Inoculation of Mother’s Own Milk Could Personalize Microbiological, Metataxonomic and Peptidomic Pro le of Pasteurized Donor Human Milk Used for Feeding Preterm Infants


 Background

Human milk is a vehicle of bioactive compounds and beneficial bacteria which promote the establishment of a healthy gut microbiome of newborns, especially of preterm infants. Pasteurized donor human milk (PDHM) is the second-best option when preterm mother’s own milk is unavailable. Since pasteurization affect the quality of donor milk, the effect on bacterial growth, human milk microbiota diversity and proteolytic phenomena in PDHM inoculated with different preterm milk samples, was evaluated.
Methods

Ten preterm milk samples was used to perform inoculation of PDHM at 10% v/v. Microbiological, metataxonomic and peptidomic analysis, on preterm milk samples at the baseline (T0) and PDHM and inoculated milk (IM) samples at T0, after 2h (T1) and 4h (T2) of incubation at 37°C, were conducted.
Results

IM samples at T2 showed a Total Bacterial Count not significantly different (p > 0.01) compared to preterm milk samples. Lactic acid bacteria (LAB) level was restored in all IM at T2. After inoculation, metataxonomic analysis showed that Proteobacteria remained the predominant phylum while Firmicutes moved from 3% at T1 to 9.4% at T2 in IM samples. Peptidomic profile of IM resembled that of PDHM incubated for the same time in terms of number and type of peptides.
Conclusion

The study demonstrated that inoculation with fresh mother’s own milk could restore bacterial growth and personalize human milk microbiome in PDHM. This effect could be beneficial thanks to the presence of maternal probiotic bacteria which make PDHM more similar to mother’s own milk.


Abstract Background
Human milk is a vehicle of bioactive compounds and bene cial bacteria which promote the establishment of a healthy gut microbiome of newborns, especially of preterm infants. Pasteurized donor human milk (PDHM) is the second-best option when preterm mother's own milk is unavailable. Since pasteurization affect the quality of donor milk, the effect on bacterial growth, human milk microbiota diversity and proteolytic phenomena in PDHM inoculated with different preterm milk samples, was evaluated.

Methods
Ten preterm milk samples was used to perform inoculation of PDHM at 10% v/v. Microbiological, metataxonomic and peptidomic analysis, on preterm milk samples at the baseline (T0) and PDHM and inoculated milk (IM) samples at T0, after 2h (T1) and 4h (T2) of incubation at 37°C, were conducted.

Results
IM samples at T2 showed a Total Bacterial Count not signi cantly different (p > 0.01) compared to preterm milk samples. Lactic acid bacteria (LAB) level was restored in all IM at T2. After inoculation, metataxonomic analysis showed that Proteobacteria remained the predominant phylum while Firmicutes moved from 3% at T1 to 9.4% at T2 in IM samples. Peptidomic pro le of IM resembled that of PDHM incubated for the same time in terms of number and type of peptides.

Conclusion
The study demonstrated that inoculation with fresh mother's own milk could restore bacterial growth and personalize human milk microbiome in PDHM. This effect could be bene cial thanks to the presence of maternal probiotic bacteria which make PDHM more similar to mother's own milk.

