Design of M13 phage systems dual-displaying two functional proteins for developing a rapid biosensing probe Authors: Qiang Meng, Huimin Zhang, Ting Liu, Ziyue Yuan, Meiling Guo, Wei Song, Qinglin Sheng, Yane Luo, Tianli Yue Published: 2026 ABSTRACT M13 phage dual-display systems, which co-present targeting and reporter elements, hold great promise for developing rapid biosensors. However, the efficiency and performance of different display-site combinations remain poorly characterized. This study systematically evaluates the dual-display capacity of M13 phage by co-displaying a Vibrio parahaemolyticus-specific nanobody (Nb20) on pIII protein and NanoLuc luciferase (Nluc) on either pVI, pVII, pVIII, or pIX capsid proteins under both monovalent and multivalent strategies. We found that display efficiency varied significantly with the coat protein site and infection time of helper phage. The pVIII-based monovalent display achieved the highest specific luminescent activity and a functional co-display proportion of up to 2.20%. Furthermore, the optimized mpVIII probe demonstrated high specificity, sensitivity (detection limit of 102 cfu/mL), and robust anti-interference capability in detecting V. parahaemolyticus in spiked shrimp samples. This work provides fundamental insights and a practical framework for constructing high-performance dual-display phage probes with one-step biosensing.Graphical abstract Supplementary InformationThe online version contains supplementary material available at 10.1186/s12951-026-04673-y. Introduction The advent of phage display technology has revolutionized the fields of biomolecular engineering and synthetic biology [1, 2]. This technique enables the fusion of exogenous peptides or proteins to phage coat proteins, facilitating their presentation on the viral particle surface and thereby directly linking genotype with phenotype [3–6]. Among various platforms, the filamentous M13 bacteriophage stands as a particularly versatile vector due to its non-lytic life cycle, capacity for accommodating large foreign inserts, and remarkable structural robustness, making it widely applicable in antibody screening, vaccine development, and biosensing research [4, 7–11]. Conventional M13 phage display systems have primarily focused on the presentation of single proteins [4, 12–14]. However, with the increasing complexity of diagnostic and targeted therapeutic demands, dual-display phages that can present two distinct functional elements simultaneously hold significant potential [15, 16]. For instance, an ideal biosensing probe could integrate both a targeting element (e.g., an antibody) and a signal-reporting element (e.g., luciferase, fluorescent protein, or luminescent inorganic material), enabling “one-step” detection of analytes [17]. Current dual-display strategies for M13 phage predominantly rely on a “pIII monovalent display combined with pVIII monovalent or multivalent display” approach [9, 18–22]. In this configuration, the minor coat protein pIII is typically used to present the targeting moiety, such as a nanobody, while the major coat protein pVIII displays a short peptide designed for conjugating the signal reporter. For example, Bae et al. utilized a phagemid system (pTV118N) to display an EGFR-specific repebody on protein pIII, while presenting the SpyTag peptide on protein pVIII [22]. Subsequent incubation with eGFP-SpyCatcher facilitated covalent coupling to the phage surface via spontaneous SpyTag-SpyCatcher interaction, thereby equipping the phage with a reporter protein [22]. A major limitation of this strategy is the requirement for additional in vitro modification steps, which involve cumbersome purification procedures for the signal proteins used in the conjugation process. Most importantly, the efficient and controlled co-display of two functional macromolecular proteins on a single phage particle remains a formidable bioengineering challenge, being mainly attributable to the limited display capacity of M13 coat proteins [9, 23]. Among the five coat proteins of M13 phage, pVI (located at the tail) and