Method for determining secretor status in human milk using infrared spectroscopy

Abstract

The present invention relates to a method for determining the maternal secretor phenotype in a human milk (HM) sample using IR spectroscopy, the method comprising the steps of: (a) providing a previously obtained HM sample; (b) subjecting the HM sample to IR spectroscopy analysis to generate an IR spectrum in the range of 4000 to 400 cm⁻¹, wherein the signals in this region are associated with the presence of specific HMOs; (c) comparing the IR spectrum of the HM sample obtained in (b) with a discriminant model, wherein the discriminant model uses a set of IR spectra from HM samples with a known maternal secretor status phenotype recorded in the same spectral range; and (d) determining whether the maternal secretor phenotype of the HM sample is a secretor (Se+) or non-secretor (Se-) based on the differences determined in (c).

Description

[0001] METHOD FOR DETERMINING THE SECRETORY STATE IN HUMAN MILK BY INFRARED SPECTROSCOPY

[0002] DESCRIPTION

[0003] Field of invention

[0004] The present invention relates generally to the Health and Nutrition Sector, and more particularly to Neonatal Care. In particular, the invention relates to a method for determining the maternal secretory phenotype in a human milk (HM) sample using infrared (IR) spectroscopy.

[0005] Background of the invention

[0006] In Spain, approximately 22,000 premature babies are born each year, with an average hospitalization cost of €40,000 and an average stay of 50 days. Personalized human milk (HM) administration, including supplementation tailored to premature infants, has significant clinical and economic importance. Optimizing nutrition through HM can improve the health outcomes of these vulnerable babies, reducing complications associated with prematurity and promoting their growth and development. This, in turn, could shorten hospital stays, directly impacting their quality of life. Furthermore, reducing a premature infant's hospital stay by just one day could represent savings of between €400 and €800, translating to annual savings of between €8 and €17 million for the healthcare system.Furthermore, studies have shown that the type of enteral nutrition plays a critical role in both clinical outcomes and healthcare costs: an exclusive LH diet resulted in a net savings of $8,167 per preterm infant in hospital costs. This underscores the transformative potential of personalized nutritional interventions, both for healthcare systems and for the well-being of preterm infants.

[0007] Human milk oligosaccharides (HMOs) play a vital role in the development of the gut microbiome and the immune function of the infant (Plaza-Díaz J, et al. Human Milk Oligosaccharides and Immune System Development. Nutrients. August 8, 2018; 10(8): 1038) and in the regulation of the microbiome and the prevention of gastrointestinal disorders (Nolan LS, et al. The Role of Human Milk Oligosaccharides and Probiotics on the Neonatal Microbiome and Risk of Necrotizing Enterocolitis: A Narrative Review. Nutrients. October 6, 2020; 12(10):3052). The most important source of variation in HMO production between individuals is attributed to genetic variations in the fucosyltransferase 2 (FUT2) and 3 (FUT3) genes, known as the secretor and Lewis genes, respectively.Mothers carrying inactivating mutations in FUT2 are called non-secretors (Se-) and represent about 20% of the European population (Samuel TM, et al. Impact of maternal features on human milk oligosaccharide composition over the first 4 months of lactation in a cohort of healthy European mothers. Scientific Reports. 2019 Aug;9(1):11767). The milk of non-secretors (Se-) mothers lacks or has only traces of 1-2 fucosylated HMOs, such as 2'-fucosyl-lactose (2'-FL), lacto-N-fucopentaose I (LNFP I), and lactodifucotetraose (LDFT), which are among the most abundant HMOs in secretors (Se+). The maternal secretory phenotype has been linked to an increased risk of preterm delivery (Caldwell J, et al. Maternal H-antigen secretor status is an early biomarker for potential preterm delivery. Journal of Perinatology: Official Journal of the California Perinatal Association.2021 Sep;41(9):2147-2155), lower risk of atopic dermatitis in babies born by cesarean section (Sprenger, N. et al. FUT2-dependent breast milk oligosaccharides and immunity at 2 and 5 years of age in infants with high hereditary immunity risk (2017) European Journal of Nutrition, 56(3), 1293-1301), and with an earlier establishment of a microbiota loaded with bifidobacteria in breastfed babies (Lewis, ZT, et al. Maternal fucosyltransferase 2 status affect the gut bifidobacterial communities of breastfed infants. (2015). Microbiome 3, 13). Furthermore, it has an influence on the fecal metabolome (Wang A, et al. Impact of milk secretor status on the fecal metabolome and microbiota of breastfed infants. Gut Microbes. 2023 Dec; 15(2) .2257273), with the content of HMOs in milk from secretor mothers correlating with the urinary and fecal content of specific HMO structures (Underwood MA, et al.Human milk oligosaccharides in premature infants: absorption, excretion, and influence on the intestinal microbiota. 2015 Dec. Pediatr Res.;78(6):670-7).

