or 8 hours at RT. Additionally, platelets of 3 single donors were combined and stored at blood bank conditions in PAS-E at RT while being gently shaken. Platelets were harvested after 1, 2, 3, 6, 7, 8, 9 and 10 days of storage. Platelets, from the two storage conditions, were fixed in 1% formaldehyde for 5 minutes after indicated storage time, then washed three times in washing buffer (36 mM citric acid, 103 mM NaCl, 5 mM KC1, 5 mM EDTA, 5.6 mM glucose, pH 6.5).

Sample preparation and staining

All unique sample conditions were seeded on poly-D-lysine coated 9 mm diameter 1.5H high-precision coverslips (Marienfeld), permeabilized, and stored in PGAS (0.2% gelatin, 0.02% azide and 0.02% saponin in PBS). Platelet content was stained for VWF and SPARC or CD63 and PF4. In all samples u-tubulin was stained as a marker for cell shape. Primary and secondary antibody staining was done in PGAS for 30 minutes at RT, washed 3x with PGAS following incubations. Finally, slides were dipped in PBS, mounted in Mowiol.

Image Acquisition

Samples were imaged with a Leica Stellaris 5 confocal microscope fitted with a 63x 1.4NA HO PL APO CS2 lens. Fluorophores were exited with 488nm, 561nm or 638nm diode lasers. Emission was filtered between 500-550nm, 570-620nm, 650,700nm for Alexa Fluor488, CF568 Alexa Fluor 647 signals respectively. Datasets were made using Leica navigator software by making a tilescan containing 1443 frames without overlap, for each frame a z-stack of 5 frames was taken with an interval of lgm. The sample was kept in focus by defining five focus point per region. The acquired images were converted to RGB png files for loading into the model.

Neural network modelling

A neural network was designed using the Keras and Tensorflow api’s in Python (Chollet, F. et al. (2015). Keras. GitHub. Retrieved from https://github.com/fchollet/keras). The model consisted of 10 convolutional layers applied in pairs interspersed with 5 maxpooling layers, followed by three dense layers combined with 3 dropout layers. Activation method for all layers with exception of the last dense layer was “relu”. The last dense layer activation method was “softmax”. The used optimizer was adam used with a learing rate of 0.00001, beta_l of 0.69 and beta_2 0.9. The used loss function was a categorical cross-entropy.

The dataset consisting of 1443 images of platelets per condition/age was used to teach the model. For validation a split of 0.2 was used, the model was run for 20-100 epochs. Accuracy and loss were monitored. The dataset was further validated using a separate dataset from an independent sample with the similar number of images.

Example 3 Platelets in PRP stored for 8 hours

Methods

The method of Example 1 was repeated, and platelets were isolated and stained with antibodies for alpha-tubulin, CD63 and/or PF4.

Results

When staining with markers VWF, SPARC or alpha-tubulin alone as training set we could already classify mean platelet age up to an accuracy of 69%, 68% and 86% respectively. When combining all three markers a maximum accuracy of 88% was reached. The additional effect of combining these three markers again highlights that the model used information in the images synergistically.

Claims

Claims

1. An in vitro method for determining the physiological age of blood platelets in a test sample wherein said physiological age is expressed in days and/or hours, the method comprising: a) providing a test sample of blood platelets; b) determining the level of expression of one or more of tubulin, VWF, SPARC, CD63 and PF4 in one or more of individual platelets within said test sample, c) determining, based on the level of expression determined under b), the physiological age of the blood platelets in said test sample.

2. An in vitro method according to claim 1, wherein for determining the physiological age of the platelets in said test sample, step c) comprises comparing the level of expression determined under b) with the level of expression in a reference sample of blood platelets of known physiological age, and determining the physiological age of the blood platelets in said test sample based on said comparison.

3. An in vitro method according to claim 1, wherein for determining the physiological age of the blood platelets in said test sample, step c) comprises using a trained machine learning data processing model for automatically associating the level of expression determined in step b) with a physiological age of the one or more platelets, such as to enable providing the age of the one or more platelets at an output of the machine learning data processing model.

