Terminology

There is currently a lack of standardisation for the technique involving FACS or MACS sorting of bacteria into Ig-bound or unbound fractions, followed by sequencing of fractions. While various terms have been given to describe this, or merely describing it as sorting followed by sequencing of fractions, we propose the use of the term “Ig-Seq” for consistency. We would like to caveat that Ig-Seq is conceptually different from the technique mFLOW (microbiota flow cytometry) found in the literature and the use of the terms should be distinguished (detailed in Supplementary Text).

Cell sorting and controls

A variety of nucleic acid dyes were used to stain bacteria. Some considerations would be the ability to stain both gram-positive and gram-negative bacteria, dye permeability, and lack of spectral overlap with the fluorescent dye used in the secondary antibody.

Choosing of the antibody isotype and type of sort is dependent on the research study. Most studies employed a two-way sort for IgA, as sIgA is the dominant antibody in both stool and breastmilk and most commonly associated with commensals (Pabst and Slack, Reference Pabst and Slack2020). Other studies looked at the difference or overlap in binding across Ig isotypes using a four-way sort. The results will be discussed below. IgA is the most important isotype to consider in Ig-Seq involving stool and breastmilk.

Cell sorting has been performed using both MACS and FACS. FACS will give fractions with greater purity than MACS, whereas MACS will focus on enrichment rather than pure fractions of bacteria bound and unbound to Ig (Figure 2). This was demonstrated by Qi et al. (Reference Qi, Tu, Zhou, Tu, Chang, Chen, Hu, Yu and Sun2022), where only the Ig+ fraction was collected, but not the Ig− fraction. However, some papers using MACS were able to assess for purity using flow cytometry or performed deconvolution to control for contamination in the sort (Gopalakrishna et al., Reference Gopalakrishna, Macadangdang, Rogers, Tometich, Firek, Baker, Ji, Burr, Ma, Good, Morowitz and Hand2019; Sánchez-Salguero et al., Reference Sánchez-Salguero, Corona-Cervantes, Guzmán-Aquino, de la Borbolla-Cruz, Contreras-Vargas, Piña-Escobedo, García-Mena and Santos-Argumedo2021). These processes can be considered if intending to use MACS rather than FACS due to a lack of access to equipment or cost constraints. Some controls used during FACS include collecting sheath fluid from the fluidics pre and post sorting (Kau et al., Reference Kau, Planer, Liu, Rao, Yatsunenko, Trehan, Manary, Liu, Stappenbeck, Maleta, Ashorn, Dewey, Houpt, Hsieh and Gordon2015; Planer et al., Reference Planer, Peng, Kau, Blanton, Ndao, Tarr, Warner and Gordon2016), and ensuring that backflush was performed between samples to maintain baseline background (Janzon et al., Reference Janzon, Goodrich, Koren, Waters and Ley2019). One study mentioned autoclaving of sheath fluid (Kau et al., Reference Kau, Planer, Liu, Rao, Yatsunenko, Trehan, Manary, Liu, Stappenbeck, Maleta, Ashorn, Dewey, Houpt, Hsieh and Gordon2015). However, autoclaving will introduce molecular contamination (Suyama and Kawaharasaki, Reference Suyama and Kawaharasaki2013) while removing microbial contamination, and controls for molecular contamination should be used in conjunction to allow for filtering of contaminant sequences. Some studies also sequenced the input fraction that was processed for sorting but not put through the cell sorter as a baseline comparison post-sorting (Kau et al., Reference Kau, Planer, Liu, Rao, Yatsunenko, Trehan, Manary, Liu, Stappenbeck, Maleta, Ashorn, Dewey, Houpt, Hsieh and Gordon2015; Planer et al., Reference Planer, Peng, Kau, Blanton, Ndao, Tarr, Warner and Gordon2016). This could also serve as a control for sequence contamination from the sorting process. Recently, a new method using 96-well plate-based MACS sorting (next-generation IgA-SEQ) has been proposed, as opposed to column-based MACS sorting (van Gogh et al., Reference van Gogh, Louwers, Celli, Gräve, Viveen, Bosch, de Boer, Verheijden, Suijkerbuijk, Brand, Top, Oldenburg and de Zoete2024). Briefly, the authors suggest that this improves yield for downstream metagenomics to be possible, maintains anaerobic conditions, and is less time-consuming compared to FACS (van Gogh et al., Reference van Gogh, Louwers, Celli, Gräve, Viveen, Bosch, de Boer, Verheijden, Suijkerbuijk, Brand, Top, Oldenburg and de Zoete2024). The authors also do not suggest column-based sorting in high microbiota abundance sample types due to problems in enrichment, and results may not accurately reflect real sample data (van Gogh et al., Reference van Gogh, Louwers, Celli, Gräve, Viveen, Bosch, de Boer, Verheijden, Suijkerbuijk, Brand, Top, Oldenburg and de Zoete2024). Because of previous findings that species-level IgA-coating between FACS and plate-based MACS sorting was only consistent in 60% of cases (van Gogh et al., Reference van Gogh, Louwers, Celli, Gräve, Viveen, Bosch, de Boer, Verheijden, Suijkerbuijk, Brand, Top, Oldenburg and de Zoete2024), interpretation of results in future studies will need to consider differences in methodology used. Comparison of Ig-unbound bacteria between plate-based MACS and FACS was not mentioned either (van Gogh et al., Reference van Gogh, Louwers, Celli, Gräve, Viveen, Bosch, de Boer, Verheijden, Suijkerbuijk, Brand, Top, Oldenburg and de Zoete2024), and this may need to be explored if using the Ig probability ratio discussed below.