Background
Human milk (HM) represents the optimal feeding for infants, not only for its nutritional quality, but also for being a rich source of several bioactive compounds and speci c peptides, essential for the development of the newborn's immature immune and digestive systems (1). HM hosts a complex community of bacteria, the HM microbiome, which comprises many viable commensals, mutualistic or potentially probiotic bacteria. Lots of these bacteria are indigenous, supported by the entero-mammary translocation, but maternal and environmental factors can contribute to shape HM microbiome composition, leading to a high inter-individual variability of microbial species in HM (2; 3). HM microbiome plays a bene cial role for the breastfed infants. The constant intake, during lactation, of bene cial bacteria present in the HM promotes the establishment of a healthy gut microbiome of newborns, which limits the growth of potential pathogens, by exerting a protective and nutritive role, especially for preterm infants (4; 5).
Mothers who delivered prematurely often experience signi cant di culties in breastfeeding their infants.
For this reason, the majority of mothers of very preterm infants are able to express small volumes of their own milk (6) and, when mother's own milk is not available or insu cient, donor HM is the second-best option (7).
As recommended by current international guidelines (8; 9), donor HM must be pasteurized, using the Holder method, to destroy potential pathogenic microorganisms and retain the ability to inhibit bacterial growth. Although it is important to guarantee microbiological safety, Holder pasteurization leads to a total or partial decrease in the biological quality and effectiveness of donor HM, inactivating most of the valuable HM microbiota species (10). Indeed, the probiotic effect of pasteurized donor HM (PDHM) is not comparable with the probiotic effect of mother's own milk (11). Moreover, HM microbial species hydrolyze milk proteins, due to the potential action of bacterial proteases/peptidases. Other proteolytic phenomena can occur upon the action of the endogenous proteases, some of these originating within the mammary gland and acting during the whole lactation period. Therefore, both HM endogenous enzymes and microbiome can release peptides or degrade speci c individual milk proteins with several functional properties (12).
Recently, it has been reported that incubating donor HM with preterm mother's own milk (preterm milk, PM) may be effective in improving the biological quality of donor HM by restoring the original milk microbiome (4). On these bases, an in vitro study was conducted to evaluate the effect on bacterial growth, HM microbiota diversity and proteolytic phenomena in PDHM inoculated with different PM samples.

Experimental design
Ten samples of inoculated PDHM (hereafter named as IM1 to IM10) were obtained by adding ten different fresh PM samples (hereafter named as PM1 to PM10) at 10% (v/v). Then, the IM samples were incubated at 37 °C for 4 h. The PDHM and IM samples were collected and analyzed at different time points: at the baseline (T0), and 2h (T1) and 4h (T2) after inoculation. The adopted experimental design is summarized in Fig. 1. Microbiological and metataxonomic analysis were performed on PDHM and PM samples at baseline and on PDHM and IM samples at T1 and T2. Peptidomic analysis was conducted on the same samples except for T1.

Subjects
Mothers who delivered prematurely between November 2018 and January 2019 at the Neonatal Intensive Care Unit (NICU) of the Fondazione IRCCS Ca' Granda Ospedale Maggiore Policlinico (Milan, Italy), were included in the study. Eligible criteria were: delivery before the 37 th week of gestational age and absence of maternal antibiotic therapy at the time of the milk collection. Basic maternal and infants' clinical characteristics were collected using the computed medical records.
Term-delivering mothers, included in this study, donated their milk to the HMB of the Fondazione IRCCS Ca' Granda Ospedale Maggiore Policlinico, according to HMB Italian guidelines (13). Written informed consent was obtained for each participating mother. The study was approved by Ethical Committee of the Fondazione IRCCS Ca' Granda Ospedale Maggiore Policlinico (Approval nr. 289_2017).

Preterm mother's own milk collection
Preterm-delivering mothers, prior to milk collection, had to perform an accurate hand washing using a hand sanitizer and a breast washing, using exclusively running water, in accordance with the NICU internal procedure. Each sample was collected, using an electric breast pump of breastfeeding room and a personal breast pump kit, into a sterile bottle. Breast pump kit had to be cleaned and sterilized before every collection.
Enrolled preterm-delivering mothers collected a sample of 50 mL of their own fresh milk between the 30 th and the 60 th day post-delivery. All PM samples were collected during the rst hours of the morning at the breastfeeding room of the NICU, immediately refrigerated and delivered to the laboratories for analyses. PM3 and PM5 samples were obtained from the same mother in different days.