pVII/pIX (located at the head) are generally suitable only for presenting short peptides within 20 amino acids, especially in multivalent display contexts [22]. While pIII is recognized for its ability to accommodate larger peptides and even full-length proteins (typically 100–150 amino acids) [9]. The major coat protein pVIII, due to structural constraints during assembly, usually permits the display of only small peptides spanning 5–10 amino acids [9, 24]. This constrained display capacity of M13 phage thus restricts its application to small heterologous proteins. However, the amino acid lengths of most reporter proteins far exceed the permissible display size of M13 [23]. Therefore, a compact, highly active reporter protein is essential for achieving efficient dual display on M13 phage. Although researchers have reported single-display characteristics of various M13 coat protein sites (e.g., the major protein pVIII and minor proteins pIII, pVI, pVII, pIX) [1], systematic comparative analyses of dual-display systems—particularly the impact of different site combinations on dual-display efficiency, protein functionality, and ultimate detection performance—remain insufficient. Moreover, to enhance signal intensity, researchers have attempted to integrate the reporter protein gene into the helper phage genome for multivalent display [20]. Nevertheless, whether this strategy invariably leads to signal amplification, and the potential effects of multivalent fusion on reporter protein activity and phage assembly stability, require further in-depth investigation. To address these challenges, this study aims to systematically evaluate the capacity of M13 phage for dual display and to develop a high-performance rapid detection probe. We selected a nanobody (Nb20) targeting V. parahaemolyticus (a food-borne pathogen) and the highly sensitive reporter protein NanoLuc luciferase (Nluc, 176 amino acids) as model functional proteins. By constructing a series of recombinant phages, we systematically compared four different coat protein sites (pVI, pVII, pVIII, pIX) under both monovalent and multivalent display strategies. We quantitatively evaluated key performance metrics, including phage yield, specific luminescent activity, and the proportion of functional particles. We further investigated the regulatory roles of host physiological state and induction conditions on display efficiency. Ultimately, the optimal probe was selected and validated for its specificity, sensitivity, and anti-interference capability in detecting V. parahaemolyticus within complex real samples. This work not only provides fundamental insights and practical strategies for constructing efficient dual-display phages but also accelerates the application of phage dual-display technology in the field of rapid biosensing. Experimental section Bacterial culture and phage propagation conditions All bacterial strains utilized in this study are listed in Table S1. S. enteritidis serotype 3, S. enteritidis serotype 4, and S. enteritidis serotype 21 were gifts from Professor Baowei Yang [25]. V. parahaemolyticus was purchased from China Center of Industrial Culture Collection (CICC). V. alginolyticus, V. harveyi, Aeromonas hydrophila, and Listeria monocytogenes were purchased from Warner Biology (China). Escherichia coli ER2738, E. coli TG1, E. coli BL21(DE3), and their recombinant derivatives were cultured in Luria−Bertani (LB) medium, supplemented with appropriate antibiotics (ampicillin 100 µg/ml, kanamycin 50 µg/ml) and the inducer IPTG (1 mmol/l) where necessary. Monovalent double-displayed recombinant phages were produced and harvested as follows: Glycerol stocks of recombinant E. coli ER2738 harboring pComb3X-Nb20-gVI-Nluc, pComb3X-Nb20-Nluc-gVII, pComb3X-Nb20-Nluc-gVIII, or pComb3X-Nb20-Nluc-gIX were inoculated (1% v/v) into LB broth and incubated overnight at 37 °C with shaking at 180 rpm. One milliliter of this starter culture was then transferred into 100 ml of fresh LB broth and grown under the same conditions until the OD600 reached 0.2, 0.4, 0.6, or 0.8. Subsequently, 100 µl of