High-fidelity genotyping and therapeutic oligonucleotides analysis by liquid-phase ion trap coupled with mass spectrometry

Abstract

Rapid and accurate nucleic acid analysis is essential for disease diagnostics and drug development, yet persistent metal cation interference severely limits mass spectrometry’s effectiveness. We developed a liquid-phase ion trap (LPIT) system that utilizes strong electric fields to simultaneously dissociate metal contaminants and enrich purified nucleic acid ions directly in solution. This dual-action mechanism simplifies complex mass spectra, achieving up to a 48-fold improvement in mass accuracy and a 100-fold enhancement in signal intensity for intact DNA molecules. The platform’s versatility was demonstrated through high-resolution genetic variation analysis, rapid influenza virus subtyping from raw PCR products, and precise impurity detection in therapeutic drugs. By eliminating laborious pretreatment, the LPIT-ESI-MS system establishes a robust platform for high-fidelity genotyping and the rigorous analysis of therapeutic oligonucleotides, accelerating both clinical diagnostics and biopharmaceutical quality control.

INTRODUCTION

The advent of electrospray ionization (ESI) (1, 2) and matrix-assisted laser desorption ionization (MALDI) (3–5) has positioned mass spectrometry (MS) as an indispensable tool for characterizing a wide array of biomolecules (6–15). The coupling of MS with high performance liquid chromatography (HPLC) has become the standard method in metabolomics (16–19), lipidomics (20–22) and proteomics (23, 24). As one of the most important types of biomolecules (25–27), analysis of nucleic acids by MS is a promising but challenging endeavor (28, 29). The greatest difficulty is the formation of metal cation adducts (30, 31), and a successful MS experiment largely relies on sample preparations. With a polyanionic backbone (32, 33), nucleic acids exhibit strong affinity to metal cations in solution, and the presence of even trace amounts of metal cations would result in abundant adduct species (34). This distribution of the total ion signal across a variety of adducts results in convoluted mass spectra characterized by notably reduced signal intensity and compromised mass accuracy (34–37).

To mitigate these issues, a variety of on-line and off-line desalting strategies have been developed (38–44), including the use of organic bases and chelating agents, precipitation, microdialysis, solid-phase extraction, anion-exchange chromatography (45) and HPLC (46–51). These methods, however, are often laborious and time-consuming, frequently requiring multiple, combined procedures to achieve sufficient salt removal. These advances make MS an increasingly important tool in nucleic acid research (52). Besides mass measurement, polynucleotide sequencing, clinical diagnostics, analyses of non-covalent complexes and chemical modifications have become feasible (53–56).

Despite these advances, the problem is exacerbated for larger nucleic acids, which have a greater propensity for cation adduction and can even be contaminated by ions leaching from container surfaces during analysis (33, 34). Consequently, large nucleic acids are typically analyzed indirectly via digestion or hydrolysis prior to MS analysis (57–59), sacrificing information about the intact molecule. While MALDI-MS has been explored for large nucleic acid analysis, its mass accuracy is often limited to ∼1% (3). As a result, the direct, rapid, and high-accuracy MS analysis of intact nucleic acids remains a notable and unresolved challenge in the field (60).

Recently developed in our lab, LPIT has been utilized to trap and separate amino acids and peptides (61–64). Here, we report the discovery that the strong electric field within an LPIT can rapidly dissociate cation adducts from nucleic acids, allowing the purified ions to be immediately adsorbed and accumulated. Based on these findings, we developed an LPIT-ESI-MS platform that enables the direct, high-fidelity and rapid analysis of intact nucleic acids with substantially improved mass accuracy and sensitivity. By eliminating the use of ion-pair reagents in a high-performance liquid chromatography (HPLC) and single-base extension (SBE) analysis (>2 hrs) (65–68), we demonstrate the power and versatility of this method through several demanding applications, including: resolving single-nucleotide polymorphisms (SNPs) in genes from SARS-CoV-2 and the drug transporter SLCO1B1; subtyping H5N1 and H7N9 influenza from polymerase chain reaction (PCR) products; and identifying subtle impurities in therapeutic oligonucleotides.