[0008] Regarding the influence of secretory status and HMO content in relation to preterm birth, it was found that HMO fucosylation, and therefore 2'-FL content, was inconsistent in women who delivered preterm infants, which could be attributed to mammary immaturity (De Leoz ML, et al. Lacto-N-tetraose, fucosylation, and secretor status are highly variable in human milk oligosaccharides from women delivering preterm. 2012 Sep 7. J Proteome Res.;11(9):4662-72). Furthermore, it has been concluded that the secretory phenotype and genotype of infants can provide strong predictive biomarkers of adverse outcomes, such as necrotizing enterocolitis (NEC) and sepsis, that cannot be inferred through standard clinical factors (Ardythe L. Morrow, et al.).Fucosyltransferase 2 Non-Secretor and Low Secretor Status Predicts Severe Outcomes in Premature Infants, The Journal of Pediatrics, Volume 158, Number 5, 2011, Pages 745-751, ISSN 0022-3476). Therefore, identifying non-secretor premature infants could be of great interest so that these infants at higher risk of developing adverse outcomes can receive milk rich in fucosylated HMOs.

[0009] This identification would also have potential implications for milk bank practices, as matching donor milk to maternal secretor status could be a strategy for personalizing donor milk administration and providing infants of non-secreting mothers with milk from secretory donors. A clinical trial is currently underway to evaluate the impact of matching donor human milk to maternal secretor status on the gut microbiome of very preterm infants (<34 weeks of gestation) (NCT04130165. Matching Donor Human Milk On Maternal Secretor Status (MMOMSS) Study), https: / / clinicaltrials.gov / study / NCT04130165#study-plan, (accessed February 29, 2024) (https: / / register.clinicaltrials.gov, identifier: NCT05646940).

[0010] Inferring maternal secretory status through quantification of 2'-FL in milk is generally performed using chromatographic techniques and is therefore not amenable to point-of-care (PoC) testing and requires trained personnel. Chung et al. developed an electrochemical PoC test for maternal secretory status in human milk (Chung, S., et al. Point-of-care human milk testing for maternal secretor status. Anal Bioanal Chem 414, 3187-3196 (2022)). However, it requires the fabrication of a dedicated biosensor and an incubation step for sample analysis, which limits its applicability. Furthermore, this technology has not been validated and is not commercially available.

[0011] The article “Xun, Yiping et al. Profile of Twenty-Three Human Milk Oligosaccharides in Han Chinese Mothers throughout Postpartum 1 year, Journal of Food Quality, 2022, 6230832, 9 pages” describes a study that aimed to identify the concentration profile of human milk oligosaccharides (HMOs), the composition of the milk microbiota, and their associations with key maternal characteristics in Chinese mothers during a one-year lactation period. Carbohydrates were detected using the MIRIS human milk analyzer, and twenty-three HMOs were quantified by ultra-performance liquid chromatography-triple quadrupole mass spectrometry (UPLC-MS). The results showed that carbohydrates were relatively stable, while total HMO concentrations ranged from 1.74 to 9.72 g / L and gradually decreased during lactation.In addition, other maternal factors, including genetics, pandad, and microbiota, are also related to individual HMO profiles, making further research studies necessary.

[0012] In the biomedical field, IR spectroscopy-based tools are emerging for the analysis and characterization of biological samples. These techniques typically rely on the rapid and direct acquisition of the sample's IR spectrum after minimal or no sample preparation, using cost-effective instrumentation and without expensive reagents or consumables.

