Newborn Screening for Methylmalonic Acidemia/propionic Acidemia: Systematic Evaluation of a Two-pronged Approach Based on MS/MS and UPLC-MS/MS


 Background: An improved second-tier test is needed to reduce the false-positive rate of newborn screening (NBS) for inborn metabolic disorders in Xuzhou, China.Methods: We designed an expanded second-tier assay using newborn dried blood spots (DBSs). Analytical and clinical performance were evaluated in 53 newborns with methylmalonic acidemia (MMA) or propionic acidemia (PA) reported by the Xuzhou Maternity and Child Health Care Hospital NBS program. Additionally, we analyzed NBS data regarding seasonal variation of metabolites, birth weight and gestational age to improve the identification of true positive MMA/PA individuals.Results: Among the 53 MMA/PA individuals assessed, two pathogenic or likely pathogenic (P/LP) variants in an MMA/PA-associated gene were identified in 46 patients, and a pathogenic variant and a variant of unknown significance (VUS) were identified in 7 patients. No such variants were detected in MMA/PA false-positive individuals or healthy controls. Ultraperformance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS)-based analysis of the initial NBS metabolic profile correctly identified MMA/PA individuals and reduced the initial NBS false-positive rate by 98.86%. MMA/PA false-positive infants in Xuzhou, China, were most likely to be summer-born.Conclusion: We established a two-pronged approach to reduce false positives by nearly 99% and provided a novel NBS strategy. Challenges in neonate metabolic testing and DNA variant interpretation regarding season, birth weight and pregnancy status remain for this Chinese population.


Introduction
Methylmalonic acidurias (MMAs) are characterized by inherited errors of metabolism that are attributed to total or partial de ciency of l-methylmalonyl-CoA mutase activity (the mut 0 and mut − subtypes, respectively), defects in the synthesis of its cofactor (5-adenosylcobalamin; the cblA, cblB, cblC and cblD MMA subtypes) or methylmalonyl-CoA epimerase de ciency [1][2][3] . The diagnosis of MMAs is based on the measurement of abnormal accumulation of MMA accompanied by increases in methylcitric acids and propionylglycine in urine, as well as propionylcarnitine (C3) in blood. Propionic acidemia (PA) is an autosomal recessive disorder with defects in the biotin-dependent enzyme propionyl-CoA carboxylase 4,5 . Large amounts of urinary 3-hydroxypropionic and methylcitric acids and tiglylglycine propionylglycine and the absence of MMA are required for the diagnosis of PA 6 . Similarly, increased amounts of C3 in the blood are detected. The worldwide prevalence of MMAs/PA varies widely depending on ethnicity and the method of ascertainment. Based on newborn screening (NBS) programs in Europe and the USA, the incidence of MMA de ciency is estimated to be 1:46,000 to 1:200,000 in Europe and the United States 7 . However, the incidence of MMA de ciency in China varies greatly and affects neonates from 1:3220 to 1:21,488 in different reports 8 . Conversely, for PA, the previously reported epidemiological data appears to be slightly lower, with an incidence of PA is a rare organic acidemia with an average estimated incidence of ~ 1:100,000-150,000 in the worldwide 9,10 . NBS using tandem mass spectrometry (MS/MS) and dried blood spots (DBSs) is a signi cant innovation in inherited metabolic disease detection 11 . Previous studies showed that determination of the C3 analyte content in DBSs, a common specimen collected by heel stick 48 ~ 72 hours after birth, can be used to con rm MMAs and PA; however, NBS with MS/MS assessment of C3 favors sensitivity over speci city and yields a number of false positive results. Although MS/MS screening is bene cial for maximizing the number of neonates identi ed (sensitivity), this approach has a slightly high false-positive rate caused by its poor speci city, which results in considerable adverse nancial and emotional burden effects given the required follow-up visits and diagnostic delays for true positive infants 12 . To reduce the false-positive rate, further evidence is needed for second-tier tests utilizing ultraperformance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS) to measure more speci c disease markers (e.g., MMA in DBSs) to con rm true positives or reject false positives from the primary screening 13 . Both primary and second-tier screening methods utilize the original DBSs and do not require a new blood draw, minimizing turnaround time.