RESULTS

Mechanism of electric field-induced cation dissociation and ion trapping

Figure 1A illustrates the schematic of our LPIT-ESI-MS system and its workflow for nucleic acid analysis. Nucleic acids extracted from biological samples (e.g., cells, tissues) can be amplified via PCR. Following a simple purification step, such as solid-phase extraction (SPE), the sample is directly infused into the LPIT and subsequently delivered to the ESI source for MS detection. We first investigated the fundamental mechanisms underpinning this process: electric field-induced cation dissociation and the subsequent adsorption of the purified nucleic acid ions. As depicted in the inset of Fig. 1A, applying a high voltage to the -HV (negative high voltage) metal connector generates a strong electric field within the LPIT region. Inside this field, nucleic acid-metal adducts are subjected to two opposing forces: the drag from the liquid flow and the electrostatic force (Fe). The electric field exerts force on the metal cations and the nucleic acid anion in opposite directions, effectively pulling them apart.

Fig. 1. Integrated workflow and mechanism.

(A) Schematic of the LPIT-ESI-MS platform and experimental workflow for nucleic acid analysis; the zoom-in view illustrates the potential distribution within the liquid-phase ion trap and the mechanisms of nucleic acid-salt adduct dissociation and subsequent ion trapping. (B) The applied waveform on the -HV metal connector. (C) The corresponding EIC of 0.1 μmol/L ssDNA-1 [M-3H]³⁻, inset: mass spectra of ssDNA-1 before turning on the -HV and after turning off the -HV, respectively. (D) Effects of electric field strength (voltage) on the dissociation of metal cations from nucleic acid anions. (E) Trapping efficiencies using GND metal connectors of different sizes.

When this electric field is strong enough, metal cations are dissociated from the nucleic acid backbone (fig. S1). This dissociation exposes additional phosphate groups, which enhances the nucleic acid’s adsorption onto the inner surface of the GND metal connector, effectively trapping it. This adsorption-based mechanism is distinct from our previous work, trapping amino acids and peptides in an LPIT (62, 64). Furthermore, we believe electrolytic reactions at the GND metal connector (anode) may create a localized increase in cation concentration, further promoting the trapping of nucleic acid anions. Once the high voltage is turned off, the trapped, desalted nucleic acid ions are eluted by the liquid flow, ionized by ESI, and detected by the mass spectrometer. Detailed descriptions of the apparatus, chemicals, and materials are provided in the Supplementary Materials.

We first characterized the LPIT’s capabilities in flow injection analysis (FIA) mode by continuously infusing a 0.1 μmol/L solution of a single-stranded DNA (ssDNA-1; details in Table 1) at 2 μL/min. The experiment was divided into three phases (Fig. 1B). In the initial phase (0–5 min), no voltage was applied, and the system functioned as a conventional ESI-MS. The resulting mass spectrum revealed a complex mixture of sodium, potassium, and iron adducts, which originate from the surfaces of containers and fluidic components (Fig. 1C, left inset). During the second phase (5–10 min), a −9000 V potential was applied. The disappearance of the analyte signal confirmed that the ssDNA-1 ions were effectively trapped within the LPIT. Upon switching off the voltage at 10 min, the trapped ions were released, producing a sharp, intense signal. The new mass spectrum (Fig. 1C, right inset) was remarkably clean: the relative intensities of cation adducts were dramatically reduced, and the deprotonated ion ([M-3H]³⁻, m/z = 796.5) became the dominant peak, with its intensity boosted approximately 167-fold. It should be noted that this intensity increase is attributed to two effects: the desalting and accumulation effects.

Table 1. Quantitative analysis of desalination efficiency for ssDNAs.