Sequencing, controls, and post-processing

The majority of studies used extraction kits for DNA extraction, including bead-beating for cell lysis. Bead-beating is able to improve DNA yield and affects genus characterisation, and should be incorporated in the extraction (Zhang et al., Reference Zhang, Brock, Arana, Dende, van Oers, Hooper and Raj2021). We recommend using extraction kits over manual phenol–chloroform extraction for greater consistency in results and in the consistency in purity of DNA extracted based on A260/280 absorbance (Douglas et al., Reference Douglas, Ivey, Papanicolas, Best, Muhlhausler and Rogers2020), and to also be consistent in the extraction procedure if also extracting overall sample material for general microbiome analysis (non-Ig-Seq) to allow for comparison. Controls, including extraction blanks, PCR negative controls, positive controls, and use of master mixes should also be detailed as per the Strengthening The Organization and Reporting of Microbiome Studies (STORMS) checklist, in line with current microbiome guidelines (Mirzayi et al., Reference Mirzayi, Renson, Furlanello, Sansone, Zohra, Elsafoury, Geistlinger, Kasselman, Eckenrode, van de Wijgert, Loughman, Marques, MacIntyre, Arumugam, Azhar, Beghini, Bergstrom, Bhatt and Bisanz2021). The majority of studies used Illumina sequencing, and only one study also included metagenomic sequencing. However, the number of samples that met the metagenomic sequencing standard was low (3 out of 19 mother–infant pairs). The majority of studies also used the regions within the V3–V5 area of the 16S rRNA gene (the most common V4), except for one study. Differences in composition based on the sequencing of different hypervariable regions of the 16S rRNA gene have previously been documented (Rintala et al., Reference Rintala, Pietilä, Munukka, Eerola, Pursiheimo, Laiho, Pekkala and Huovinen2017; Lopez Leyva et al., Reference Lopez Leyva, Brereton and Koski2021). The majority of studies used OTUs and the 2013 Greengenes database, however we recommend the use of denoising algorithms such as DADA2 or Deblur to generate ASVs or sOTUs, respectively with single nucleotide resolution and the use of updated databases. This has been discussed in detail elsewhere (Nearing et al., Reference Nearing, Comeau and Langille2021). Strict filtering of reads post-classification and decontamination should also be performed to remove potential contaminants from further analysis (McKnight et al., Reference McKnight, Huerlimann, Bower, Schwarzkopf, Alford and Zenger2019; Reitmeier et al., Reference Reitmeier, Hitch, Treichel, Fikas, Hausmann, Ramer-Tait, Neuhaus, Berry, Haller, Lagkouvardos and Clavel2021).

Ig index and other methods of analysis

Various methods were used to analyse Ig-Seq data. While some studies only reported relative abundance, others used some form of differential abundance analysis to compare differences between groups or fractions. A few Ig indices have also been proposed in analysing the differences between the composition of Ig-bound and unbound taxa. An IgA index was used in Dzidic et al. (Reference Dzidic, Abrahamsson, Artacho, Bjorksten, Collado, Mira and Jenmalm2017) and Dzidic et al. (Reference Dzidic, Mira, Artacho, Abrahamsson, Jenmalm and Collado2020) (first used in Bunker et al., Reference Bunker, Flynn, Koval, Shaw, Meisel, McDonald, Ishizuka, Dent, Wilson, Jabri, Antonopoulos and Bendelac2015), with the log of the ratio of the abundance of a taxon between the IgA+ and − fractions being measured (log (IgA+/IgA−)). This is a modification of the “Palm index” (Jackson et al., Reference Jackson, Pearson, Ilott, Huus, Hegazy, Webber, Finlay, Macpherson, Powrie and Lam2021) used in Palm et al. (Reference Palm, Zoete, Marcel, Cullen, Barry, Stefanowski, Hao, Degnan, Hu, Peter, Zhang, Ruggiero, Cho, Goodman and Flavell2014), which did not use a logarithmic transformation of the ratio of relative abundance values between the IgA+ and IgA− fractions for each taxon (IgA+/IgA−). A different index was developed by Kau et al. (Reference Kau, Planer, Liu, Rao, Yatsunenko, Trehan, Manary, Liu, Stappenbeck, Maleta, Ashorn, Dewey, Houpt, Hsieh and Gordon2015) as a measure of overall binding of a taxon and is based on relative abundance of IgA+ and IgA− fractions ( $ -\frac{\log \left({IgA}^{+}+c\right)-\log \left({IgA}^{-}+c\right)}{\log \left({IgA}^{+}+c\right)+\log \left({IgA}^{-}+c\right)} $ , where c is a pseudo-count). Termed the “Kau index” by Jackson et al. (Reference Jackson, Pearson, Ilott, Huus, Hegazy, Webber, Finlay, Macpherson, Powrie and Lam2021), this IgA index is an improvement of the Palm index, by using a logarithmic transformation and taking into consideration the difference of the IgA+ and − abundances relative to the total (Kau et al., Reference Kau, Planer, Liu, Rao, Yatsunenko, Trehan, Manary, Liu, Stappenbeck, Maleta, Ashorn, Dewey, Houpt, Hsieh and Gordon2015). This also provides directionality in the interpretation of the Ig index, with positive scores indicating greater abundance of a taxon in the Ig+ fraction and negative scores indicating greater abundance in the Ig− fraction.