Here, we adapted a novel validated supplemental UPLC-MS/MS technology using three 3-mm DBS punches and used it to evaluate archived DBSs from newborns with different types of positive screens for MMAs and PAs. C3associated disorder screening is fraught with false-positive cases that require second-tier con rmation using UPLC-MS/MS, and DNA sequencing is necessary to reach a nal diagnosis and to identify which pathogenic genes are responsible and the severity of the speci c variant. On the basis of 509,313 newborns born in Xuzhou, China, over a 5-year period between November 2015 and December 2020 C3, C3/C2 and methionine (Met) data measured by rst-tier MS/MS technologies, we developed a novel strategy using MMA, methylcitrate (MCA) and homocysteine (HCY) values as second-tier test parameters for differential screening in initially positive samples 14 . In addition to the primary MS/MS analytes currently used in NBS for MMAs and PA, our statistical approach utilizes information from the entire MS/MS metabolic pro le measured at birth. Determination of the C3 level in DBSs by MS/MS for NBS NBS for MMAs and PA was used to detect the C3 level in DBSs. Samples were collected on neonatal screening cards at 36 or 72 hours after birth and dried at room temperature using commercial kits. One DBS of 3 mm in size (equivalent to 3.2 µL whole blood) was placed in a 96-well U-shaped plate and preprocessed with 100 µL extraction solution containing internal standard mixture (NeoBase™ Nonderivatized MS/MS Kit, PerkinElmer™, Turku, Finland.). Then the plate was incubated at 45 ℃ for 45 min, transferred to a 96-well V-shaped plate and covered with aluminum lm before detection. After the samples were allowed to rest at room temperature for 2 hours, the concentrations of C3 in the DBSs were quanti ed by MS/MS (Xevo TQD/TQD, Waters, USA). The data were collected using electrospray positive ion scanning and the multiresponse monitoring mode. The total determination time was 3 min for each sample with a 15 µL injection volume. Masslynx V4.1 software was used to analyze the screening data.

Molecular Genetics analysis
Mutation analysis was performed for further diagnosis, involving DNA sequencing as described, and "site validation" of variants relevant to the screened disorders was analyzed. Using an ABI-3100 automated sequencer (Applied Biosystems, Foster City, CA, USA), molecular genetics were analyzed with mutation sites in pathogenic genes. The DNA sequencing pipeline developed for screening purposes, which included MUT, MMAB, MMACHC, PCCA, and PCCB genes as alternative targets, was employed to clinically diagnose individuals positive by MS/MS screening. As shown in supplementary Table 1, DNA panels used in our center included 78 genes which divided into two groups based on evidence for association with Xuzhou area metabolic conditions. Statistical analysis NBS metabolic data analysis: We performed a retrospective analysis of NBS data from 509,313 newborns that focused on 53 con rmed individuals' MS/MS analytes. Three analytes, namely, C2, C3 and Met, were interesting parameters in C3-associated disorders. We rst compared analyte levels between different MMA/PA subtype individuals and controls (Fig. 1). Analysis of variance (ANOVA) was used to compare the 46 analytes between three speci c phenotypic subgroups of 53 MMA patients (5 isolated MMA, 40 cblC de ciency and 8 PA individuals). As differences in seasons, brith weight and gestational age (GA) may be associated with distinct metabolic pro les, we further strati ed newborns into three subgroups: control, MMA/PA-negative (MMA/PA-N) and MMA/PA-true positive (MMA/PA-TP). Of 509,313 newborns, some individuals might have other abnormal metabolic levels that were detected by MS/MS and were removed from analysis. For comparison of distributions of C3, C3/C2, C3/C0 and Met values among neonates with Mann-Whitney U test and Kruskal-Wallis test, a p value < 0.05 was considered signi cant. Only C3 and C3/C2 assignments from screening samples were used to plot the receiver operating characteristic (ROC) curve ( Fig. 3A and B).