ssDNA No.	Sequence (5′-3′)	Length (nt)	Formula	Molecular weight (Da)
				Theoretical	Measured	Difference (ppm)
ssDNA-1	att ccg aa	8	C78H99N30O45P7	2392.5	2392.4	−29.0
ssDNA-2	gaa ggt ttt cct tcg	15	C147H187N51O92P14	4573.8	4573.7	−17.3
ssDNA-3	acg ttg gat gga gac taa gac aga aag gg	29	C287H353N127O165P28	9088.6	9088.4	−25.0
ssDNA-4	cct ctc tat ggg cag tcg gtg atc gca gta cga tcg tat cta gca gag ctg cg	53	C516H650N198O319P52	16339.7	16340.9	70.6
ssDNA-5	ctc tct tat cta cat agg ttg ttt aaa gga atc tgg gtc ata cat gtg gat ata tgt gtt cat ggg taa tat gct tcg tgg aat agg gga gac tcc cat agt acc att ggg gct t	115	C1132H1420N422O699P114	35651.9	35653.8	52.2
ssDNA-6	ctc tct tat cta cat agg ttg ttt aaa gga atc tgg gtc ata cat gtg gat ata tgc gtt cat ggg taa tat gct tcg tgg aat agg gga gac tcc cat agt acc att ggg gct t	115	C1131H1419N423O698P114	35636.9	35639.0	58.3
ssDNA-7	acc agg ttg ctg ttc ttt atc agg ttg tta act	33	C324H411N111O204P32	10114.7	10115.7	98.9
ssDNA-8	acc agg ttg ctg ttc ttt atc agg atg tta act	33	C324H410N114O202P32	10123.7	10124.3	59.3
ssDNA-9	acc agg ttg ctg ttc ttt atc agg gtg tta act	33	C324H410N114O203P32	10139.7	10140.6	88.8

To further quantify this desalting effect, the relative fractional abundances of target ions (nucleic ions without metal adducts) were calculated against the total ion signals (table S1) at different charge states. The relative fractional abundance of the [M-3H]³⁻ peak of ssDNA-1 before and after LPIT treatment was 16.1% and 81.2% respectively. However, metal ion adducts are less prevalent at higher charge states, particularly for shorter nucleic acid ions. For ssDNA-1 (8 nt), the relative abundance of target ions increased from 16.1% to 96.4% as the charge state rose from 3- to 6-. In contrast, for the longer ssDNA-3 (29 nt), the relative abundance only increased from 15.6% to 28.3% as the charge state increased from 13- to 16- (table S2). In all cases, LPIT effectively reduced metal adducts. These results suggest that LPIT plays an increasingly critical role in enhancing the signal intensity of target ions as nucleic acid length increases.

To elucidate the underlying mechanism, we further investigated the effects of electric field strength and electrode surface area. We found that the dissociation of metal cations improves with a stronger electric field, and larger nucleic acids require a higher field strength for efficient desalting. For example, an electric field of ∼1.8 × 10⁵ V/m was sufficient for the 8-nucleotide (nt) ssDNA-1, whereas the 53 nt ssDNA-4 required a field of ∼3 × 10⁵ V/m (Fig. 1D). To test our hypothesis that ions are adsorbed onto the GND metal connector, we compared two connectors with different inner surface areas (“Small” with a 0.15 mm bore and “Large” with a 0.25 mm bore). The “Large” connector yielded a ∼143-fold signal improvement, notably higher than the ∼86-fold enhancement from the “Small” connector. In this experiment, 1 μmol/L, ssDNA-1 was tested. This result suggests that a larger surface area facilitates the adsorption and accumulation of nucleic acid ions.

Next, we optimized parameters of the LPIT (figs. S2 and S3). After optimization, a flow rate of 2 μL/min, a -HV with an amplitude of - 9000 V and duration of 3 mins, and the “Large” connector were chosen in later experiments, otherwise specified. As shown in figs. S4 and S5, the LPIT also exhibits good repeatability, and a blank mobile phase between runs is adequate for avoiding carryover effects. Quantitative analyses were performed by using the quantitative loop injection method (fig. S6). Detection sensitivity of ssDNA-8 was characterized (fig. S7), an LOD of 0.1 nmol/L was achieved under the FIA mode, which corresponds to ∼1 fmol in terms of total analyte quantity.

Performance enhancement in mass accuracy and signal intensity

The accurate mass measurement of larger nucleic acids is notoriously difficult due to their stronger affinity for metal cations. Figure 2A shows the conventional ESI-MS mass spectra of three ssDNAs with increasing lengths (15, 29, and 115 nt), revealing that cation adduction worsens as the analyte size increases. For the 29 nt ssDNA-3, metal adducts dominated the spectrum, and for the 115 nt ssDNA-6, the desired deprotonated ion was completely obscured (Fig. 2A). Furthermore, large nucleic acids yield higher charge states observed in the mass spectrum, leading to reduced m/z differences between deprotonated and metal-adducted ions, as well as between adjacent metal adducts. In contrast, LPIT-ESI-MS dramatically reduced the metal adduct peaks for all three ssDNAs (Fig. 2B). The intensities of the target deprotonated ions were enhanced by ∼10-fold in all cases, with the most notable improvement observed for the larger, more problematic ssDNAs. A higher voltage of −15 kV was required for the larger 115 nt fragment to achieve efficient desalting, consistent with our findings in Fig. 1D.