[0013] Currently, milk banks measure milk macronutrients using mid-infrared (mid-IR) milk analyzers. These systems provide information on the amount of energy, fat, carbohydrate, and protein content in human milk, but do not provide information on human moieties (HMOs) (Billard H. et al. Calibration Adjustment of the Mid-infrared Analyzer for an Accurate Determination of the Macronutrient Composition of Human Milk. (2016) J Hum Lac 32(3), 19). Therefore, determining maternal secretory status using IR spectroscopy-based techniques for milk analysis could revolutionize both the nutritional practices of preterm infants and the practices of milk banks, significantly impacting the well-being of preterm infants.

[0014] However, as revealed in the article “Belfort, MB et al. Deciphering macronutrient information about human milk. (2024). J Perinatol 44, 1377-1381”, there are significant difficulties in distinguishing lactose and HMOs using IR devices due to similarities in their chemical structures, specifically that of terminal lactose in HMOs. Carbohydrate estimates using IR are influenced by calibration and instrumentation. In a study of four different IR analyzers used in three milk bank settings, they yielded significantly different results on a shared set of samples. The calibration of IR instruments for lactose or total carbohydrates will strongly influence carbohydrate estimates.Instruments calibrated for total carbohydrates systematically overestimate available energy in human milk (HM) by assigning the same energy values ​​to digestible (lactose) and indigestible (HMOs) carbohydrates. Given this context, there is an urgent need for accurate, accessible, and efficient methods to determine maternal secretory phenotype from HM samples, considering its relevance to personalized nutrition for preterm infants, the design of donor milk compatibility strategies in HM banks, and the optimization of clinical outcomes related to the infant gut microbiome and immune system development. Existing chromatographic methods, while accurate, require specialized equipment and trained personnel and are not suitable for point-of-care (POC) applications.Similarly, emerging technologies, such as electrochemical sensors, face significant limitations in terms of commercial availability and usability.

[0015] To address these needs, the authors of the present invention, after carrying out significant experimental work, have developed a method that uses IR spectroscopy to determine the maternal secretory phenotype by analyzing specific spectral absorption bands associated with HMOs.

[0016] By integrating IR spectroscopy, this method represents an innovative solution that directly addresses the technical and operational limitations of the state of the art, providing a fast, non-invasive and cost-effective method, compatible with the analytical systems available in milk banks.

[0017] This invention has the potential to transform LH nutrition and management practices by allowing for personalized milk administration, particularly in premature infants and scenarios where the secretor phenotype directly influences clinical outcomes.

[0018] Brief description of the figures

[0019] Figure 1. ATR-FTIR spectra of dried residues obtained from 2 pL of an HMO standard panel. Abbreviations: LNFP I (lacto-N-fucopentaose I); 2'-FL (2'-fucosyl-lactose); 3-FL (3-fucosyl-lactose); 6'-SL (6'-sialyl-lactose); LNDFH I (lacto-N-difucopentaose I); LNT (lacto-N-tetraose); DSLNT (disialyl-lacto-N-tetraose); 3'-SL (3'-sialyl-lactose).

[0020] Figure 2. ATR-FTIR spectra of average dry residues obtained from 2 pL of 37 LH samples from secretory mothers and 23 LH samples from non-secretory mothers. Figure 3. PLS-DA model score chart for maternal secretory status determination (A). Results of secretory status classification of LH samples for the non-secretory group in calibration (B) and cross-validation (CV) (C). Note: the red dashed line denotes the discrimination limit between the groups.

[0021] Figure 4. Fractional variance Y* captured for self-prediction (calibration, C) and CV versus the correlation of the swapped Y-block with the original Y-block. The sum of squares of Y for calibration (SSQY.C) is expected to increase to a value of "1" when the model captures the entire response of the Y-block. The sum of squares of Y for CV (SSQY.CV) is expected to be approximately the same as SSQY.C as long as the model is not overfitted.

[0022] Figure 5. Results of the predicted secretory status of 25 LH samples from secretory and non-secretory mothers.

[0023] Detailed description of the invention

[0024] Based on the needs of the state of the art, the authors of the present invention have developed a method that allows the in situ determination of the maternal secretory state phenotype by direct measurement of LH by IR spectroscopy.