Second-tier test data analysis: Based on the guidelines of the Clinical and Laboratory Standards Institute (CLSI) (EP28-A3c: De ning, Establishing, and Verifying Reference Intervals in the Clinical Laboratory, Approved Guideline, 3rd edition), the D/R ratio was ≤ 1/3, and the two values of MMA and HCY were not outliers. Cases that lacked quantitative data (less than 5000 cases) were assessed using Kolmogorov-Smirnov (K-S) testing for normal distribution analysis. Signi cant data (p value < 0.05) were not identical to the normal distribution using K-S testing, and the percentile of a nonparametric test was used to establish the cutoff. A nonparametric test requires two values to be calculated, including the minimum reference value r1 [e.g. if the 2.5th percentile is the corresponding observed value r1 = 0.025 (n + 1)] and the maximum reference value r2 [e.g. if the 97.5th percentile is the corresponding observed value r2 = 0.975 (n + 1)]. Here, n represents the sampled data measurement.  Table 1. Indeed, not only isolated MMA individuals but also PA individuals were con rmed to carry compound heterozygous variants. We observed a strikingly high frequency of cblC-type individuals in our cohort (75.47%, n = 40/53), in which 80 independent cblC alleles were affected. The sequencing results con rmed the compound heterozygous or homozygous variants in the MMACHC gene in 40 infants, of which 7 were homozygous for c.609G > A (71.4%, 5/7), c.658_660del (14.3%, 1/7) and c.80A > G (14.3%, 1/7) (Supplemantal Table 1). To calculate metabolite levels, individuals homozygous for the MMACHC gene (patients 3,9,17,30,31, 33 and 40) were placed into one group for further analyses, with the other individuals placed in a second group. The mean C3/C2 ratio of the homozygous group obtained from the initially measured NBS data was 0.926, which was signi cantly different from that of the compound heterozygous group (C3/C2 ratio = 0.483). Correspondingly, the Met values in the homozygous group with MMACHC variants ranged from 3.34 to 11.01 µmol/L, with a mean value of 7.03 µmol/L, which was near the minimum value of the normal range of Met and lower than that in the compound heterozygous group. However, no signi cant difference in C3 values for individuals with cblC de ciency was found based on the biochemical data (Fig. 1).

Evaluation of NBS for MMAs and PA based on season, gestational age and birth weight
The levels of speci c metabolites detectable (e.g., C2) by MS/MS for NBS are known to have seasonal variation.
We assessed the seasonal pro le of MMA/PA screening parameters ( Fig (Fig. 2D). Within the MMA/PA-N group, there was a signi cantly higher proportion of preterm (GA ≤ 37 weeks) than full-term (> 37 weeks) births after analysis with the Mann-Whitney U test. An additional analysis of neonate birth weight (normal: 2,500-4,000 grams) revealed a relatively lower birth weight for MMA screening-positive newborns (both true positives and false positive newborns) than that for the healthy group (Fig. 2E).
To evaluate the sensitivity of MS/MS analytes for MMA/PA screening, receiver operating characteristic (ROC) curves were utilized to distinguish true and false positive MMA/PA individuals. The optimal screening e ciency was determined using a sensitivity of 0.957 and a speci city of 0.972 under the curve area with 0.996 for parameter C3 (Fig. 3A and B). However, the parameter C3/C2 had a better sensitivity and speci city than C3 with ROC analysis. The mean decrease in the accuracy (MDA) index (Fig. 3D) suggested that the C3/C2 ratio was a signi cant parameter for MMA/PA screening. However, other metabolic analytes, including the C0 and Met levels, showed similar rankings by MS/MS. Of the above several covariates analyzed, seasonal variation played a fundamental role in primary NBS. Ranking analysis of the MS/MS metabolites showed signi cant differences based on phenotypic subtype. Individuals with cobalamin de ciency might have lower Met levels than those with other phenotypes (Fig. 1C). Moreover, the C2 and Met levels declined sharply during the summer due to the higher air temperature in Xuzhou, and parameter C2 was crucial for identifying MMA/PA in the initial MS/MS screening (Fig. 3E/F).