Fig. 2. Desalting efficiency for ssDNAs.

(A) Conventional ESI-MS mass spectra of ssDNA-2 (15 nt), ssDNA-3 (29 nt), and ssDNA-6 (115 nt). The concentration for all three ssDNAs was 1 μmol/L. (B) LPIT-ESI-MS mass spectra for the same three ssDNA fragments. The heat maps show the results of UniDec deconvolution of the respective mass spectra, which separates the charge dimensions into m/z vs. charge matrices. Note: −15 kV was applied for ssDNA-6.

This spectral simplification has a profound impact on data analysis. Deconvolution heat maps generated by UniDec software visualize the charge state distribution and reveal how spectral complexity escalates with nucleic acid length in conventional ESI-MS (69), making mass determination difficult. Our LPIT-ESI-MS method effectively minimizes these adducts, resulting in cleaner data. This directly translates to a notable improvement in mass accuracy for long nucleic acid analyses. For smaller oligonucleotides such as ssDNA-2 and ssDNA-3, both conventional ESI-MS and LPIT-ESI-MS are capable of detecting the deprotonated ions. However, with LPIT-ESI-MS, the mass error for ssDNA-6 ([M-42H]⁴²⁻) was 0.0566 Th (∼66 ppm), a result primarily limited by the inherent accuracy of the mass analyzer. In contrast, conventional ESI-MS had much larger mass errors of 2.7027 Th (∼3189 ppm). This represents a 48-fold enhancement in mass accuracy for the 115-nt nucleic acid. While a Q-TOF instrument was primarily used in this study, additional experiments conducted with a high-resolution Orbitrap mass spectrometer (Q Exactive HF-X) at a resolving power of 240,000 further illustrate the necessity of the LPIT system. Even with high-resolution instrumentation, target ion peaks for longer nucleic acids (e.g., 53 nt and 115 nt) remain unresolved or heavily obscured by a dense distribution of metal adducts (figs. S8 and S9). These results confirm that chemical heterogeneity, rather than instrumental resolution alone, is the primary bottleneck for large nucleic acid analysis, underscoring the critical role of LPIT in enabling the identification of these larger fragments.

High-resolution genotyping of single-nucleotide polymorphisms

As the most common genetic variation, SNP detections are important issues in disease diagnosis, virus classification, as well as in fundamental biology (70–73). We applied LPIT-ESI-MS to genotype the pharmacogenetically notable SLCO1B1 (T521C) polymorphism using ssDNA-5 and -6 (Table 1). This variant influences simvastatin and cerivastatin metabolism, and identification of the 521C allele is critical as it necessitates careful dosing to mitigate the risk of myocardial disease. The top, middle and bottom of Fig. 3 plot the mass spectra of ssDNA-5, −6, and a mixture of them, in which the dominant peaks ([M-42H]⁴²⁻) are also displayed in the zoom-in views. As the m/z differences between deprotonated nucleic acid and its metal adduct ions become smaller for larger ions, these peaks are not resolved when using ESI-MS (black lines). Furthermore, the 521T and 521C were also not able to be distinguished in conventional ESI-MS mass spectra. The red lines in Fig. 3 are the mass spectra acquired using the LPIT-ESI-MS. The m/z difference between the [M-42H]⁴²⁻ ions of 521T and 521C is merely 0.3520 Th. While indistinguishable by conventional ESI, the mixture is clearly resolved after LPIT-ESI-MS treatment. The corresponding deconvolution results (Fig. 3, right) further confirm the successful identification of both species. Based on the measured intact masses, the mass accuracies for 521T and 521C are 52.2 ppm and 58.3 ppm, respectively, as detailed in Table 1.

Fig. 3. SNP analysis of SLCO1B1.

Mass spectra (left) and corresponding deconvolution results (right) for ssDNA-5 (top), ssDNA-6 (middle), and their mixture (bottom). Black lines: collected using conventional ESI-MS; red lines: collected using LPIT-ESI-MS.