[0025] In one main aspect, the invention relates to a method for determining the maternal secretory phenotype in an LH sample by IR spectroscopy ('method of the invention'), comprising:

[0026] (a) provide a previously obtained LH sample;

[0027] (b) subject the LH sample obtained in a) to an IR spectroscopy analysis to generate an IR spectrum in the range of 4000 to 400 cm -1, where the signals in this region are associated with the presence of specific HMOs;

[0028] (c) compare the IR spectrum of the LH sample obtained in b) with a discriminative model, wherein the discriminative model uses a set of IR spectra of LH samples with known maternal secretory state phenotype recorded in the same spectral range; and (d) determine the maternal secretory phenotype of the LH sample as secretor (Se+) or non-secretor (Se-) based on the differences determined in c).

[0029] In a preferred embodiment, the IR spectrum in b) is generated in the range of 925 to 585 cor 1In this region, fucosylated HMOs ± 1-2, related to the maternal secretor phenotype, such as 2'-fucosyl-lactose (2'-FL), lacto-N-fucopentaose I (LNFP I), and lactodifucotetraose (LDFT), three of the most abundant HMOs in secretor mothers (Se+), exhibit bands. In particular, (2'-FL) and LNFP I show two bands (820 cm⁻¹). -1 and 764 cm 1 ) that could be distinguished from the bands presented in other non-fucosylated measured HMOs « 1-2. The band at 820 cm -1 It is attributed to wobbling and deformation vibrations of the C-6 methyl group of fucose, coupled with out-of-plane bending vibrations of the CH bond of the pyranose ring, while the band at 746 cm -1 It is assigned to bending modes of the ring skeleton and the COC glycosidic bond, which are strongly influenced by the geometry of the δ1-2 bond with the terminal galactose. Taken together, the bands at 820 and 746 cm⁻¹ -1They therefore represent diagnostic bands.

[0030] In the present invention, the term “LH sample” refers to the milk produced by the mammary glands of women's breasts (breast milk). It is defined as a Substance of Human Origin (SoHO) and is regulated according to REGULATION (EU) 2024 / 1938 OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL of 13 June 2024.

[0031] The set of IR spectra from LH samples with a known maternal secretory state phenotype used in c) allows for the construction of a discriminatory model. The discriminatory model is constructed using a multivariate statistical model, selecting the 4000 to 400 cm⁻¹ region. -1 , preferably 925 to 585 cm -1, yi) applying the second derivative transform to the set of IR spectra of the LH samples to obtain the transformed data followed by i) calculating the “mean centering” of the transformed data.

[0032] Based on the results of the multivariate statistical model, the sample can be classified as belonging to a secretory or non-secretory mother.

[0033] Therefore, by directly measuring the IR spectrum of LH and using the predefined criteria established in the validated multivariate model, the maternal secretory state phenotype can be accurately determined, allowing the identification of secretory (Se+) and non-secretory (Se-) mothers with a sensitivity = 1.00 and a specificity = 0.97 in a validation study (n = 60), as shown in the examples described below (see Table 1).

[0034] In the method of the present invention, the spectral analysis can be carried out using different IR spectral techniques used in the prior art, such as, for example, Fourier transform infrared spectroscopy (FTIR), transmission infrared spectroscopy, reflectance infrared spectroscopy, infrared microscopy, mid-infrared scattering spectroscopy, and photoacoustic spectroscopy.

[0035] In one particular embodiment, the spectral analysis performed in the method of the invention comprises attenuated total reflectance Fourier transform IR (ATR-FTIR) spectroscopy analysis.

[0036] In another particular embodiment, the analysis performed in step (b) may employ one type of spectral method, for example ATR-FTIR, together with one or more different spectral methods.

[0037] In one particular embodiment, the spectrometer used in the method of the present invention comprises an FTIR spectrometer. Preferably, for clinical purposes, the spectrometer is a portable spectrometer, which allows for the in situ determination of the maternal secretory status phenotype.

[0038] In another aspect of the invention, the determination of the maternal secretory state phenotype using the method of the present invention is used to prevent, diagnose, or monitor health conditions associated with the maternal secretory state phenotype. In one particular embodiment, these conditions are selected from among gastrointestinal disorders, allergies, and alterations in the growth and / or development of the immune and neurological systems in infants (including premature infants).