Generalization of a novel second-tier approach: determination of cutoffs for MMA, MCA and HCY From January 2019 to March 2020, 263 samples with increased C3 concentrations at the initial NBS test were further assessed by novel second-tier UPLC-MS/MS screening. As the dataset was not symmetrical, a total of 1,830 negative subjects were used to calculate the cutoff value based on the percentile distribution of metabolites. The statistical results of the percentile distribution of the negative controls are shown in Table 1. Notably, the value "4.54" of the MMA metabolite and the value "12.64" of the HCY metabolite were two suspicious outliers. According to the NCCLS C28-A3 le, D/R (MMA)= (4.54-3.96)/(5.55-0.00) = 0.10, while D/R (HCY)= (12.64-10.03)/(17.14-0.9) = 0.16; however, both D/R ratios were ≤ 1/3, and there was no outlier in any of the three metabolite indexes. Previous literature and clinical research have indicated that sex does not affect the three analytes that were measured with UPLC-MS/MS. Based on the distribution of the histogram chart, the 99.5th percentile and the 99.7th percentile were considered the cutoff values for the second-tier test (Table 1). When using the 99.5th percentile as the cutoff value, the positive rate of MMA/PA detection was estimated to be 0.45%, 0.40% and 0.76%, respectively. When the 99.7th percentile was reached, the positive rate of detection for each disorder was reduced, and the number of recall cases decreased from 43 to 29 (Table 2). A comprehensive analysis of several published laboratory parameters was performed, and as shown in Table 3 Here, we developed a following nonderivatized approach using UPLC-MS/MS instead of primal derivatized testing. Despite there being only 207 negative controls for nonderivatized detection for the second-tier test, the new cutoffs for MMA and HCY showed signi cantly increased performance (Supplementary Table 2 and Supplementary Figure II). For comparison of the distribution characteristics of MMA, MCA and HCY values before and after modi cation, the Mann-Whitney test was performed. As shown in Fig. 2, there were signi cant differences among the three indexes before and after experimental improvement. First, we compared the 10th, 25th, 75th and 90th percentile MMA, MCA and HCY values. The 75th percentile values of MMA and HCY with primal derivatized detection were similar to the 25th percentile values with the nonderivatized approach, and the difference between the 10th and 90th percentiles was greater than 30%; there were only small differences in the MCA values between the 10th and 90th percentiles for the second-tier test. Taken together, these data show that the distribution of MCA values from the derivatized and nonderivatized tests was similar, which allowed us to conclude that the previous cutoff values could be generalized. In a speci c test of the generalizability of the cutoff values for MMA and HCY, however, the previous cutoff values were not equal to the novel nonderivatized test values.
Comparison of the second-tier test with NBS analysis for newborns suspected to have MMA/PA disorders Thirteen individuals with con rmed MMA/PA disorders were identi ed by second-tier screening from January 2019 to March 2020, and their disease was con rmed by urinary GC/MS and DNA sequencing. As shown in Fig.  1C, the individuals were identi ed with a one-step second-tier test before genetic analysis. The C3 level, C3/C2 ratio and Met level were analyzed in patients grouped based on the initial MS/MS NBS results: the isolated MMA group (n = 1), the Cbl group (n = 6), the PA group (n = 2) and the healthy group (n = 1,830). As shown in Fig. 1A, C3, Met and C3/C2 were generally increased in the three groups compared to the healthy group, while the C3/Met ratio was not signi cantly different. Unfortunately, none of the differences between the three disease groups were signi cant. However, the second-tier UPLC-MS/MS test could directly distinguish these groups, and further statistical analysis of the parameters revealed signi cant differences (Fig. 4B, C). The mass spectrum peaks (shown in Fig. 4) revealed that a solely increased MMA peak likely indicated isolated MMA, an increased MMA peak combined with an increased HCY peak likely indicated Cbl de ciency, an increased MCA peak likely indicated PA, and other patterns likely indicated none of these disorders. Patients who fall within the uncertain category in initial NBS would be differentially screened using a second-tier test, as con rmation via genetic analysis and clinical follow-up is required, which means that this second-tier test can reduce the initial recall rate of patients suspected of having these disorders on NBS. The consideration of MMA, MCA and HCY as markers for MMA/PA would have eliminated 88.97% (234/263) of unnecessary referrals and recall tests.

Discussion
NBS for MMAs/PA is performed by measuring C3 concentrations in DBSs, which is considered a poor discriminator. Although MS/MS-based testing might identify numerous neonates with propionate metabolic disorders, the exact number of false positive cases appears to be affected by the unfavorable balance between screening speci city and sensitivity [15][16][17] . Combined consideration of the C3/C2 ratio and Met improves the PPV, but the false positive rate for C3-associated disorders remains high without further differentiation 18 . This shortcoming necessitates further diagnostic workup and may lead to an unbalanced cost-bene t ratio of NBS, in addition to placing unnecessary emotional burden on the affected families. Here, we expanded a novel secondtier test based on UPLC-MS/MS speci cally for the analysis of three speci c metabolites (MMA, MCA and HCY 19 ), which also provides useful information for differential screening for MMAs/PA, as these metabolites are currently recognized as candidate biomarkers for MMAs and PA. All metabolites and ratios calculated in the initial MS/MS screening or from the second-tier test, if altered, still require urinary GC/MS testing and DNA sequencing to con rm the suspected diagnosis 11 .