Two different measures were proposed by Jackson et al. (Reference Jackson, Pearson, Ilott, Huus, Hegazy, Webber, Finlay, Macpherson, Powrie and Lam2021) that are available in the “IgAScores” R package (https://github.com/microbialman/IgAScores), to account for the total reads in the pre-sort fraction. The IgA+ probability is the likelihood that enough of a taxon is bound by IgA such that it is able to appear in the IgA+ fraction ( $ {Prob}_{ij}^{IgA+}=\frac{IgA_{ij}^{+}{FracSize}_j^{IgA+}}{PreSort_{ij}} $ or $ \frac{IgA_{ij}^{+}{FracSize}_j^{IgA+}}{\max \{{PreSort}_{ij},{IgA}_{ij}^{+}{FracSize}_j^{IgA+}\}} $ ), where i is a taxon in sample j, FracSize is the fraction of IgA+ sorted bacteria, and PreSort is the pre-sort taxon abundance) (Jackson et al., Reference Jackson, Pearson, Ilott, Huus, Hegazy, Webber, Finlay, Macpherson, Powrie and Lam2021). It requires the pre-sort abundances and the percentage of IgA-bound sorted bacteria in the FACS run (Jackson et al., Reference Jackson, Pearson, Ilott, Huus, Hegazy, Webber, Finlay, Macpherson, Powrie and Lam2021). The advantage is that this measure can be used if the aim is only to look at bacteria with high proportions of Ig binding, and sequencing of the Ig− fraction is not required (Jackson et al., Reference Jackson, Pearson, Ilott, Huus, Hegazy, Webber, Finlay, Macpherson, Powrie and Lam2021). The disadvantage is that it requires FACS to be used for sorting or for flow cytometric analysis post-MACS to determine the percentage of sorted bacteria that are Ig+. It is also affected by purity of the sorted fractions, especially in the case of column-based MACS, where lower purity of sorted fractions is seen (Jackson et al., Reference Jackson, Pearson, Ilott, Huus, Hegazy, Webber, Finlay, Macpherson, Powrie and Lam2021). The second measure is the IgA probability ratio, which is the ratio of the IgA+ probability to the IgA− probability ( $ {ProbRatio}_{ij}=\mathit{\log}2\left(\frac{\left({IgA}_{ij}^{+}{FracSize}_j^{IgA+}\right)+c}{\left({IgA}_{ij}^{-}{FracSize}_j^{IgA-}\right)+c}\right)/\mathit{\log}2\left(\frac{1+c}{c}\right) $ ), where the ratio of the IgA+ and IgA− probabilities is divided by a scaler such that ProbRatio values range from −1 to 1) (Jackson et al., Reference Jackson, Pearson, Ilott, Huus, Hegazy, Webber, Finlay, Macpherson, Powrie and Lam2021). This can distinguish taxa with lower relative Ig binding and is more sensitive than the IgA+ probability alone. An additional advantage is that sequencing of the pre-sort fraction is not required, while pre-sort differences are accounted for (Jackson et al., Reference Jackson, Pearson, Ilott, Huus, Hegazy, Webber, Finlay, Macpherson, Powrie and Lam2021). However, sequencing of the Ig− fraction is required to determine taxon abundance in the Ig− fraction. We will refer to these measures as the Ig+ probability and Ig probability ratio in this paper for general application to other Ig isotypes. These measures are advantageous over other previous Ig indices by accounting for pre-sort differences that may affect significance of group comparisons, and having a significantly lower coefficient of variation with more consistent results (Jackson et al., Reference Jackson, Pearson, Ilott, Huus, Hegazy, Webber, Finlay, Macpherson, Powrie and Lam2021). Between the two measures, the probability ratio was more robust to pre-sort differences than the Ig+ probability and had a significantly lower coefficient of variation (Jackson et al., Reference Jackson, Pearson, Ilott, Huus, Hegazy, Webber, Finlay, Macpherson, Powrie and Lam2021).

While papers have used relative abundance or differential abundance analysis of each taxon, the Ig probability ratio or the Ig+ probability are promising approaches for consideration in analysis of Ig-Seq data. It has been reported that the same taxon can be both Ig-bound and -unbound when using relative abundance (Ding et al., Reference Ding, Chen, Yu, Ross, Stanton, Zhang, Yang and Chen2022). While this can also be caused by differences in Ig-binding patterns at the lower taxonomic levels (Planer et al., Reference Planer, Peng, Kau, Blanton, Ndao, Tarr, Warner and Gordon2016), it was found that even the same species can be in both Ig+ and Ig− fractions (Ding et al., Reference Ding, Chen, Yu, Ross, Stanton, Zhang, Yang and Chen2022). While acknowledging the limitations of 16S gene sequencing in resolving species-level classifications (Janda and Abbott, Reference Janda and Abbott2007), I–microbiota associations are mostly low-affinity and the likelihood of binding of a taxon is based on probability. The goal of Ig-Seq is therefore to look at the overall relative preference of Ig binding of taxa, and the Ig probability ratio and Ig+ probability are able to do so, while relative abundance measures are less informative. The Ig probability ratio and Ig+ probability have an advantage over differential abundance measures by taking into account pre-sort abundances and relative abundances of taxa in the same formula. We believe that the Ig probability ratio can be considered for use where possible, or the Ig+ probability as a second option in the analysis of Ig-Seq data (Figure 2). Where multiple Ig isotypes are used in a four-way sort, it is possible to use the Ig probability, followed by linear mixed effect modelling to account for the FACS quadrant (Conrey et al., Reference Conrey, Denu, O’Boyle, Rozich, Green, Maslanka, Lubin, Duranova, Haltzman, Gianchetti, Oldridge, De Luna, Vella, Allman, Spergel, Tanes, Bittinger, Henrickson and Silverman2023). Use of standardised measures as suggested will also contribute to consistency in results, which are comparable across studies, which is currently missing in the field.

Comparing sorted fractions to the overall microbiota

In comparing the common overall unsorted microbiota to the sorted fractions across studies, we found differences in taxa after sorting. Various taxa could only be found in the overall fraction and lost after sorting while others were enriched after sorting (Table 1). Differences might be explained from processing of samples for sorting resulting in loss of bacteria from the sorted fractions. Overall, Bifidobacterium was found in sorted fractions in all three sample types. This suggests a possibility for enrichment of Bifidobacterium in the sorted fractions and a high affinity for Ig.

A possibility of a sorting bias resulting in enrichment of Alphaproteobacteria in sorted fractions was suggested by one study (Janzon et al., Reference Janzon, Goodrich, Koren, Waters and Ley2019), due to loss of bacteria that aggregate in the sorting process. However, we do not find this to be the case when comparing taxa found to be recurrent across studies according to sample type. The post-sort fractions are more likely to be enriched in certain taxa based on high affinity of binding to Ig, with higher confidence in infant stool results where multiple studies report similar findings despite differences in analysis methods. It is likely that the enrichment of Alphaproteobacteria after sorting is study specific or due to insufficient filtering of sequencing reads, and we recommend that stringent filtering and contaminant controls are employed to reduce enrichment of possible contaminants.

