AOAC HMO-06 Reviewer Forms (July 25, 2023)

AOAC Stakeholder Program on Infant Formula and Adult Nutritionals (SPIFAN)

HMO-06 REVIEWER FORMS

July 25, 2023

AOAC INTERNATIONAL 2275 Research Blvd., Suite 300 Rockville, MD, 20850 USA

dboyd@aoac.org 301.924.7077 x126

FOR EXPERT REVIEW PANEL USE ONLY NOT FOR DISTRIBUTION

Expert Review Panel for Infant Formula and Adult Nutrition Evaluation of Method HMO-06

Title: AOAC Official Method 20xx.xx Analysis of Eight Human Milk Oligosaccharides (HMOs) in Infant Formula and Premix Samples by Labelling with Benzocaine and Analysis by HPLC-UV First Action 20xx Author: Dora Molnar-Gabor, Marton Lengyel, Teddy Krongaard Reviewer Name: Reviewer 1 Summary of Method: This method uses a labeling procedure using 4-aminobenzoic acid ethyl ester (benzocaine) and picoline borane as the reducing agent, then applies HPLC separation with UV detection for the analysis of the eight HMOs. Method Scope/Applicability: Method has been applied to the analysis of 8 HMOs (2’-FL, 3-FL, DFL, LNT, LNnT, 6’-SL, 3’-SL, and LNFP-I) in ingredients (premixes) and selected milk protein based infant formula matrices containing maltodextrin. General comments about the method: Document appears to already be largely formatted in alignment with AOAC guidelines. Recommend some additional guidance on establishing equivalence for the reagents listed (purity >x?, impurity

1

FOR EXPERT REVIEW PANEL USE ONLY NOT FOR DISTRIBUTION

Cons/Weaknesses • Few matrices and/or potential interferences studied • LNT and LNnT are not baseline resolved, this may complicate analysis of one or the other when one is present at significantly higher levels than the other. o LOQ of LNT especially may then be problematic due to need to increase injection volume to improve LOQ of this analyte. In many samples it would then not pass the SMPR criterion. • IS not added until relatively late in the prep. Supporting Data • General Comment: I wish more matrices were evaluated against this method to test the ruggedness of the technique. I’d especially like to see samples containing GOS as well as protein sources other than bovine milk.

Method Optimization: Consider adding IS earlier in prep

• Performance Characteristics:

Analytical Range: 1.25-50,000 mg/100g for all analytes except 2’-FL which was 2-50,000 mg/100g

LOQ: (mg/100g)

2’-FL

3-FL

DFL 0.2

3’-SL

6’-SL

LNT 0.3*

LNnT

LNFP-I

0.1

0.1

0.3

0.4

0.3

0.2

*May be debatable due to noted resolution/injection volume concern noted above.

Accuracy/Recovery: (%)

2’-FL

3-FL

DFL

3’-SL

6’-SL

LNT

LNnT 91-98

LNFP-I 92-99

99-104

95-102

98.107

94-107

97-105

94-101

Precision (RSD r ): (%)

2’-FL

3-FL

DFL

3’-SL

6’-SL

LNT

LNnT

LNFP-I 0.4-3.5

0.6-2.6

1.3-2.6

1.4-2.4%

0.3-4.1

0.7-4.3

0.3-1.9

0.3-1.8

Reproducibility (RSD R ): N/A

• System suitability: None listed

2

FOR EXPERT REVIEW PANEL USE ONLY NOT FOR DISTRIBUTION

1. Is the Validation Study Report in a format acceptable to AOAC? - Yes with some minor edits

2. Is the method described in sufficient detail so that it is relatively easy to understand, including equations and procedures for calculation of results (are all terms explained)? - Yes

3. Are the figures and tables sufficiently explanatory without the need to refer to the text?

- Yes

4. Are all the figures and tables pertinent?

- Yes

5. Could some be omitted and covered by a simple statement?

- No

6. Are the references complete and correctly annotated?

- Yes

7. Does the method contain adequate safety precaution reference and/or statements? - Yes

Recommendation: I do not recommend this method for first action at this time since it does not have data on enough sample matrices to fully assess it.