[0039] The following examples illustrate the present invention:

[0040] EXAMPLES

[0041] Materials and methods

[0042] 1.1 Breast Milk Samples. Eighty-five breast milk samples were collected as part of the NUTRISHIELD study (Ramos García, V. & Ten-Doménech, I. et al. (2023). Fact-based nutrition for infants and lactating mothers—The NUTRISHIELD study. Frontiers in Pediatrics. Vol. 11). Specifically, 19 samples were collected from 16 mothers of preterm infants and 49 samples from 26 mothers of full-term infants. Milk samples were collected in the morning (between 7 and 10 am) from one breast after complete expression. In addition, 10 pasteurized breast milk samples and 7 unpasteurized breast milk samples were included. The latter samples were provided by 13 donors admitted after routine screening at the breast milk bank (Bancompass of the Valencian Community, Valencia, Spain).

[0043] The study was approved by the Biomedical Research Ethics Committee of the La Fe Health Research Institute, La Fe University and Polytechnic Hospital (Valencia, Spain) under registration number 2019-289-1, and all methods were performed in accordance with relevant guidelines and regulations. Written informed consent was obtained from breastfeeding mothers prior to sample collection, along with demographic and clinical information. The study is registered at https: / / register.clinicalthals.gov, identifier: NCT05646940.

[0044] 1.2 Chemical products and reagents

[0045] LC-MS grade acetonithol (>99.9%) was obtained from Scharlau (Barcelona, ​​Spain), formic acid (>95%), trifluoroacetic acid (TFA) (>99%), ammonium acetate (>98%), and chloroform (>99.5%) from Sigma-Aldhch Química SL (Madrid, Spain). LC-MS grade ethanol (>99.5%) and methanol (>99.9%) were purchased from Labkem (Barcelona, ​​Spain). Ultrapure water was generated using a Merck Millipore Milli-Q integrated water purification system (Darmstadt, Germany). The sodium salts of 3'-sialyl-lactose (3'-SL) (>98%), lacto-N-tetraose (LNT) (>95%), lacto-N-fucopentaose I (LNFP I) (>95%), 2'-fucosyl-lactose (2'-FL) (>95%), 3-fucosyl-lactose (3-FL) (95%), and lacto-N-difucohexaose I (LNDFH I) (>95%) were obtained from Carbosynth (Bratislava, Slovakia). The sodium salts of 6'-sialyl-lactose (6'-SL) (>97%), disialyl-lacto-N-tetraose (DSLNT) (>95%), and isomaltothose DP3 were obtained from Merck (Darmstadt, Germany).

[0046] 1.2.1 Preparation of stock solutions

[0047] Individual stock solutions of standards were prepared (i.e., 12.5 mmol L' 1 2'-FL, 4.4 mmol L' 1 LNT, 5.3 mmol L' 1 LNFP I, 1 mmol L' 1 LNDFH I, 7.2 mmol L' 1 3-FL, 1 mmol L 1 3'-SL, 1.6 mmol L -1 6'-SL and 1 mmol L' 1 DSLNT) in water and were stored at -20 °C.1.3 ATR-FTIR Spectroscopy

[0048] IR spectra in the range of 4000 to 400 cm -1Spectra were acquired using an Alpha II FTIR spectrometer (Bruker Optics GmbH, Ettlingen, Germany) equipped with a platinum ATR with a monolithic diamond measuring interface element (single reflection), a CenterGIow™ IR source, and a temperature-stabilized DTGS detector without a purge system. OPUS 8.5 software (Bruker Optics GmbH) was used to control the instrument. 2 µL of the LH sample or stock solution were poured onto the ATR crystal using a micropipette and air-dried for 5 min at room temperature. Spectra were collected by adding 24 scans with a resolution of 4 cm⁻¹ -1 Using a previously recorded air spectrum under the same instrumental conditions as a background, with an acquisition time of 30 s. Between samples, the ATR interface was cleaned with a cotton swab and H2O until the reference signal was recovered. The order of spectral acquisition was random.