Type of Ig and characterisation of Ig–microbiota associations

In comparison of binding based on Ig isotype, three studies that investigated binding of multiple isotypes were referred to. Overall, IgA bound a greater percentage of microbes than IgM and IgG (Gopalakrishna et al., Reference Gopalakrishna, Macadangdang, Rogers, Tometich, Firek, Baker, Ji, Burr, Ma, Good, Morowitz and Hand2019). While there was a large overlap of microbes bound by IgA, IgM and IgG, IgA, and IgM were more similar in binding patterns compared to IgG, which bound a subset of microbes (Janzon et al., Reference Janzon, Goodrich, Koren, Waters and Ley2019). This is in agreement with another study in mice, where both studies concluded that the difference in binding specificity to microbes in IgG was to target pathogens or pathobionts which escaped binding of IgA and were able to cross the epithelial barrier (Koch et al., Reference Koch, Reiner, Lugo, Kreuk, Stanbery, Alison, Ansaldo, Seher, Ludington and Barton2016; Janzon et al., Reference Janzon, Goodrich, Koren, Waters and Ley2019). IgG is also not produced locally in the gut and tends to reflect systemic responses instead, such as targets of vaccination including Haemophilus in infancy (Janzon et al., Reference Janzon, Goodrich, Koren, Waters and Ley2019). In humans, the role of IgG binding of the microbiota is more accepted in the context of controlling specific members of the microbial population rather than aiding in colonisation of broad classes of bacteria. IgA is the most dominant antibody in mucosal immunity, and is the main antibody implicated in binding of microbiota (Chen et al., Reference Chen, Magri, Grasset and Cerutti2020). The similar binding specificities of IgA and IgM provide an allowance for functional redundancy, as seen in selective IgA deficiency, where IgM contributes to a degree of protection in place of IgA (Fadlallah et al., Reference Fadlallah, El Kafsi, Sterlin, Juste, Parizot, Dorgham, Autaa, Gouas, Almeida, Lepage, Pons, Le Chatelier, Levenez, Kennedy, Galleron, de Barros, Malphettes, Galicier, Boutboul, Mathian, Miyara, Oksenhendler, Amoura, Doré, Fieschi, Ehrlich, Larsen and Gorochov2018; Catanzaro et al., Reference Catanzaro, Strauss, Bielecka, Porto, Lobo, Urban, Schofield and Palm2019). The most important isotype to consider in the context of the microbiota is IgA, followed by IgM. IgG is also able to bind to the microbiota, but seems to be more important in the context of protection against pathogens or pathobionts.

Humans have two subclasses of IgA–IgA1 and IgA2 (Chen et al., Reference Chen, Magri, Grasset and Cerutti2020). One study compared the binding patterns of both subclasses in colostrum and infant stool (Sánchez-Salguero et al., Reference Sánchez-Salguero, Corona-Cervantes, Guzmán-Aquino, de la Borbolla-Cruz, Contreras-Vargas, Piña-Escobedo, García-Mena and Santos-Argumedo2021). IgA1 and IgA2 were found to target different bacteria (Sánchez-Salguero et al., Reference Sánchez-Salguero, Corona-Cervantes, Guzmán-Aquino, de la Borbolla-Cruz, Contreras-Vargas, Piña-Escobedo, García-Mena and Santos-Argumedo2021). IgA1-bound bacteria in colostrum included Staphylococcus, Streptococcus, and Clostridium, while IgA2-bound bacteria in colostrum were Bifidobacterium, Lactobacillus, and Bacteroides, and IgA2-bound bacteria in neonatal stool were Bifidobacterium, Pseudomonas, and Paracoccus (Sánchez-Salguero et al., Reference Sánchez-Salguero, Corona-Cervantes, Guzmán-Aquino, de la Borbolla-Cruz, Contreras-Vargas, Piña-Escobedo, García-Mena and Santos-Argumedo2021). There was no IgA1-associated microbiota in neonatal stool and 60% of IgA2-associated bacteria in neonatal stool potentially originated from IgA2-associated colostrum, according to a Source Tracker analysis (Sánchez-Salguero et al., Reference Sánchez-Salguero, Corona-Cervantes, Guzmán-Aquino, de la Borbolla-Cruz, Contreras-Vargas, Piña-Escobedo, García-Mena and Santos-Argumedo2021). The authors suggest that there is a role for IgA2-associated bacteria in maternal to infant transfer in early life, in the establishment of the infant microbiome (Sánchez-Salguero et al., Reference Sánchez-Salguero, Corona-Cervantes, Guzmán-Aquino, de la Borbolla-Cruz, Contreras-Vargas, Piña-Escobedo, García-Mena and Santos-Argumedo2021). IgA1 is more abundant in the small intestine and is sensitive to degradation by proteases in the large intestine, and is not found in stool samples (Brandtzaeg and Johansen, Reference Brandtzaeg and Johansen2005). IgA2 is protease-resistant and predominates in the large intestine (Brandtzaeg and Johansen, Reference Brandtzaeg and Johansen2005), and will be the main subclass of IgA in stool samples. The association of Bifidobacterium with IgA2 in both milk and neonatal stool samples may suggest a role of IgA2-associated bacteria in the development of the infant microbiome. Bifidobacteria are also important taxa in the infant microbiota due to their ability to utilise human milk oligosaccharides (HMOs) in breastmilk (Lawson et al., Reference Lawson, O’Neill, Kujawska, Gowrinadh Javvadi, Wijeyesekera, Flegg, Chalklen and Hall2020; Stuivenberg et al., Reference Stuivenberg, Burton, Bron and Reid2022), and transfer via maternal antibodies may be a mechanism to ensure infants obtain these beneficial microbes. Given that IgA1 is present in the small intestine, where the majority of Ig–microbiota associations are thought to occur (Bunker et al., Reference Bunker, Flynn, Koval, Shaw, Meisel, McDonald, Ishizuka, Dent, Wilson, Jabri, Antonopoulos and Bendelac2015), there is a possibility that IgA1 from breastmilk is more important when transferred to the small intestine, where microbial antigens are known to be immunostimulatory (Mowat and Agace, Reference Mowat and Agace2014; Bunker et al., Reference Bunker, Flynn, Koval, Shaw, Meisel, McDonald, Ishizuka, Dent, Wilson, Jabri, Antonopoulos and Bendelac2015; Million et al., Reference Million, Tomas, Wagner, Lelouard, Raoult and Gorvel2018). When considering subclasses of IgA, both IgA1 and IgA2 can be found in colostrum, and the connection of IgA2-bound bacteria in the colostrum and infant stool may be of interest in infant gut development.