Donor HM collection and pasteurization
Term delivering mothers, belonging to the HMB, collected their milk at home following the the same instructions of mothers who delivered preterm, in terms of personal hygiene and material disinfection. Each sample was collected using an electric breast pump and a personal breast pump kit, and placed into sterile bottle. Breast pump kit had to be cleaned and sterilized before every collection.
Donor HM was stored at donors' home in a refrigerator for maximum 24 h and thereafter frozen at -20° C. During the transport, donor HM was preserved by maintaining the cold chain until the arriving to the HMB.
Donor HM samples, used to perform inoculation, were collected by donors in different periods of lactation. These samples were thawed, pooled (total volume of 3 L) and then Holder pasteurized at 62.5°C for 30 min. PDHM was stored at -80° C until inoculation with fresh PM samples.

Inoculation of PDHM with PM
To perform the inoculation, frozen PDHM was gently thawed and 100 mL and inoculated with 10 mL of each fresh PM sample. The low milk volume expressed by preterm-delivering mothers, which necessarily must be used to breastfeed their babies, did not allow to consider higher inoculation level.
After inoculation, PDHM and IM samples were kept at 37 °C for 4 h (incubation time).
PDHM and PM were analyzed immediately after their arrival in the laboratory. PDHM and IM samples were analyzed after incubation at different time points too, as described in the experimental design ( Fig.  1).

Enumeration of microorganisms
Serial decimal dilutions of the inoculated aliquots in sterile quarter-strength Ringer's solution (Scharlab, Barcelona, Spain) were prepared, and the following microbiological determinations were carried out.  (19). At any sampling time, PDHM sample was also analysed for detection of mesophilic aerobic bacteria, Enterobacteriaceae, coliforms, E. coli, yeasts, Pseudomonas spp., coagulase-positive and negative staphylococci and B. cereus.

Search for Listeria monocytogenes and Pseudomonas aeruginosa
Listeria monocytogenes was searched in all IM and PDHM samples by SureFast® Listeria monocytogenes PLUS real-time PCR (RT-PCR) assay (R-Biopharm, Darmstadt, Germany) according to the manufacturer's instructions. The RT-PCR ampli cation reactions were performed on an Eco Real-Time PCR System (Illumina, San Diego, CA, USA). Detection of Pseudomonas aeruginosa was performed in samples where the presence of Pseudomonas spp. had been detected by microbiological cultivation method. The adopted protocol was as follows: ten colonies from PP agar plates were randomly picked and sub-cultured overnight in Brain Heart Infusion (BHI) broth (Scharlab) at 30 °C. After growth, the DNA was extracted by the MicroLYSIS kit (Clent Life Science, Stourbridge, UK). The identi cation of isolates was carried out using P. aeruginosa speci c primer as previously reported in Cremonesi et al. (20).

Search for Staphylococcus aureus virulence and enterotoxin genes
Staphylococcus aureus virulence and enterotoxin genes were explored in samples found to be contaminated with coagulase-positive staphylococci. The DNA was extracted from 1 mL of breast milk as previously described by Cremonesi et al. (21). The extracted DNA was ampli ed by a multiplex PCR for the detection of genes encoding for the coagulase (coa) and thermonuclease (nuc) regions and for the main staphylococcal enterotoxins (sea, sec, sed, seg, seh, sei, sej and sel) according to Cremonesi et al. (22).

Metataxonomic analysis
Five mL of milk sample were centrifuged at 500 g for 10 min at 4 °C; the supernatant was discarded, and the pellet was washed with one mL of saline solution (0.9% NaCl) and centrifuged at 500 g for 5 min at 4°C . The supernatant was discarded, and the bacterial DNA was extracted from the samples as described previously (23), by using a method based on the combination of a chaotropic agent, guanidium thiocyanate, with silica particles, to obtain bacterial cell lysis and nuclease inactivation. DNA quality and quantity were assessed using a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA). The isolated DNA was then stored at -20 ˚C until use.