In addition, we further evaluated the capability of LPIT-ESI-MS for the direct analysis of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) variants (fig. S10). ssDNA-8 and -9 correspond to the wild type and the D614G mutant of SARS-CoV-2 spike protein (74), respectively. To rigorously challenge the system’s resolution, we also included a third variant (ssDNA-7) containing an A > T substitution. This represents the smallest mass difference between natural nucleobases in DNA (9 Da). We then compared the performance of our LPIT-ESI-QTOF system against conventional MALDI-TOF MS. As illustrated in fig. S10, the LPIT-ESI-MS method successfully resolved all three components in a mixture, clearly distinguishing both the 9 Da and 16 Da mass differences. The system demonstrated superior mass fidelity with errors of approximately 100 ppm, whereas MALDI-TOF failed to resolve the mixture and exhibited notably higher errors (∼336 ppm). Notably, the entire LPIT-based SNP analysis can be completed within 20 minutes, offering a substantial time advantage over the traditional 2-hour SBE-based techniques.

Direct subtyping of influenza virus PCR products

Beyond synthetic oligonucleotides, we tested the system’s ability to analyze large, biologically relevant PCR products for influenza virus subtyping. Precise molecular weight (Mw) determination is crucial for monitoring mutations in highly pathogenic avian influenza (HPAI) strains like H5N1 and H7N9. We analyzed two PCR products targeting the hemagglutinin (HA) gene: an 82-bp H5N1 fragment and an 85-bp H7N9 fragment, using primers specified in the Chinese national standard GB/T 18936–2020. After SPE desalting, samples were injected into the LPIT for analysis. To highlight the advantages of our approach, we compared the analysis of PCR products using conventional ESI-MS and LPIT-ESI-MS. While target peaks were undetectable in conventional ESI-MS, LPIT-ESI-MS successfully captured high-quality signals for both samples (figs. S11 and S12).

As shown in the left part of Fig. 4, LPIT-ESI-MS allowed for the simultaneous detection of both the PCR product strands and the unconsumed primers in a single spectrum, eliminating the need for a separate primer removal step. After deconvolution of the charge state envelope, we obtained precise Mw values for both strands of H5N1 (25,079.2 Da and 26,094.0 Da) and H7N9 (26,698.2 Da and 26,327.9 Da) (Fig. 4, right), the mass errors for each sense and antisense strand summarized in Table 2. The platform’s high resolution was further confirmed by the detection of minor peaks corresponding to strands with an additional deoxyadenosine monophosphate (+dAMP) (75), a common polymerase artifact. Resolving single-nucleotide differences directly within large PCR products enables rapid and reliable screening for infectious disease diagnostics.

Fig. 4. Identification of avian influenza subtypes.

Mass spectra obtained by LPIT-ESI-MS of single chains from influenza H7N9 (top) and H5N1 (bottom). Right subpanels: Deconvoluted spectra showing molecular weights of each strand. The +A strand represents the 3′ A-tailing product generated by Taq polymerase. The green and purple represent the sense (5′ → 3′) and antisense (3′ → 5′) strands of H7N9, respectively; blue and red represent the sense (5′ → 3′) and antisense (3′ → 5′) strands of H5N1, respectively.

Table 2. DNAs of avian influenza virus subtypes synthesized by PCR.

Name	Sequence	Length (nt)	Formula	A-tailing Molecular weight (Da)
				Theoretical	Measured	Difference(ppm)
H7N9	5′-3′ cta att gat ggt tgg tat ggt ttc aga cac cag aat gca cag gga gag gga act gct gca gat tac aaa agc act caa tcg gca a	85	C844H1051N344O503P85	26694.5	26698.2	138.6
	3′-5′ gat taa cta cca acc ata cca aag tct gtg gtc tta cgt gtc cct ctc cct tga cga cgt cta atg ttt tcg tga gtt agc cgt t	85	C837H1061N300O523P85	26324.4	26327.9	133.0
H5N1	5′-3′ agg gag gat ggc agg gaa tgg ttg atg gtt ggt atg ggt acc atc ata gca atg agc agg gga gtg ggt acg ctg cag aca a	82	C820H1013N347O489P82	26093.4	26094.0	23.0
	3′-5′ tcc ctc cta ccg tcc ctt acc aac tac caa cca tac cca tgg tag tat cgt tac tcg tcc cct cac cca tgc gac gtc tgt tt	82	C796H1019N281O501P82	25078.2	25079.2	39.9