Some other considerations to take note of include IgA production in formula-fed infants and characteristics of Ig–microbiota associations. It was reported in one study that formula-fed infants had less than 1% of IgA-bound microbiota, except for some that had up to around 40% of IgA-bound bacteria (and could not be explained), while breast-fed infants had a mean of ~25% IgA-bound microbiota (Gopalakrishna et al., Reference Gopalakrishna, Macadangdang, Rogers, Tometich, Firek, Baker, Ji, Burr, Ma, Good, Morowitz and Hand2019). IgA is maternally transmitted through breastmilk in the first month of life before the infant produces its own IgA, and meconium from neonates before breastfeeding does not contain IgA (Gopalakrishna et al., Reference Gopalakrishna, Macadangdang, Rogers, Tometich, Firek, Baker, Ji, Burr, Ma, Good, Morowitz and Hand2019; Sánchez-Salguero et al., Reference Sánchez-Salguero, Corona-Cervantes, Guzmán-Aquino, de la Borbolla-Cruz, Contreras-Vargas, Piña-Escobedo, García-Mena and Santos-Argumedo2021). Infant samples in the study were collected between 7 and 40 days of life (Gopalakrishna et al., Reference Gopalakrishna, Macadangdang, Rogers, Tometich, Firek, Baker, Ji, Burr, Ma, Good, Morowitz and Hand2019), and most infants would have been unable to produce IgA if formula-fed. This is important to consider for designs in Ig-Seq studies as formula-fed infants would not be expected to have IgA-bound microbiota in the first month of life. More generally, Ig–microbiota binding is independent of the relative abundance of a taxon in the overall microbiota and of total Ig levels (Dzidic et al., Reference Dzidic, Abrahamsson, Artacho, Bjorksten, Collado, Mira and Jenmalm2017). This allows Ig-Seq to provide more depth in understanding of Ig–microbiota interactions that cannot be captured in current routine analyses.

Comparing Ig+ and Ig− bacteria in different sample types

Differences in Ig-binding patterns were compared across sample types. In infant stool, several taxa were able to be found in both Ig+ and Ig− fractions. The taxa were Bifidobacteriaceae, Streptococcaceae, and Bifidobacterium (Table 2). Given that the taxa were found in several studies, it is likely that differences in binding exist at the lower taxonomic levels, as was found in the analysis of species or strains (Palm et al., Reference Palm, Zoete, Marcel, Cullen, Barry, Stefanowski, Hao, Degnan, Hu, Peter, Zhang, Ruggiero, Cho, Goodman and Flavell2014; Planer et al., Reference Planer, Peng, Kau, Blanton, Ndao, Tarr, Warner and Gordon2016). It is also important to use measures that are discriminatory for overall relative binding in analysis to avoid having a taxon being both Ig+ and Ig−, as well as to look at the lower taxonomic levels for better resolution of taxa. Enterobacteriaceae were only found in the Ig+ fraction in infant stool, while Bacteroides was Ig− in maternal stool, and Ig+ bacteria in breastmilk were Staphylococcus, Streptococcus, and Pseudomonas. Enterobacteriaceae have been described as pro-inflammatory and Ig binding of the taxon in a healthy population is in line with reduced IgA binding of the taxon in disease states such as NEC (Gopalakrishna et al., Reference Gopalakrishna, Macadangdang, Rogers, Tometich, Firek, Baker, Ji, Burr, Ma, Good, Morowitz and Hand2019. Ig binding of Enterobacteriaceae is likely to control the population to prevent overgrowth and disease. On the other hand, Bacteroides has been reported to have lipopolysaccharide (LPS) which is anti-inflammatory or immune-inhibitory, which may explain its presence in the IgA− fraction (Pither et al., Reference Pither, Illiano, Pagliuca, Jacobson, Mantova, Stornaiuolo, Colicchio, Vitiello, Pinto, Silipo, Fischbach, Salvatore, Amoresano, Molinaro and Di Lorenzo2022; Li et al., Reference Li, Xu, Zhai, Yang, Yang, Ke, Chen, Ou, Sha and Xiao2023). While LPS is typically thought to be pro-inflammatory, studies have shown that the inflammatory status of LPS is dependent on the bacteria and may change in the host (Paciello et al., Reference Paciello, Silipo, Lembo-Fazio, Curcurù, Zumsteg, Noël, Ciancarella, Sturiale, Molinaro and Bernardini2013; Vatanen et al., Reference Vatanen, Kostic, d’Hennezel, Siljander, Franzosa, Yassour, Kolde, Vlamakis, Arthur, Hamalainen, Peet, Tillmann, Uibo, Mokurov, Dorshakova, Ilonen, Virtanen, Szabo and Porter2016; Li et al., Reference Li, Xu, Zhai, Yang, Yang, Ke, Chen, Ou, Sha and Xiao2023).

Across sample types, Bifidobacterium was only Ig+ in maternal stool and breastmilk, and both Ig+ and Ig− in infant stool, and it is possible that this is a result of age-related differences in Ig–microbiota binding patterns. The presence of Bifidobacterium in maternal samples corresponds with results from the WTR analysis, where it was reported that B. longum was the main bacterium transferred from the maternal gut to breastmilk (Qi et al., Reference Qi, Tu, Zhou, Tu, Chang, Chen, Hu, Yu and Sun2022). The strains B. longum subsp. infantis BT1, B. longum subsp. infantis JCM 11347, and L. gasseri HL75 were also common to both breastmilk and infant stool in one study (Qi et al., Reference Qi, Tu, Zhou, Tu, Chang, Chen, Hu, Yu and Sun2022). Given that Bifidobacterium Ig binding may contribute to transfer from maternal breastmilk to the infant gut, and that Bifidobacterium can also be transferred to the infant during delivery, IgA2-binding of Bifidobacterium may reflect transfer from mother to infant across sample types (Makino et al., Reference Makino, Kushiro, Ishikawa, Kubota, Gawad, Sakai, Oishi, Martin, Ben-Amor, Knol and Tanaka2013; Sánchez-Salguero et al., Reference Sánchez-Salguero, Corona-Cervantes, Guzmán-Aquino, de la Borbolla-Cruz, Contreras-Vargas, Piña-Escobedo, García-Mena and Santos-Argumedo2021).