Metagenomic library preparation and sequencing
Bacterial DNA was ampli ed using the primers described by Caporaso et al. (24), which targeted the V3-V4 hypervariable regions of the 16S rRNA gene. All PCR ampli cations were performed in 25 μL volumes per sample. A total of 12.5 μL of Phusion High-Fidelity Master Mix 2× (Thermo Fisher Scienti c, Waltham, MA, USA) and 0.2 μL of each primer (100 μM) were added to 2 μL of genomic DNA (5 ng/μL). Ampli cation was performed in an Applied Biosystem 2700 thermal cycler (Thermo Fisher Scienti c) using the ampli cation cycle as follows: samples were denatured at 98 ˚C for 30 s, followed by 25 cycles with a denaturing step at 98 ˚C for 30 s, annealing at 56 ˚C for 1 min and extension at 72˚C for 1 min, and a nal extension at 72˚C for 7 min. Amplicons were cleaned with Agencourt AMPure XP kit (Beckman Coulter, Brea, CA, USA) and libraries were prepared following the 16S Metagenomic Sequencing Library Preparation Protocol (Illumina, San Diego, CA, USA). The libraries obtained were quanti ed by RT-PCR with KAPA Library Quanti cation Kit (KapaBiosystems, Cape Town, South Africa), pooled in equimolar proportion and sequenced during a single MiSeq (Illumina) run with 2×250-base paired-end reads.

Bioinformatics and Statistical analysis
Demultiplexed paired-end reads from 16S rRNA-gene sequencing were rst checked for quality using FastQC (25) for an initial assessment. Forward and reverse paired-end reads were joined into single reads using the C++ program SeqPrep (26). After joining, reads were ltered for quality based on: i) maximum three consecutive low-quality base calls (Phred < 19) allowed; ii) fraction of consecutive high-quality base calls (Phred > 19) in a read over total read length ≥ 0.75; iii) no "N"-labeled bases (missing/uncalled) allowed. Reads that did not match all the above criteria were ltered out. All remaining reads were combined in a single FASTA le for the identi cation and quanti cation of OTUs (operational taxonomic units). Reads were aligned against the SILVA closed reference sequence collection release 123, with 97% cluster identity (27; 28) applying the CD-HIT clustering algorithm (29). A pre-de ned taxonomy map of reference sequences to taxonomies was then used for taxonomic identi cation along the main taxa ranks down to the genus level (domain, phylum, class, order, family, genus). By counting the abundance of each OTU, the OTU table was created and then grouped at each phylogenetic level. OTUs with total counts lower than 10 in fewer than 2 samples were ltered out. All of the above steps, except the FastQC reads quality check, were performed with the QIIME open-source bioinformatics pipeline for microbiome analysis (30).
The milk microbial diversity was assessed within-(alpha diversity) and across-(beta diversity) samples. All indices (alpha and beta diversity) were estimated from the complete OTU table (at the OTU level), ltered for OTUs with more than 10 total counts distributed in at least two samples. Besides the number of observed OTUs directly counted from the OTU table, within-sample microbial richness and diversity were estimated using the following indices: Chao1 and ACE (Abundance-based coverage Estimator) for richness, Shannon, Simpson and Fisher's alpha for diversity (31; 32; 33; 34; 35; 36), Simpson E and Pielou's J (Shannon's evenness) for evenness (37). The across-sample milk microbiota diversity was quanti ed by calculating Bray-Curtis dissimilarities (38). Prior to the calculation of the Bray-Curtis dissimilarities, OTU counts were normalized for uneven sequencing depth by cumulative sum scaling CSS, (39). Among groups (PDHM, PM, IM) and pairwise Bray-Curtis dissimilarities at different timepoints were evaluated non-parametrically using the permutational analysis of variance approach (999 permutations; (40). Details on the calculation of the mentioned alpha-and beta-diversity indices can be found in Biscarini et al. (41).
Descriptive data related to clinical demographic data of preterm/term -delivering mothers and respective newborns, involved in this study, were reported as mean and standard deviation. Bacterial counts were expressed as mean and standard deviation and the signi cance level was set at 0.01. Metataxonomic data were reported as mean and relative abundance (%) and the signi cance level was set at 0.05.