[0049] 1.4 HILIC-MS / MS Analysis (reference analysis)

[0050] 1.4.1 Preparation of standard solutions

[0051] By mixing appropriate volumes of the individual stock solutions, a working solution with 600 mol L was prepared 1 of 2'-FL and LNT; 300 ¡i mol L' 1 of LNFP I and LNDFH I; and 80! mole L' 1 of 3-FL, 3'-SL, 6'-SL and DSLNT in H2O. A calibration curve was then prepared by serial dilution in water:acetonitrile (1:1, v / v) with concentrations ranging from 3 to 600 pmol L' 1 depending on the HMO, and with isomaltothose as an internal standard (IS) at a concentration of 8 mol L' 1 .

[0052] 1.4.2 Extraction and purification of HMOs

[0053] The extraction of HMOs was performed following a previously described protocol (V. Navarro-Esteve, et al. Simultaneous screening and quantitation of human milk oligosaccharides by liquid chromatography — Mass spectrometry, Carbohydrate Polymer Technologies and Applications, Volume 9, 2025, 100644, ISSN 2666-8939). Briefly, the human milk samples were thawed at room temperature and mixed by inverting them ten times. A 50 µL aliquot of human milk was extracted, diluted with 200 µL of Milli-Q H2O, and centrifuged at 14,000 x g at 4 °C for 30 min. 200 µL of the liquid layer were mixed with 800 µL of chloroform:methanol (2:1 v / v) and centrifuged at 14,000 x g at 4 °C for 30 min. Then, 300 µL of the aqueous phase were transferred and mixed with 600 µL of ethanol. The resulting mixture was incubated at 4 °C overnight for protein precipitation and then centrifuged at 14,000 x g at 4 °C for 30 min.The liquid phase was carefully collected and evaporated to dryness using a miVac centrifugal vacuum concentrator (Genevac LTD, Ipswich, UK) at 50 °C. The dried residue was redissolved in 192 L of H2O and 8 L of a 250 mol L⁻¹ solution. 1IS prior to solid-phase extraction (SPE). A Thermo Scientific™ HyperSep Hypercarb 96-well SPE plate (Rockwood, TN, USA) (50 mg bed weight) was pre-conditioned with 500 µL of water, 500 µL of 80% v / v acetonitrile with 0.1% v / v TFA, and 500 µL of water. After loading the sample, unwanted and weakly bound components were washed with 600 µL of water. Neutral and acidic HMOs were then co-eluted with 400 µL of 20% v / v acetonitrile (0.05% v / v TFA) and 400 µL of 40% v / v acetonitrile (0.05% v / v TFA). The eluted samples were evaporated to dryness in a miVac centrifugal vacuum concentrator and reconstituted in 250 µL of 50% v / v acetonitrile. A blank extract was also prepared following the steps described for the LH samples, but replacing the LH with water. A pooled quality control (QC) sample was prepared by mixing 130 µL of each LH sample extract.The accuracy and precision levels of the method, the performance of the SPE extraction, and the effect of the matrix were evaluated in triplicate by adding an LH sample, a preprocessed LH sample before SPE, and an LH extract recovered after SPE, respectively.