Changes in Ig–microbiota associations with time

In comparison of Ig–microbiota associations at common timepoints, a total of four timepoints for infant stool, one timepoint for maternal stool, and two timepoints for breastmilk were analysed (Table 3). In the stool of breast-fed infants, Bifidobacterium, Escherichia, and Pseudomonas were recurrent Ig-associated bacteria at 1 week of age, and there were no recurrent Ig-associated bacteria at 1 month of age. There were insufficient studies to compare Ig- bacteria at 1 week of age, and Ig- bacteria at 1 month of age were Clostridiales and Bifidobacteriales, based on two studies using relative abundance. At 12 months of age, IgA+ bacteria were Ruminococcaceae and Bifidobacteriaceae. Bifidobacteriaceae was also detected at 16 months of age. IgA− bacteria were Bacteroidaceae, Ruminoccocaceae, Streptococcaceae, and Lachnospiraceae at 12 months of age and Lachnospiraceae at 16 months of age. Ig–microbiota associations in maternal stool at 1 month postpartum were similar to the Ig-bound microbes in infant stool at 1 week of age (Bifidobacterium and Escherichia), with an additional Blautia. Breastmilk contained Bifidobacterium in the Ig+ fraction at 1 week postpartum during the colostrum stage, which was also found in infant stool and maternal stool in the first month postpartum.

The similarities in Ig–microbiota associations among maternal stool, breastmilk, and infant stool in the first week or month of life indicate the importance of Ig-bound bacteria. Bifidobacterium is beneficial to the infant and able to gain a selective advantage in the microbiota with HMO degradation as previously discussed, and Ig binding of the taxa may aid in colonisation of the mucosa (Lawson et al., Reference Lawson, O’Neill, Kujawska, Gowrinadh Javvadi, Wijeyesekera, Flegg, Chalklen and Hall2020; Stuivenberg et al., Reference Stuivenberg, Burton, Bron and Reid2022). Bifidobacterium or Bifidobacteriaceae was also detected as Ig-bound even at 12 and 16 months of age in infant stool, and remains an important taxon in the child’s gut microbiome. In one study, B. bifidum was found to be consistently IgA-bound after 3 months of age, and this is also shown in this analysis (Planer et al., Reference Planer, Peng, Kau, Blanton, Ndao, Tarr, Warner and Gordon2016). Escherichia, on the other hand, is potentially pathogenic and Ig binding of the microbe as transmitted from the mother may be to prevent it from entering systemic circulation, especially before weaning, when the infant’s immune system has not fully matured. (Zeng et al., Reference Zeng, Cisalpino, Varadarajan, Hellman, Warren, Cascalho, Inohara and Núñez2016). This finding is reflected in other studies, where breastfeeding increased the Ig-binding of E. coli, and Escherichia-Shigella was found to be IgA-bound across maternal stool, infant stool, and breastmilk (Planer et al., Reference Planer, Peng, Kau, Blanton, Ndao, Tarr, Warner and Gordon2016; Ding et al., Reference Ding, Chen, Yu, Ross, Stanton, Zhang, Yang and Chen2022). IgA-binding of Escherichia was also found to decrease colonisation of the genus in an in vitro model of human microbiota (Sasaki et al., Reference Sasaki, Mori, Hoshi, Sasaki, Inoue, Shinkura and Kondo2021), while N-glycan-mediated IgA binding conferred protection against E. coli O55 challenge in GF piglets (Raskova Kafkova et al., Reference Raskova Kafkova, Brokesova, Krupka, Stehlikova, Dvorak, Coufal, Fajstova, Srutkova, Stepanova, Hermanova, Stepankova, Uberall, Skarda, Novak, Vannucci, Tlaskalova-Hogenova, Jiraskova Zakostelska, Sinkora, Mestecky and Raska2021). Additionally, adherent-invasive E. coli (AIEC) associated with Crohn’s disease induced production of IgA and specific binding of AIEC, while commensal E. coli did not induce IgA production or binding in mice (Tanaka et al., Reference Tanaka, Imai, Tsugawa, Eap, Yazawa, Kaneko, Ohno, Sugihara, Kitamoto, Nagao-Kitamoto, Barnich, Matsushima, Suzuki, Kagawa, Nishizaki, Suzuki, Kamada and Hozumi2023).

There was also a shift in Ig-bound or -unbound bacteria in infant stool with time. While at 1 week of life, Ig-bound microbes consist of important commensals and pathogens, there is a high level of individuality in Ig–microbiota binding patterns at around 1 month of age, and no recurrent Ig-bound taxa were recorded. This is in agreement with the finding that Ig–microbiota binding patterns in unrelated infants were weakly correlated at 6 months of age, suggesting a lack of recurrent Ig–microbiota binding patterns early in life (Planer et al., Reference Planer, Peng, Kau, Blanton, Ndao, Tarr, Warner and Gordon2016). There was also a genetic component affecting Ig–microbiota interactions, with Ig–microbiota binding patterns in twins being more similar than in unrelated individuals up to 21 months of age (Planer et al., Reference Planer, Peng, Kau, Blanton, Ndao, Tarr, Warner and Gordon2016). There is a shift in the Ig-bound and -unbound microbes detected at 12 months onwards, such as Ruminococcaceae and Lachnospiraceae, corresponding to a shift in predominant microbes found in the infant microbiome due to transitioning from breastmilk to solid food as a dietary source (Backhed et al., Reference Backhed, Roswall, Peng, Feng, Jia, Kovatcheva-Datchary, Li, Xia, Xie, Zhong, Khan, Zhang, Li, Xiao, Al-Aama, Zhang, Lee, Kotowska, Colding, Tremaroli, Yin, Bergman, Xu, Madsen, Kristiansen, Dahlgren and Wang2015; McKeen et al., Reference McKeen, Roy, Mullaney, Eriksen, Lovell, Kussman, Young, Fraser, Wall and McNabb2022). While there were insufficient timepoints common across studies to study Ig–microbiota binding patterns beyond 16 months of age, it was reported that Ig–microbiota binding patterns were more correlated at 2 years of age, when they resembled those of the adult (Planer et al., Reference Planer, Peng, Kau, Blanton, Ndao, Tarr, Warner and Gordon2016). This again corresponds to maturation of the infant gut microbiome, and shows that the Ig–microbiota binding patterns are dynamic with time and shifts in the overall microbiome due to environmental and host factors (Backhed et al., Reference Backhed, Roswall, Peng, Feng, Jia, Kovatcheva-Datchary, Li, Xia, Xie, Zhong, Khan, Zhang, Li, Xiao, Al-Aama, Zhang, Lee, Kotowska, Colding, Tremaroli, Yin, Bergman, Xu, Madsen, Kristiansen, Dahlgren and Wang2015; Wernroth et al., Reference Wernroth, Peura, Hedman, Hetty, Vicenzi, Kennedy, Fall, Svennblad, Andolf, Pershagen, Theorell-Haglöw, Nguyen, Sayols-Baixeras, Dekkers, Bertilsson, Almqvist, Dicksved and Fall2022).