Peptidomic profiling by UPLC/HR-MS/MS
Peptidomic analyses were performed using an Acquity UPLC module (Waters, Milford, MA, USA) coupled to a Q Exactive hybrid quadrupole-Orbitrap mass spectrometer (Thermo Fisher Scientific, San Jose, CA, USA) and the peptides were identified using the Proteome Discoverer v1.4 software (Thermo Fisher Scientific). The MS data were processed, and the peptides were identified using the Proteome Discoverer v1.4 software (Thermo Fisher Scientific). Automatic peak detection was performed setting signal-to-noise ratio to 4 as suggested by Mangé et al. (42). The sequences of peptides were identified from MS/MS spectra using SequestHT algorithm (43) against a HM protein library constructed based on results from previous studies (42; 44; 45; 46; 47). A non-specific enzyme cleavage pattern was defined, and 12 missed cleavage sites (maximum allowed for the algorithm) were allowed. No static modifications were set.
Phosphorylation of serine and threonine, deamidation of asparagine, glutamine and arginine, oxidation of methionine and cyclisation of an N-terminal glutamine to pyro-glutamic acid were selected as dynamic modifications. Mass error tolerance for precursor ions was 5 ppm and for fragment ions was 0.02 Da. A strict false discovery rate of peptide identification was set (FDR = 0.01).