[0054] 1.4.3 HILIC-MS / MS Method

[0055] For HMO detection, a UPLC system coupled to an Orbitrap QExactive Plus MS (Thermo Fisher, Waltham, MA, USA) was used. Chromatographic conditions were adapted from other locations. An ACQUITY Glycoprotein BEH Amide precolumn (1.7 µm, 2.1 x 5 mm, 130 Å, Waters, Milford, MA, USA) was installed connected to an ACQUITY Glycoprotein BEH Amide column (1.7 µm, 2.1 x 100 mm, 130 Å, Waters) on a Vanquish UHPLC binary pump (Thermo Fisher). The flow rate was set to 600 L / min. -1 and the column temperature at 35 °C. The injection volume was 5 L. The composition of the two mobile phases was: 10 mmol L'1 ammonium formate in H2O with 0.1% v / v formic acid (A) and 99.9% v / v acetonitrile with 0.1% v / v formic acid (B). The gradient elution program was performed as follows: 0.75 min with 95% v / v B; 0.75 to 5 min with a linear gradient of 95% v / v to 80% v / v B; 5 to 30 min with a linear gradient of 80% v / v to 50% v / v B followed by 2.5 min of column washing with a linear gradient of 50% v / v to 2% v / v B, including column re-equilibration with 95% v / v B. During column washing, the flow rate was set at 250 L / min -1The total run time was 32.5 min. MS detection was carried out in electrospray ionization (ESI) alternating positive and negative modes (polarity change) with the ion source voltage set at ±3.5 kV; the capillary temperature at 250 °C; the enveloping gas at 15 (arbitrary units); and the auxiliary gas at 10 (arbitrary units). Two inclusion lists were constructed (one per polarity) that include the mass-charge (m / z) ratios of the precursors of all HMO adducts reported for LH that are listed in the NIST milk oligosaccharide MS library (CA Remoroza, T. et al. Creating a mass spectral reference library for oligosaccharides in human milk. Analytical Chemistry, 90 (15) (2018), pp. 8977-8988; CA Remoroza, et al. Increasing the coverage of a mass spectral library of milk oligosaccharides using a hybrid-search-based bootstrapping method and milks from a wide variety of mammals.Analytical Chemistry, 92 (15) (2020), pp. 10316-10326). An exclusion list was also constructed, including m / z peaks detected in a blank injection. MS / MS spectra were acquired using high-energy dissociation beam (HCD) collision cell fragmentation, with a normalized collision energy (NCE) of 25%.

[0056] 1.4.4 LC-MS Data Processing

[0057] 2'-FL, 3-FL, 3'-SL, 6'-SL, LNT, LNFP I, LNDFH I, and DSLNT were quantified using Xcalibur Quan Browser version 4.1 software (ThermoFisher) using linear regression curves and relative responses (i.e., peak areas of compounds normalized to the peak area of ​​the IS). Based on the concentration of 2'-FL and LNFP I, the maternal phenotype (i.e., secretor or non-secretor) was established.

[0058] 1.5 Visualization, processing and modeling of spectra

[0059] Spectral visualization was performed using OPUS 8.5 software. Spectral processing and data analysis were carried out in MATLAB R2021a (MathWorks, Natick, MA, USA) and using PLS Toolbox 9.3 (Eigenvector Research Inc., Manson, WA, USA). A partial least squares discriminant analysis (PLS-DA) model was constructed with 60 LH samples by selecting a spectral region of interest (i.e., 925 to 585 cm⁻¹). -1 ) and using the second derivative and mean centering during spectral preprocessing. For cross-validation (CV), Venetian blinds with 10 data divisions and a sample thickness of 1 were selected. Then, a set of 25 LH samples was used as an external validation set to evaluate the prediction performance of the PLS-DA model. RESULTS

[0060] 1. FTIR spectra of HMO patterns

[0061] ATR-FTIR spectra of a panel of HMO patterns were measured, and the spectral range with characteristic bands was evaluated for particular HMOs known to be abundant in secretory mothers and scarcely present or absent in non-secretory mothers (i.e., 2'-fucosyl-lactose (2'-FL) and lacto-N-fucopentaose I (LNFP I)) (Figure 1). In the region between 925 cm -1 and 585 cm -1 , 2' FL and LNFP I presented two bands (i.e., 820 cm -1 and 764 cm -1 ) that could be distinguished from the bands presented in other measured HMOs.

[0062] 2. 60 LH samples from mothers with known secretory status as a calibration set. The ATR-FTIR spectra of 60 LH samples were measured using known quantities of 2'-FL and LNFP I, determined by UPLC-MS. Based on the LC-MS results, 34 LH samples were attributed to secretory mothers and 26 LH samples to non-secretory mothers. The mean spectra of the two sample groups are shown in Figure 2. As can be seen, the samples from secretory mothers showed a band at 820 cm⁻¹. -1 .

[0063] 3. Statistical model for determining the maternal secretory state phenotype A partial least squares discriminant analysis (PLS-DA) model was constructed with the 60 LH samples by selecting the spectral region of interest (i.e., 925 to 585 cm -1) and using the second derivative and mean centering during spectral preprocessing. For cross-validation (CV), “Venetian blinds” with 10 data divisions and a sample thickness of 1 were selected. Figure 3 presents the PLS-DA model score graph, where a clear separation between LH samples from secretory and non-secretory mothers can be observed.