Ig–microbiota binding patterns in health and disease

Differences in Ig–microbiota binding patterns were also observed between populations of health and disease. This adds greater depth to the analysis and may explain the development of clinical disease better than the overall microbiota (Gopalakrishna et al., Reference Gopalakrishna, Macadangdang, Rogers, Tometich, Firek, Baker, Ji, Burr, Ma, Good, Morowitz and Hand2019). This was investigated in four studies on undernutrition, allergic disease, and NEC (Kau et al., Reference Kau, Planer, Liu, Rao, Yatsunenko, Trehan, Manary, Liu, Stappenbeck, Maleta, Ashorn, Dewey, Houpt, Hsieh and Gordon2015; Dzidic et al., Reference Dzidic, Abrahamsson, Artacho, Bjorksten, Collado, Mira and Jenmalm2017; Gopalakrishna et al., Reference Gopalakrishna, Macadangdang, Rogers, Tometich, Firek, Baker, Ji, Burr, Ma, Good, Morowitz and Hand2019).

Undernutrition was studied in the context of kwashiorkor in humans and using mice. Enterobacteriaceae were enriched in the IgA+ fraction in twins discordant for kwashiorkor as compared to concordant healthy twins at diagnosis, as well as in GF mice transplanted with stool from 21-month-old kwashiorkor twins fed undernourished diets (Kau et al., Reference Kau, Planer, Liu, Rao, Yatsunenko, Trehan, Manary, Liu, Stappenbeck, Maleta, Ashorn, Dewey, Houpt, Hsieh and Gordon2015). Enterobacteriaceae Ig binding was modifiable with a peanut-based ready-to-use-therapeutic food treatment (Kau et al., Reference Kau, Planer, Liu, Rao, Yatsunenko, Trehan, Manary, Liu, Stappenbeck, Maleta, Ashorn, Dewey, Houpt, Hsieh and Gordon2015). Differences in Ig–microbe associations were also found in diet alone, with Erysipelotrichaceae being IgA-bound in mice fed undernourished diets compared to those fed standard diets (Kau et al., Reference Kau, Planer, Liu, Rao, Yatsunenko, Trehan, Manary, Liu, Stappenbeck, Maleta, Ashorn, Dewey, Houpt, Hsieh and Gordon2015). The IgA index and presence of enteropathogenic E. coli in the microbiota were also correlated to stunting and wasting scores for clinical outcomes in kwashiorkor (Kau et al., Reference Kau, Planer, Liu, Rao, Yatsunenko, Trehan, Manary, Liu, Stappenbeck, Maleta, Ashorn, Dewey, Houpt, Hsieh and Gordon2015). Differences in Ig binding patterns in undernutrition were also found in another study using mice, where loss of binding of beneficial bacteria such as Lactobacillus and Bifidobacterium was found in undernourished mice in stool samples, and loss of binding of Lactobacillus in stool and jejunal samples in undernourished mice with additional gavage of bacterial strains to induce environmental enteric dysfunction (Huus et al., Reference Huus, Bauer, Brown, Bozorgmehr, Woodward, Serapio-Palacios, Boutin, Petersen and Finlay2020). It was found that loss of IgA-Lactobacillus associations and resultant reduced mucosal colonisation were independent of host IgA (Huus et al., Reference Huus, Bauer, Brown, Bozorgmehr, Woodward, Serapio-Palacios, Boutin, Petersen and Finlay2020). Lactobacillus adapts to the lack of nutrients by changing surface glycans, which is also linked to mutations in carbohydrate transport and metabolism (Huus et al., Reference Huus, Bauer, Brown, Bozorgmehr, Woodward, Serapio-Palacios, Boutin, Petersen and Finlay2020). This then affects the ability of Lactobacillus to bind to IgA via glycan-mediated interactions and reduces mucosal colonisation (Huus et al., Reference Huus, Bauer, Brown, Bozorgmehr, Woodward, Serapio-Palacios, Boutin, Petersen and Finlay2020). This shows the effect of diet and the overall microbiota on Ig–microbiota associations in undernutrition, and how glycan-mediated Ig-binding in the host is affected by microbial adaptation in undernutrition.