3

FOR EXPERT REVIEW PANEL USE ONLY NOT FOR DISTRIBUTION

Expert Review Panel for Infant Formula and Adult Nutrition Evaluation of Method: HMO-06

Title: Analysis of Eight Human Milk Oligosaccharides (HMOs) in Infant Formula and Premix Samples by Labelling with Benzocaine and Analysis by HPLC-UV Author: Molnar-Gabor,D.; Lengyel,M. Krongaard,T. Reviewer Name: Reviewer 2 Summary of Method: • Powder samples are reconstituted in water (as SPIFAN recommendation) and homogenized in a sonic bath. • A sample of reconstituted sample is further diluted to get HMO within range of calibration curve • Optional steps to remove maltodextrins (enzymatic treatment with AMG) and proteins (ethanol precipitation) are included. • An aliquot of the diluted (and optionally treated) sample is taken and an internal standard (pannose) is added. • To the above solution the labeling reagent (benzocaine + picoline borane in DMSO/Acetic acid) is added and the labelling reaction carried out at 60°C for 3 h. • The labelled samples are further diluted by addition of acetonitrile then analysed by HPLC-UV (detection @ 300nm) on a HILIC stationary phase using a mobile phase of acetonitrile / ammonium formate (6.66 mM, pH 3.0) • A guard column and switching valve are used to remove excess reagents on-line prior to the analytical column. • Quantification is performed against a standard curve plotting peak area (analyte / IS) against concentration.

Method Scope/Applicability:

Method claims to be “applicable for analysis of 8 HMOS (2’FL, 3FL, DFL, 3’SL, 6’SL, LNT, LNnT, LNFP-I) in infant formula, premixes and blend samples”

General comments about the method: Method appears to be fairly straight-forward to follow, and easy to implement. The LC run time is very short (only 14 min in total) which is good, however I do have some concerns regarding the gradient, as the “column wash” appears to start at 10 min and run until 11 min (during which time LNT and LNnT are still eluting) and LNFP-I elutes at 11.3 min which is after the re-equilibration has started. Clearly this works well in the authors laboratory, but if the method is transferred to different systems with different plumbing / gradient delays there is a risk that we could run in to problems. Personally, I would be tempted to sacrifice some speed to gain some robustness to ensure easier method transfer. As with all methods for HMOs, we

1

FOR EXPERT REVIEW PANEL USE ONLY NOT FOR DISTRIBUTION

recommend adding some notes about the importance of the calibration standards and how they should be controlled (currently not included in the method).

Method Clarity: Fairly clear and simple to follow

Pros/Strengths: • Easy method to follow • Uses standard HPLC-UV equipment found in almost all routine labs • Short LC run time

Cons/Weaknesses • Requires HMOs to be derivatized (and the derivatization reaction is long – 3h) • Several “optional” steps which can complicate the calculation and impact LoQ/sensitivity since they introduce additional dilutions. • Very short LC run time may be difficult to transfer between labs (may be better to add some additional time to enable easier transfer, otherwise introduce additional information to specify instrument gradient delays and how to correct for them). • In the “principle” section it is stated that AMG treatment can be used to remove maltodextrin, starch, dried glucose syrup, “or other glucose polymers”. However it is clear that AMG will not remove any type of glucose polymer (not beta-glucans, not polydextrose, etc) so this should be corrected. • In the “derivatization” section it is stated “….pipette 50-50 uL of each …..” This is not clear. Should it just be 50 uL ? • In the section “evaluation” some expected relative retention times are listed, and a statement is given that “some variation” may be expected especially for 3’SL and 6’SL. It may be useful to put some quantitative measurement of how much variability may be acceptable for each HMO, otherwise the information is not so useful. • In the section “evaluation” the units are included in the equation. The units should not appear in the equation, but in the key to the abbreviations (it will also make the equation easier to read). •

Supporting Data

2

FOR EXPERT REVIEW PANEL USE ONLY NOT FOR DISTRIBUTION

• General Comment:

- Insufficient data have been provided to support the method applicability statement, or to confirm that the method meets the SMPRs. I would recommend the authors to: o Better describe the infant formula and “preblend” matrices o Add additional formula matrices including (but not limited to): Intact protein formula / partially hydrolyzed protein formula / elemental formula / soy-based formula / RTF formula o Make matrix-based LoQ estimates - Please report all data in the SMPR Units (mg/100g of reconstituted product) for the infant formula (for the preblends we have no SMPR, so in principle a different unit could be OK, since concentrations will be very high). This may be a “simple conversion” but it is extra work for busy reviewers – the “simple conversion” can be done by the authors to make everyone’s life easier. - when comparing vs SMPR I recommend making the comparison only on formula since the premixes are a different matrix for which we have no SMPR and the concentration range is completely different. - Also recommend improving the description of the validation experiment design. It is not clear to me how the precision experiments were performed. From the description it sounds like a single measurement was made on six different days (but the data set looks like samples were analysed in triplicate on 2 different days). This needs clarification. - The authors should be aware that there are currently no SMPRs for LNFP-I nor for premixes. I would recommend that the data be published in the SLV report, but the method itself may need to be limited to 7 HMO in formula if we want to go forward before any other SMPRs are available (with a possible view to upgrading in future if/when SMPRS do become available). Also, since there are currently no formula on the market containing LNFP-I, it will make generating samples for MLT a bit challenging.

• Method Optimization:

The authors have not considered the impact of other ingredients on their analysis. Particularly important are galactooligosaccharides, but it is also prudent to consider other potential interferants, for example polydextrose, probiotics, sialyllactoses containing N-glycolylneuraminic acid (found in significant concentrations in sheep or goat milk-based formula).

• Performance Characteristics:

Analytical Range:

IF I understood the data set correctly:-

2’FL: ~ 70 mg/100g 3FL: ~ 20 mg/100g DFL: ** ~ 9 mg/100g ** (Blank contains 3 mg/100g – likely overestimation !)

3

FOR EXPERT REVIEW PANEL USE ONLY NOT FOR DISTRIBUTION

LNT: ~ 20 mg/100g LNnT: ~ 11 mg/100g 3’SL: ~1.5 mg/100g – 7 mg/100g 6’SL: ~ 11 mg/100g

LOQ: LoQ data have been established on solutions in water as:-

IF I understood the data set correctly:-

2’FL: ~ 0.16 mg/100g (meets SMPR) 3FL: ~ 0.19 mg/100g (meets SMPR) DFL: ~ 0.35 mg/100g (meets SMPR) LNT: ~ 0.59 mg/100g (meets SMPR) LNnT: ~ 0.55 mg/100g(meets SMPR) 3’SL: ~0.58 mg/100g (meets SMPR) 6’SL: ~ 0.78 mg/100g (meets SMPR) LNFP-I: ~ 0.44 mg/100g

The authors have stated that their data do not meet the required LoQ but I think they have based this on their LoQ in powder samples …. Maybe I am wrong, but this would be easier to understand if the authors just used the correct SMPR units.

Accuracy/Recovery:

In formula for all HMOs: 91.2 – 103% (meets SMPR) In premixes for all HMOs: 94.3 – 101 %

Precision (RSD r ):

In formula for all HMOs: 1.3 – 4.1 % (meets SMPR) In premixes for all HMOs: 0.79 – 4.26 %

Reproducibility (RSD R ): n/a

• System suitability:

No specific section for this but under section I “sequence” it is mentioned that the r 2 of the calibration curve should be at least 0.999 and a check standard to be injected at least every 6 sample injections. In section J “evaluation” some guidance is given for expected relative retention times (however no specifications are really set).

4

FOR EXPERT REVIEW PANEL USE ONLY NOT FOR DISTRIBUTION

1. Is the Validation Study Report in a format acceptable to AOAC? Yes

2. Is the method described in sufficient detail so that it is relatively easy to understand, including equations and procedures for calculation of results (are all terms explained)? Yes

3. Are the figures and tables sufficiently explanatory without the need to refer to the text? No – In some cases the titles could be improved and it is not always easy to understand exactly what data is in the tables. 4. Are all the figures and tables pertinent? YES – I would also like to see plots of residuals next to calibration curves (they are easier to read than tables)

5. Could some be omitted and covered by a simple statement?

No

6. Are the references complete and correctly annotated? Yes. However I would like to see an additional reference to support the statement “ The most widely adopted reaction for this task is reductive amination, which has been used to determine HMOs from human milk as well as infant formula” (e.g. Anumula 2006 Anal Biochem 350: 1-23) or some papers analyzing HMOs using such approaches (there are plenty of examples).

7. Does the method contain adequate safety precaution reference and/or statements? Yes

Recommendation: Recommend no action at this time. Method looks promising, however the validation is insufficient. Authors should use wider range of formula matrices and multiple spike levels (preferably covering the range in the SMPR). Also the LoQs should be assessed in-matrix, not in water.

5