Results
Nine preterm-delivering mothers were enrolled in the study and ten samples of fresh PM were collected (PM3 and PM5 were from the same mother) from November 2018 to January 2019. Clinical demographic data regarding mothers and respective newborns were reported in Table 1. The pathogenic bacteria B. cereus, E. coli and L. monocytogenes were not detected. P. aeruginosa was revealed only in IM3, IM5 and IM7 samples at T2 (Table 2). In the same samples, S. aureus was found at all sampling times (data not shown), achieving at the end of incubation values comparable to those detected in the corresponding PM samples. Moreover, S. aureus strains harboring toxin genes (sea and seh) were detected in PM3, PM5 and IM3, IM5 samples. Coliforms and Enterobacteriaceae were present at similar level (about 2 log 10 CFU/mL) in IM3 and IM5 samples, and the content of these microorganisms was about 1.5 log higher than in the corresponding PM samples ( Table 2).
As for LAB, Lactobacillus spp. and Streptococcus spp. were recovered in many of the PM samples, while enterococci were found in PM4 and PM10. At T2, the corresponding IM samples showed LAB levels not signi cantly different (p > 0.01). The applied experimental conditions did not promote the growth of Propionibacteria, since their counts in the IM samples were generally 1 log lower than in the PM samples at T3. A similar trend was observed for Bi dobacterium and yeasts. Sequence-based rarefaction curves were obtained from the QIIME pipeline; the sample-based rarefaction curve was produced with ad hoc R functions. The observed number of OTUs detected was plotted as a function of the number of reads (up to 75,000) in each sample and of the number of samples (see Additional Figure 1). Both curves tend to plateau asymptotically suggesting that sampling and depth of coverage in this analysis were enough to describe the biological diversity of the maternal milk microbiome, although a few samples have been under-sequenced.
Results from 16S rRNA-gene sequencing from all samples have been used to characterize the core microbiome in PDHM, PM and IM samples. OTUs were grouped taxonomically at the phylum level. In terms of relative abundances, most of the reads for PDHM and IM samples belonged to the Proteobacteria and Firmicutes phyla (Fig. 2), which accounted for almost 100% of the entire microbiota at all time points (  % at T1 to 9.4% at T2 (Table 3). A complete list of the bacterial groups at phylum, family and genus level as well as their relative abundances are reported in Additional Table 1.
These data were further investigated in terms of the relative abundance in bacterial distribution, analyzing each PM sample and IM at T1 and T2 (Fig. 3). Differences in microbial composition between PDHM and PM samples found in the core microbiome analysis are shown in Fig. 4.
At the family level, the microbiota of the PM samples was characterized by Staphylococcaceae, with the prevalence of the genus Staphylococcus, and Bacillales with the prevalence of the genus Bacillus. Lactobacillales family members were observed in few PM samples, with low relative abundance. In PDHM sample, the class of Gammaproteobacteria and the Enterobacteriaceae family were the most predominant with Enterobacterales, Enterobacter and Escherichia-Shigella genus.
Variation of different alpha-diversity and beta-diversity indices at different time points of incubation.
The alpha-diversity indices were all signi cantly moving over time (  Beta-diversity analysis was evaluated with (Fig. 6A) and without (Fig. 6B) the PDHM sample. From PERMANOVA (999 permutations), the analysis showed a statistically signi cant (p=0.001) separation between timepoints and hence between PM and IM samples, revealing major differences in the principal constituents of the microbial community (Fig. 6B). PDHM presented a different microbial community as compared to the PM (Fig. 6A) samples, and, being the substrate for the inoculum, its microbial composition was predominant within 4 hours of incubation.
Peptidomic analysis of PDHM, PM and IM samples.
A qualitative peptidomic analysis was conducted on PDHM, PM and IM samples at different time points.
Since PM3 and PM5 were collected from the same mother and their peptidomic pro les overlapped, only the PM5 was considered for the discussion. In PM samples, about 800 unique peptides were identi ed.
Each PM contained a different number of peptides, higher (PM1, PM2, PM4, PM7, PM10) or lower (PM5, PM6, PM8, PM9) than PDHM. As shown in Fig. 7, PM7 and PM10 appeared as the most proteolyzed samples, whereas the lowest extent of proteolysis characterized PM 5, 8 and 9. About 60% of total peptides identi ed in PDHM and PM at T0 were of a molecular weight range between 1 and 3 kDa, , while those smaller than 1kDa represented only 3% of total peptides (Fig. 7A). Human β-casein was the most representative protein with 62% of total peptides, followed by α S1 -casein (13%), osteopontin (10%) and polymeric immunoglobulin receptor (6%) (Fig. 7B). Peptides belonging to α-lactalbumin, one of the most abundant proteins in HM, were not found. At T2, the number and type of peptides in PDHM sample roughly overlapped those revealed at T0. As a matter of fact, incubation for 4h at 37 °C did not alter the peptidomic pro les of PDHM sample. The peptidomic pro le of IM samples, except for IM5 and IM6, resembled that of PDHM incubated for the same time ( Fig. 7C and Fig. 7D). These results re ected in very similar patterns the precursor proteins for the peptide identi ed in IM samples at the end of incubation (T2) (Fig. 7D). The dominance of β-casein derived peptides (74% of total peptides identi ed at T2) was con rmed in all IM samples. Peptides in the molecular weight range of 1-3kDa were still the most abundant (Fig. 7C).