Differences in Ig–microbiota binding patterns were also described in children who developed allergic disease and those who remained healthy at 7 years of age (Dzidic et al., Reference Dzidic, Abrahamsson, Artacho, Bjorksten, Collado, Mira and Jenmalm2017). Differences in the IgA–microbiota associations were already detected early in life at 1 month of age in allergy and asthma, and were present at 12 months of age in allergy, asthma, and ARC (Dzidic et al., Reference Dzidic, Abrahamsson, Artacho, Bjorksten, Collado, Mira and Jenmalm2017). Differences in IgA–microbiota associations were also observed for the three different types of allergic disease, with increased binding of Lachnospiraceae and decreased IgA binding of Bacteroides in allergy, reduced IgA binding of Escherichia/Shigella in asthma and decreased IgA binding of Rhodobacter and Bacteroides in ARC as compared to healthy controls (Dzidic et al., Reference Dzidic, Abrahamsson, Artacho, Bjorksten, Collado, Mira and Jenmalm2017). Microbiota predominant in Bacteroides has been associated with reduced immune education in early life and resultant autoimmune disease (Vatanen et al., Reference Vatanen, Kostic, d’Hennezel, Siljander, Franzosa, Yassour, Kolde, Vlamakis, Arthur, Hamalainen, Peet, Tillmann, Uibo, Mokurov, Dorshakova, Ilonen, Virtanen, Szabo and Porter2016). Reduced binding of Bacteroides in allergy and ARC compared to healthy individuals may reflect greater immune-inhibition from Bacteroides LPS affecting immune education and immune-related disease development. Probiotic supplementation in the study was treated as a confounding factor in the analysis, and there was no effect of the supplementation on IgA–microbiota binding patterns or proportions of binding in allergic disease and healthy children (Dzidic et al., Reference Dzidic, Abrahamsson, Artacho, Bjorksten, Collado, Mira and Jenmalm2017). The same authors followed up by studying IgA–microbiota binding patterns in the breastmilk of mothers, as they showed that the probiotic affected the overall breastmilk microbiota (Dzidic et al., Reference Dzidic, Abrahamsson, Artacho, Bjorksten, Collado, Mira and Jenmalm2017). Specifically, the genus Rothia was reduced in the probiotic group (Dzidic et al., Reference Dzidic, Abrahamsson, Artacho, Bjorksten, Collado, Mira and Jenmalm2017). However, there was no significant difference in IgA–microbiota associations between the breastmilk of mothers of children who were diagnosed with allergic disease or not, or between probiotic and placebo groups (Dzidic et al., Reference Dzidic, Abrahamsson, Artacho, Bjorksten, Collado, Mira and Jenmalm2017). It was found that reduced richness of the breastmilk microbiota was correlated with allergy development instead (Dzidic et al., Reference Dzidic, Abrahamsson, Artacho, Bjorksten, Collado, Mira and Jenmalm2017). As the maternal cohort was predominantly allergic, and IgA–microbiota associations are distinct from the overall microbiota composition, is it possible that this affected the breastmilk IgA–microbial interactions (Dzidic et al., Reference Dzidic, Abrahamsson, Artacho, Bjorksten, Collado, Mira and Jenmalm2017). Additionally, they found that Bifidobacterium was the main taxon implicated in possible Ig–microbiota-related vertical transmission through breastmilk, as mentioned earlier. IgA binding of Bifidobacterium, or its presence in the overall microbiota, was not described in the study. IgA–microbiota binding patterns differ between types of allergic disease compared to healthy populations in infant stool, but further research is needed in non-allergic maternal cohorts to determine if differences in IgA-microbiota associations also apply to breastmilk and allergic disease outcomes in children.

One study investigated the IgA–microbiota associations in the development of NEC, focusing specifically on Enterobacteriaceae (Gopalakrishna et al., Reference Gopalakrishna, Macadangdang, Rogers, Tometich, Firek, Baker, Ji, Burr, Ma, Good, Morowitz and Hand2019). While Enterobacteriaceae was enriched in the overall microbiota of infants who developed NEC, reduced IgA binding of Enterobacteriaceae was able to explain the clinical development of NEC better than the overall microbiota alone (Gopalakrishna et al., Reference Gopalakrishna, Macadangdang, Rogers, Tometich, Firek, Baker, Ji, Burr, Ma, Good, Morowitz and Hand2019). Increasing loss of IgA-Enterobacteriaceae binding was also detected prior to NEC diagnosis. Further analysis in mice showed that IgA from breastmilk was required to prevent NEC development (Gopalakrishna et al., Reference Gopalakrishna, Macadangdang, Rogers, Tometich, Firek, Baker, Ji, Burr, Ma, Good, Morowitz and Hand2019). This suggests that IgA–microbiota associations have a role in the prevention of disease, and loss of healthy associations results in disease states. Enterobacteriaceae are inflammatory taxa related to pathogenesis, and binding of this taxon by IgA in stool samples collected in the first month of life reinforces that pathogens are likely to be bound by IgA early in infant gut development. The transfer of IgA, able to bind to pathogenic Enterobacteriaceae through breastmilk, highlights the importance of breastfeeding and the passive transfer of immunity to the infant while the infant’s immune system is still developing and unable to produce its own IgA (Gopalakrishna et al., Reference Gopalakrishna, Macadangdang, Rogers, Tometich, Firek, Baker, Ji, Burr, Ma, Good, Morowitz and Hand2019). Loss of IgA-bound Enterobacteriaceae both precedes and is indicative of NEC development according to this study, and transfer of maternal IgA able to bind to Enterobacteriaceae through breastfeeding is possibly protective of infant NEC development. However, more studies are required to confirm this.

Current knowledge gaps include Ig–microbiota associations in maternal stool both pre- and postpartum, and the differences between the Ig-bound and -unbound taxa in breastmilk at different stages. Even in infant stool, there is a lack of common timepoints between 1 week and 1 month and between 1 month and 12 months. The changes in Ig–microbiota binding patterns according to critical periods of development have also not been reported. It also needs to be confirmed if the IgA2-association of microbes is the mechanism of transfer of commensals from mother to infant. Given studies in undernutrition reporting changes in surface glycans in microbes in response to nutritional pressures affecting IgA binding and mucosal colonisation, the importance of diet in Ig–microbiota associations and the ability to correct for associations using diet should also be studied. It is not known how diet changes in a healthy cohort are able to influence Ig–microbiota binding patterns, or if maternal diet is able to do so in the infant when breastfeeding. In allergic disease, the possibility of Bifidobacterium Ig binding in breastmilk and infant stool should be studied in a non-allergic maternal cohort. There is also a lack of understanding of how maternal and infant stool Ig–microbiota associations are related, and whether maternal Ig–microbiota associations would affect those of the infant. Future studies involving all three sample types are required to arrive at more robust conclusions. Dietary and probiotic interventions have been used in some studies included; however, while one study showed that dietary intervention was able to influence Ig–microbiota binding patterns, other studies have not shown any effect of probiotic interventions on Ig–microbiota binding patterns. More studies are required to confirm this and determine factors able to influence Ig–microbiota binding patterns to correct them in the prevention of disease development.