Quality control and impurity detection in therapeutic oligonucleotides

Finally, this platform was applied to tackle a critical challenge in pharmaceutical analysis: the quality control of therapeutic oligonucleotides. These emerging drugs are commonly produced by solid-phase synthesis (76), which can introduce impurities such as chain-length variants (e.g., n-1 deletions) and incomplete backbone modifications. Mipomersen (Kynamro; Ionis Pharmaceuticals), an oligonucleotide therapeutic for homozygous familial hypercholesterolemia (77, 78), was employed as a model to evaluate the technique’s ability to characterize synthesis-related impurities, such as chain length variants and phosphate modifications. A total of five oligonucleotide fragments were synthesized to represent both modification and impurity types. These fragments include a dual-modified fragment with methoxyethyl and phosphorothioate (PS) backbone modifications (fully PS-modified fragment), a modified fragment with only PS modifications, fragments containing one or two phosphodiester (PO) bond impurities (relative to the fully PS-modified counterpart), and a fragment one nucleotide shorter than the fully PS-modified reference. Figure 5A presents the sequences of the five oligonucleotides.

Fig. 5. Analysis of therapeutic oligonucleotide mixtures and synthesis-related impurities.

(A) Chemical structures of representative modified nucleotides and sequences of the five synthesized oligonucleotide fragments. (B) LPIT-ESI-MS analysis of a mixture containing the five oligonucleotide fragments (insets: enlarged ESI views; right: deconvoluted spectra). (C) Identification of 1% and 5% PO impurities in PS oligonucleotides via LPIT-ESI-MS.

To evaluate the practical utility of the LPIT-ESI-MS method in identifying oligonucleotide therapeutics and their associated impurities, we combined the 5 synthetic oligonucleotide fragments for analysis using LPIT-ESI-MS, demonstrating that each fragment can be simultaneously identified with exact molecular weight following deconvolution (Fig. 5B). Additionally, we spiked the fully PS-modified oligonucleotide with 1% and 5% of the PO impurity (fragment with one phosphodiester bond), respectively, which was also successfully identified by LPIT-ESI-MS (Fig. 5C). This showcases the method’s capability for low-level impurity detection. This high level of discrimination is essential for drug development and demonstrates the notable potential of LPIT-ESI-MS for ensuring the purity and safety of oligonucleotide-based therapeutics.

DISCUSSION

In this work, we introduced a LPIT-ESI-MS platform that overcomes the long-standing challenge of high-fidelity analysis of intact oligonucleotides. We discovered that a strong electric field within the LPIT simultaneously dissociates metal adducts while trapping and enriching the purified nucleic acid ions. This dual-action mechanism dramatically simplifies mass spectra, and leads to enhancements in both signal intensity (up to 100-fold) and mass accuracy (up to 48-fold). This enables the direct, rapid and high-fidelity analysis of intact nucleic acids, from short oligonucleotides to long PCR products. The versatility and capability of this platform were demonstrated through several demanding applications, including high-resolution SNP genotyping (such as the most challenging A-T transition), direct subtyping of influenza viruses from PCR products without single-base extension, and the precise identification of subtle impurities in therapeutic oligonucleotides.

Both the aqueous-acetonitrile buffer system and the flow rate used in LPIT are compatible with micro- and nano-HPLC configurations. This suggests that LPIT could potentially be coupled with HPLC to improve separation and desalting performance. It should be noted that volatile salts widely used in nucleic acid MS, such as ammonium acetate, are generally incompatible with the current LPIT setup. Under the strong electric field inherent to LPIT, even low concentrations of salt (5 mM ammonium acetate, for example) can induce excessive Joule heating and subsequent gas outgassing, which prevents the maintenance of a stable ion trap. Nevertheless, further optimization of the synchronization between the two systems is required—a challenge analogous to that encountered in implementing two-dimensional liquid chromatography. By removing a fundamental barrier in nucleic acid MS, the LPIT platform opens the door to routine, high-throughput characterization of complex RNA structures, nucleic acid-protein interactions, and the next generation of oligonucleotide-based therapeutics.