Discussion
In this study, the effect of PM inoculation in PDHM on bacterial growth, HM microbiome diversity and proteolytic phenomena was evaluated. Several studies (48; 49) showed that HM is not sterile and contains a variety of mother-speci c probiotic and commensal bacteria, which constitute HM microbiome, a source of potential bene cial bacteria for the colonization of preterm infants' gut. The present study is one of the few investigations on the microbiome and peptidome of PDHM, PM and of the derived IM samples, and the rst one dealing with the occurrence and potential growth of pathogenic bacteria in PM and related IM incubated samples.
The origin of bacteria in HM depends on different factors. Currently it is possible to distinguish between a "maternal microbiota" and an "exogenous microbiota", as supported by Moossavi & Azad (50). Beyond the transfer of microbes from breast skin ora to the milk, other pathways are considered relevant to shape a speci c HM microbiome among which contamination related to milk handling.
A recent review of Fernández & Rodríguez con rmed that a bacterial load of HM from healthy women, collected hygienically, usually was about 3 log 10 CFU/mL while a bacterial load up to 6 log 10 CFU/mL was characteristic of mastitis suffering women or milk after breast pumping (51). Despite the adherence to hygienic condition in the Neonatal Intensive Care Unit, the bacterial load of our PM samples ranged from 3 to 5 log 10 CFU /mL that didn't exclude an external contamination of expressed HM during breast milk collection. Our results show that the inoculation of PDHM with PM (10%, v/v) allowed the restoration of a bacterial load samples similar to that of PM samples after 4 hours of incubation at 37 °C, although an high variability of bacterial growth between PM samples and then in IM samples was detected.
Anyway, all microorganisms grown in PM samples were found in the respective IM samples at T2.
Speci cally, in this study, cultivable bacteria as LAB or belonging to Staphylococcaceae, Streptococcaceae, Enterobacteriaceae and Pseudomonaceae and other gram-positive bacteria (Propionibacterium and Bi dobacterium) were detected in PM samples and in their related IM samples.
Most of the species belonging to the genera Lactobacillus and Bi dobacterium, commonly considered probiotic bacteria, can confer health bene ts to HM-fed premature infants. LAB were found in all PM samples, were not found in PDHM but reached the same load in all IM samples at the end of incubation period. Some species of lactobacilli, commonly present in human breast milk, colonize the neonatal gut and contribute to protection against infant infections, suggesting that lactobacilli could potentially wellmodulate immunity (52). In gastrointestinal tract, HM lactobacilli strains are metabolically active in the infant gut and increase the production of functional metabolites such as butyrate, which is the main energy source for colonocytes and a relevant compound in the modulation of intestinal function (53).
Concerning Bi dobacteria, PM contains lower levels compared to milk of term-delivering mothers (54). Moreover, indirect breastfeeding could affect the concentration of Bi dobacteria in HM, which can be further lowered by the need to use breast-pump (2). In addition, Bi dobacteria grow only in strict anaerobic condition. This could explain our di cult to detect Bi dobacteria in PM and IM samples using traditional plate count method, as con rmed also by Boix-Amorós et al. (55).
The presence of microbial families belonged to Staphylococcaceae, Streptococcaceae, Enterobacteriaceae and Pseudomonaceae could be consistent with contamination during breast milk collection, even if hygienic protocols are used (2; 56; 57).
Given that pasteurization of mother's own milk is not recommended (58), also potential pathogenic microorganisms are normally conveyed to preterm infants fed with fresh HM. Nevertheless, breastfeeding is recommended because the potential bene ts of breast milk outweigh the possible negative effects: only few articles reported that neonatal infection was caused by the ingestion of contaminated HM (59).
The feeding of HM to the newborn should be avoided or stopped only in speci c circumstances (60) and the implementation of the hygienic measures during milk collection, both in the Neonatal Intensive Care Unit and at home, could avoid proliferation of undesirable bacteria in expressed HM.
Holder pasteurization is necessary to neutralize pathogens potentially present in donor HM but unfortunately it reduce initial microbial load, destroying bene cial bacteria, as LAB and bi dobacteria (61). In our study, metataxonomic analysis underlined that the addition of a small amount of fresh HM can enhance the microbiome of PDHM, making it more similar to HM microbiome.
In accordance with other studies (3; 62), the microbiome of our PM samples was mostly characterized by the presence of Staphylococcaceae, with the prevalence of the Staphylococcus genus, and Bacillales with the prevalence of the Bacillus genus. The large number of Firmicutes in these samples could be due to the presence on areolar skin of bacteria belonging to this phylum.
Conversely, microbiome of our PDHM samples was mostly characterized by Proteobacteria, the dominant phylum in HM as also con rmed by Pannaraj et al. (62).
After inoculation and incubation, the microbiome of IM became richer of Firmicutes, predominant phylum of our PM samples. Therefore, Staphylococcaceae and Bacillales, found in IM samples at T2, make PDHM more similar to mother's own milk.
In addition, an increment in microbial diversity in IM samples four hours after the milk incubation was detected.
Fehr et al. demonstrated that the microbiota of milk and infants gut of mother-infants dyads are dissimilar but related (63). Milk and infants' gut microbiota become slightly more similar over time, sharing speci c microbial strains, in association with breastfeeding mode and duration.
Preterm infants present a marked vulnerability to dysbiosis with a change in abundance, diversity, and progressive acquisition of the intestinal microbiota. Disrupted maturation of the intestinal microbiota occurred in extremely preterm infants with postnatal growth failure (64).
In the last years, metagenomics analysis underlined that HM microbiome modulate preterm infants' gut microbiota but, to the best of our knowledge, studies on the impact of different nutritional approaches on the gut microbiota of preterm infants are very limited.
Some studies (5; 65; 66; 67) suggest that donor HM favors an intestinal microbial pro le more similar to mother's own milk. On the other hands, the diversity of gut microbiota increased over time and was constantly higher in infants fed with mother's own milk relative to infants who was fed with donor or formula milk.
Ford et al. demonstrated that preterm infants fed with mother's own milk developed an increased gut microbial community compared to preterm infants fed with PDHM. In addition, infants fed mother's own milk had a better weigh gain and an improved feeding tolerance than donor HM-fed infants (66).
The inoculation strategy could personalize PDHM which could shape infants' gut microbiota increasing microbial diversity and promoting healthier short and long-term outcomes in preterm infants. In addition, increasing volume of own mother milk for inoculation could be more useful. Each PM sample presented different degree of proteolysis that re ected in characteristic peptidomic pro les. Despite this inter-variability, negligible differences in proteolysis level and peptidomic pro le among IM samples were detected over time. Overall, these features mostly re ected those of PDHM sample after the same time of incubation, thus underlying a minor additional exogenous proteolysis. The largest number of peptides in PDHM and PM samples derived from β-casein and, therefore, the same applied for IM samples. Different authors demonstrated the prevalence of β-casein derived peptides in HM (72; 73; 74). Scarce presence of peptides derived from major whey protein, in particular from αlactalbumin, characterized the studied samples, as reported for HM by Gan et al. (75).