[0064] The model details are presented in Table 1:

[0065] Table 1. Merit figures of the PLS-DA model.

[0066]

[0067]

[0068] (*) These values ​​depend on the prevalence of non-secretors.

[0069] Note: Cl, Confidence Interval

[0070] The model showed excellent sensitivity (i.e., identification of true positives) and specificity (i.e., identification of true negatives) for both groups with very low classification error rates (both calibration and cross-validation), demonstrating its high classification performance.

[0071] Overall, the PLS-DA model exhibits excellent performance metrics, indicating high predictive accuracy and reliability. To provide further insight into the robustness and significance of the constructed PLS-DA model, a permutation test with 100 iterations was performed, confirming the statistical significance (95% confidence level) of the classification model (see Table 2 and Figure 4).

[0072] In summary, the results demonstrate that the model is suitable for predicting the maternal secretory state phenotype based on the IR spectrum of LH.

[0073] Table 2. Permutation test results: Probability of insignificance of the model versus permuted samples for the 2-component model. p-values ​​< 0.05 indicate that the model is significant at the 95% confidence level.

[0074] < < <

[0075]

[0076] 4. External validation set

[0077] ATR-FTIR spectra were measured from 25 LH samples collected from mothers with a known maternal secretor status phenotype determined by UPLC-MS as described in section 3 (14 secretors and 11 non-secretors). These samples were not used during model development in any way and were therefore used as an external validation set to assess the predictive performance of the PLS-DA model. Table 3 and Figure 5 show details of the predictive performance.

[0078] Table 3. Merit figures for the prediction performance of the PLS-DA model for maternal secretory status.

[0079]

[0080] (*) These values ​​depend on the prevalence of non-secretory bacteria.

[0081] Note: Cl, Confidence Interval

[0082] In the external validation set, the model correctly classified 24 of the 25 analyzed samples (Figure 5), with only one classification error. The associated performance metrics (Table 3) show a sensitivity of 100% and a specificity greater than 90%, confirming the model's ability to discriminate between samples from secretory and non-secretory mothers, even when applied to independent data. The slight decrease in performance compared to the calibration phase is expected when the model is faced with entirely new samples. However, the classifier maintains high predictive stability and generalizability.

[0083] Overall, the PLS-DA model shows strong predictive performance with minimal overfitting, high accuracy, and good generalizability, making it a reliable tool for classification tasks within the given dataset.

Claims

CLAIMS 1. A method for determining the maternal secretory phenotype in a human milk (HM) sample by infrared (IR) spectroscopy, comprising: (a) provide a previously obtained LH sample; (b) subject the LH sample to IR spectroscopy analysis to generate an IR spectrum in the range of 4000 to 400 cm -1 , where the signals in this region are associated with the presence of specific HMOs; (c) comparing the IR spectrum of the LH sample obtained in b) with a discriminatory model, wherein the discriminatory model uses a set of IR spectra of LH samples with a known maternal secretory state phenotype recorded in the same spectral range; and (d) determine the maternal secretory phenotype of the LH sample as secretor (Se+) or non-secretor (Se-) based on the differences determined in c).

2. The method of claim 1 wherein the IR spectrum in b) is generated in the range of 925 to 585 cm -1 .

3. The method according to claim 1 or 2, wherein the specific HMOs are α1-2 fucosylated HMOs, at least one selected from 2'-fucosyl-lactose (2'-FL) and lacto-N-fucopentaose I (LNFP I) and lactodifucotetraose (LDFT).

4. The method according to any of the preceding claims wherein the analysis in step b) is carried out by attenuated total reflectance Fourier transform IR (ATR-FTIR) spectroscopy.

5. The method according to claim 1, wherein the discriminatory model is constructed by: i) applying the second derivative transform to the set of IR spectra of the LH samples to obtain the transformed data; and i) calculating the centering on the mean of the transformed data.

6. The method of claim 1, wherein the determination of the maternal secretory state phenotype is used to prevent, diagnose or monitor health conditions associated with the secretory phenotype selected from among gastrointestinal disorders, allergies, alterations in growth and / or development of the immune and neurological systems in infants.

Images