MATERIALS AND METHODS

Chemicals and materials

Single-stranded DNAs (ssDNA) listed in Table 1 were synthesized by Sangon (Sangon Biotechnology Co., Ltd., Shanghai, China). HPLC grade acetonitrile (ACN) was obtained from Fisher Chem. Alert Co. (NJ, USA). Deionized water was purchased from Watson (Hong Kong, China). All these reagents were used without any further purification. Sample solutions were prepared in water as stock solutions at the concentration of 100 μmol/L and diluted to final concentrations with a running buffer of ACN/water (1:1 v/v)

PCR Amplification and Sample Pretreatment

The double-stranded DNA (dsDNA) templates, as listed in Table 2, were synthesized in accordance with the GB/T 18936–2020 standard. PCR amplification was performed using Takara Taq polymerase (TaKaRa Taq, RO01A) and 1.5 mM MgCl₂. The thermal cycling conditions were as follows: initial denaturation at 95°C for 2 min; 45 cycles of denaturation at 95°C for 30 s, annealing at 60°C for 30 s, and extension at 72°C for 40 s; followed by a final extension at 72°C for 6 min. Specific primer sequences for H5N1 and H7N9 are detailed in fig. S13. To disrupt the hydrogen bonding between the strands prior to mass spectrometry analysis, all PCR products underwent a denaturation step in a 95°C water bath for 5 minutes.

LPIT-ESI-MS Instrument Setup

The LPIT system consists of three parts: a sample injection part, a capillary fluid channel, stainless steel connectors for electrodes, and a mass spectrometer as the detection system. In the flow injection analysis (FIA) mode, samples loaded in a 0.5 mL syringe were injected directly into the liquid channel by a syringe pump (Harvard, No.704500), which also provides a constant liquid flow rate for the liquid-phase ion trap. It should be noted that a six-way microvalve (Agilent 1200 series 2 position/6 port, No. G1162A) could also be used for sample introduction. Samples first flow through an inline MicroFilter kit (Agilent, No. 5067–1582) to remove microparticles and air bubbles in the channel. PEEKsil capillaries (I.D. 150 μm, O.D. 1/32 inch) were used as the liquid channels in experiments. Capillary length between the MicroFilter and the first stainless steel metal connector (VICI, No. ZU1C) was 50 cm, and 5 cm from the first to the GND metal connector. The liquid-phase ion trap utilizes stainless steel unions (300-series; VICI Valco Instruments, Houston, TX, USA) as electrodes. The “Small” connector (P/N: ZU1XC) has a 0.15 mm bore, while the “Large” connector (P/N: ZU1C) has a 0.25 mm bore. Both are compatible with 1/16-inch O.D. PEEKsil capillaries. A tunable negative high voltage (-HV) was applied on the -HV metal connector, which was provided by a high voltage generator module (PRB15R8M1LR2RT2LC1, Wisman Inc. Xianyang, China). The GND metal connector was grounded during experiments. An ESI source was used to connect the liquid-phase ion trap with the following mass spectrometer (Agilent G6520B Accurate-Mass Q-ToF). By default, it was directly connected with the GND connector, otherwise specified.

Mass Spectrometry Parameters

The following MS parameters of Agilent G6520B were used in experiments, data acquisition: negative mode, ionization voltage: 4000 V, gas temperature: 250°C, gas flow rate: 8.0 L/min, fragmentor: 120 V, mass range (m/z): 300–3000. In the SNP detection part, the MALDI-TOF mass spectrometry was Bruker autoflex maX, operated in linear, positive ion mode (Ionization Mode LD+). Spectra were acquired with an acceleration voltage of 19.5 kV over a mass range of m/z 3000–20,000. Delayed extraction (DE) was used with a delay time of 340 ns. The laser energy was set to 30% above the threshold, and 10,000 laser shots were accumulated per spectrum. Mass spectrometry analysis was performed on the Q Exactive HF-X Orbitrap mass spectrometer (Thermo Fisher Scientific) in negative ion mode. The key parameters were as follows: Scan Parameters: Full MS scans were acquired at a resolution of 240,000 over the range of m/z 300–1,500. The AGC target was 3 × 10⁶ with a maximum injection time of 50 ms. In-source CID was set to 20.0 eV, and each spectrum was an average of 5 microscans. ESI Source: The spray voltage was 3.00 kV, and the capillary temperature was 320°C. The flow rates for sheath gas and auxiliary gas were 5 and 1 (arb. units), respectively. The funnel RF level was maintained at 55.0.

Supplementary Materials

This PDF file includes:

Supplementary Text

Figs. S1 to S13

Tables S1 and S2