Conclusions
The present research study revealed an inter-individual microbiological variability in fresh PM samples thus underlying that each preterm mother presents a unique ngerprint-like microbiota, which can be potentially delivered to own infant. The inter-individual variability of PM samples also re ected in their peptidomic pro les. In addition, the study demonstrated that inoculation with fresh PM partially restore and personalize HM microbiome of PDHM. This effect could be bene cial thanks to the presence of maternal probiotic bacteria which make PDHM more similar to mother's own milk. Nevertheless, the adoption and respect of strict hygienic measures during collection and storage of PM is paramount to avoid contamination of inoculated milk. Overall, the feasibility of inoculating PDHM with fresh PM for preterm infant feeding needs further investigation. In this view, the enlargement of the sample size of PM samples could reinforce the reliability of the results reducing variability and bias.

Declarations
Authors' contributions DM contributed to the study concept, data interpretation and wrote the manuscript. CT contributed to the discussion of the results and the revision of the article. PP conceptualized and designed the protocol and reviewed the manuscript. SM, TS, FB, PC, BC, VP and MS carried out the analyses, collected data and wrote the manuscript. ID contributed to the study concept, data interpretation and critically reviewed the manuscript. OA and NL contributed to the discussion of the results and critically reviewed the manuscript.
FM provided suggestions with regard to the content and concept of the article. PR conceptualized and designed the protocol and critically reviewed the manuscript. All authors red and approved the nal manuscript.

Figure 1
Experimental design of the study. Heatmap of the within-phylum relative abundance of microbial taxa identi ed in PDHM, PM and IM samples at three time points.

Figure 5
Alpha-diversity analyzed using different indices per time points.

Figure 6
First two dimensions from the (non-metric) multi-dimensional scaling of the Bray-Curtis dissimilarity matrix.