4. An in vitro method according to any of the preceding claims, wherein the expression level is determined using immunostaining in a multitude of individual platelets from said platelet sample, preferably in combination with confocal or super-resolution microscopy and image analysis of individually stained platelets.

5. An in vitro method according to any of claims 2-4, wherein the level of expression is compared using a machine learning data processing model.

6. An in vitro method according to any of the preceding claims, wherein the level of expression is determined by immunostaining of at least tubulin, VWF and SPARC proteins in said platelets or by immunostaining of at least tubulin, CD63 and PF4 proteins in said platelets, preferably in combination with image analysis of fluorescence intensities measured from said immunostaining.

7. An in vitro method according to any of claims 4-6, wherein the image analysis further comprises collection of data on the morphology of the platelet, preferably the discoid shape of said platelets and the presence of granules in said platelets, preferably wherein staining for tubulin is used as a marker for the shape of the platelets.

8. An in vitro method according to claim 6, wherein tubulin is alphatubulin.

9. A method of typing an individual suffering from thrombocytopenia, the method comprising: providing a blood sample from said individual; determining the physiological age of platelets in a blood sample from said individual using the method of any one of claims 1-8, and typing the individual based on the physiological age of the platelets determined as (a) an individual suffering from thrombocytopenia with increased platelet clearance, or

(b) an individual suffering from thrombocytopenia with decreased platelet production.

10. The method of typing an individual suffering from thrombocytopenia according to claim 9, further comprising assigning a treatment regime to said patient wherein said treatment regimen is aimed at decreasing platelet clearance in said individual in case said individual is typed as (a) an individual suffering from thrombocytopenia with increased platelet clearance, or wherein said treatment regimen is aimed at increasing platelet production in case said individual is typed as (b) an individual suffering from thrombocytopenia with decreased platelet production.

11. A method of treating an individual suffering from thrombocytopenia, comprising the steps of

- typing the patient using a method according to claim 9 or 10, and

- administering to said individual a therapeutically effective amount of a medicament aimed at decreasing platelet clearance or aimed at increasing platelet production, depending on the outcome of the typing method.

12. A treatment regime for use in a method of treating an individual suffering from thrombocytopenia according to claim 11, wherein said regimen is aimed at decreasing platelet clearance for the treatment of an individual typed as (a) an individual suffering from thrombocytopenia with increased platelet clearance, or wherein said regime is aimed at increasing platelet production for the treatment of an individual typed as (b) an individual suffering from thrombocytopenia with decreased platelet production.

13. A method for determining the quality of a platelet concentrate for transfusion medicine, comprising performing a method according to any one of claims 1-8 on a sample of said platelet concentrate.

14. Method of training a machine learning data processing model for determining the age of platelets in a blood sample, the method comprising: a) receiving, by the machine learning data processing model, a training data set comprising training data indicative of a first level of expression of one or more of tubulin, VWF, SPARC, CD63 and PF4 in a representative number of first individual platelets within a platelet sample, wherein preferably the first level of expression is determined from the platelet sample using immunological staining techniques in combination with super-resolution microscopy and image analysis; b) receiving, by the machine learning data processing model, a ground truth data set comprising ground truth data indicative of the physiological age of the first platelets, wherein the physiological age of the first platelets is determined using the method of claim 1; c) training the machine learning data processing model based on the training data received in step a) and the ground truth data received in step b) for enabling, after completion of the training period, the step of automatically associating measurement data indicative of a second level of expression of one or more of tubulin, VWF, SPARC, CD63 and PF4 in at least one second individual platelet with an physiological age of the at least one second platelet, such as to enable providing the age of the at least one second platelet at an output of the machine learning data processing model.

15. A system for determining the age of platelets in a blood sample, the system comprising: one or more processors for receiving measurement data indicative of a level of expression of one or more of tubulin, VWF, SPARC, CD63 and PF4 in at least one individual platelet within a platelet sample obtained from a blood sample; and a memory storing a trained machine learning model, wherein the trained machine learning model is configured to output a classification of the at least one individual platelet based on the measurement data; and wherein the one or more processors are configured to determine, data indicative of the age of the at least one individual platelet based on the output classification.