Analysis of the regional season pro les of metabolites described in our center revealed that MMA/PA false positive cases were most increased in the summer in Xuzhou, China, during which the heat and humid environment accelerated the degradation of C2 before measurement 20,21 . While the birth prevalence for several disorders is known to vary among different GA groups, the nding of a higher MMA/PA false positive rate in groups of individuals with inadequate duration of pregnancy and low birth weight was surprising 22,23 . MMAs/PA are detected with NBS based on an elevated C3 level and/or C3/C2 ratio. Our rst hypothesis was that summerborn infants might have a naturally lower level of C2 or a slightly higher C3/C2 ratio, which could directly result in an increased rate of MMA/PA false positives compared to that in non-summer-born infants. Another hypothesis was that neonates who are born prematurely might have higher levels of C3 or a higher C3/C2 ratio. The data suggest that preterm infants are very similar with respect to metabolite concentrations regardless of whether they have MMAs/PA; as such, it is di cult to reduce the false positive rate of NBS. However, the current limited results did not support a substantially higher preterm birth rate in summer-born newborns. Thus, further study is needed in a large-scale newborn population to test these hypotheses.
Interestingly, molecular genetic analysis for individuals with MMAs/PA is strongly recommended for cases in which the plasma Met and C3/C2 values are signi cantly lower (as seen with mild alterations with compound heterozygotes compared to homozygotes in cblC de ciency 24 ). Moreover, the detection of C3 is now well recognized as a complicating factor in initial MS/MS screening, primarily because most cases of MMA/PA de ciency may present with elevated C3 concentrations. However, distinguishing features that limit the speci city of MS/MS screening for MMA/PA are lacking 25,26 .
It is worth noting that the utility of the second-tier test (which assesses DBSs via UPLC-MS/MS) highlights the importance of speci c metabolites for differential disorder screening. Second-tier tests have been available for MMA/PA disorder screening to enable comprehensive detection in a time-and cost-effective manner. To facilitate broader application of the second-tier assay, we rst established regional reference intervals for the clinical parameters. In the present study, when the 99.7th percentile MMA, MCA and HCY values (3.13, 0.29 and 9.77 µmol/L, respectively) were used as cutoffs, 234 out of 263 (88.97%) false positive babies from the initial MS/MS screening were excluded without recall testing. An HCY cutoff value of 9.77 µmol/L has been proposed to yield higher sensitivity, and the second-tier test of HCY seems to separate the clbC type from isolated MMAs or PA. The detection of MCA, which is a speci c metabolic analyte for PA disorders, can be expected to in uence the proportion of mixed cases found on initial MS/MS screening. Newborns with elevated MMA alone might also have signi cantly lower MCA and HCY levels than clbC subtype newborns or newborns with PA disorders 27,28 . Overall, following initial MS/MS screening, a novel approach with a second-tier test effectively reduced the false positive rate and avoided unnecessary recall detection.

Conclusion
In summary, MS/MS-based NBS can re ect toxic metabolite alterations in MMA/PA cases as a rst-tier test in neonates 29 , but it results in numerous false positives. However, second-tier UPLC-MS/MS testing reduced the false positive rate by nearly 88.97% without resampling 12,30 . Applying this second-tier test, which enables direct and speci c analyte analysis, could partly avoid false positives and help focus efforts on true positive individuals who require follow-up testing. The second-tier testing strategy enables the detection of a wide range of metabolic disorders with a single laboratory test and signi cantly reduces the incidence of morbidity and mortality in infants. Collectively, the combination of such a second-tier approach with initial MS/MS screening would provide more speci c metabolic information for the physician, while subsequent molecular genetic validation could be implemented for rapid and inexpensive screening to bene t newborns with MMAs/PA and even other disorders.

Con ict of interest statement
We declare that we have no nancial and personal relationships with other people or organizations that can inappropriately in uence our work, there is no professional or other personal interest of any nature or kind in any product, service and/or company that could be construed as in uencing the position presented in the manuscript entitled, 'Newborn screening for methylmalonic acidemia/propionic acidemia: systematic evaluation of a twopronged approach based on MS/MS and UPLC-MS/MS'. Tables   Table 1  Statistical percentile of IVG concentrations