helper phage M13KO7 (1012 pfu/ml) was added. Following gentle mixing and a 1-hour static infection period at 37 °C, the culture was shaken for 12 hours. The culture was centrifuged at 10,000 × g for 10 min, and the supernatant was transferred to a new tube. This centrifugation step was repeated twice to ensure complete removal of bacterial cells. A sterile 20% of PEG8000 solution (v/v, containing 2.5 mol/l of NaCl), equivalent to one-quarter of the supernatant volume, was added, mixed thoroughly, and incubated to precipitate phage particles. The mixture was centrifuged at 12,000 × g for 15 min. The resulting phage pellet was resuspended in 1 ml of TBS buffer (18 mmol/l of Tris, 150 mmol/l of NaCl, pH 7.4). After a final centrifugation at 12,000 × g for 15 min to remove any residual cellular debris, the supernatant containing the monovalent double-displayed recombinant phage was collected. For the production of multivalent double-displayed recombinant phages, glycerol stocks of recombinant E. coli TG1/M13KO7-gVI-Nluc, TG1/M13KO7-Nluc-gVII, or TG1/M13KO7-Nluc-gIX were inoculated (1% v/v) into LB broth and grown overnight at 37 °C with shaking at 180 rpm. Recombinant phages M13KO7-p(Nluc-pVII), M13KO7-p(pVI-Nluc), and M13KO7-p(Nluc-pIX) were isolated from the cultures using the procedure described above. Separately, E. coli ER2738/pComb3X-Nb20 was grown to an OD600 of 0.2, 0.4, 0.6, or 0.8. To this culture, 100 µl of the harvested recombinant helper phage (M13KO7-p(pVI-Nluc), M13KO7-p(Nluc-pVII), or M13KO7-p(Nluc-pIX), 1012 pfu/ml) was added. After a 1-hour static infection at 37 °C, the culture was shaken for 12 hours. Phages were purified as previously outlined, yielding the multivalent double-displayed recombinant phage suspensions. Plasmid design and molecular cloning All recombinant plasmids (Table S1) were constructed using standard Gibson assembly methods [26]. All primers used are listed in Table S2. Gene sequences for gVI, gVII, gVIII, and gIX were obtained from the M13mp18 plasmid map (https://www.snapgene.com/plasmids/basic_cloning_vectors/M13mp18). The Nluc gene (PDB accession 7SNS), gVI-Nluc, Nluc-gVII, Nluc-gVIII, Nluc-gIX, and Nb20 genes were codon-optimized and synthesized commercially (Tsingke Biotechnology). To generate phagemids for monovalent double-display, the backbone vector pComb3X was used to construct pComb3X-Nb20-gVI-Nluc, pComb3X-Nb20-Nluc-gVII, pComb3X-Nb20-Nluc-gVIII, and pComb3X-Nb20-Nluc-gIX. Initially, linearized pComb3X vector and the Nb20 gene fragment were amplified using primer pairs pComb3X-F/pComb3X-R and Nb20-F/Nb20-R, respectively. These fragments were assembled using the ClonExpress II One Step Cloning Kit (C112-01, Vazyme) to create pComb3X-Nb20, which was then transformed into E. coli ER2738 to generate ER2738/pComb3X-Nb20. Subsequently, linearized pComb3X-Nb20 vector and the fusion gene fragments (gVI-Nluc, Nluc-gVII, Nluc-gVIII, Nluc-gIX) were amplified using specific primer pairs (pComb3X-Nb20-F/pComb3X-Nb20-R, pVI-Nluc-F/pVI-Nluc-R, Nluc-pVII-F/Nluc-pVII-R, Nluc-pVII-F/Nluc-pVIII-R, Nluc-pVII-F/Nluc-pIX-R, respectively). Each linear vector was ligated with its corresponding insert using the ClonExpress II One Step Cloning Kit, resulting in the final phagemids. These phagemids were transformed into E. coli ER2738 to produce the respective recombinant strains. For constructing multivalent double-display phage vectors, the helper phage vector M13KO7 was engineered to create M13KO7-gVI-Nluc, M13KO7-Nluc-gVII, and M13KO7-Nluc-gIX. Linearized M13KO7 fragments (targeting genes gVI, gVII, and gIX) and the Nluc gene fragment were amplified using primers M13KO7-6-F/M13KO7-6-R, M13KO7-7-F/M13KO7-7-R, M13KO7-9-F/M13KO7-9-R, and Nluc-F/Nluc-R, respectively. The linearized M13KO7 vectors were individually ligated with the Nluc fragment using the ClonExpress II One Step Cloning Kit. The resulting plasmids were transformed into E. coli TG1. An expression plasmid for Nluc, pET28a-Nluc, was constructed using the pET-28a vector. Linearized pET-28a and the Nluc gene fragment, amplified with primers pET-28a-F/pET-28a-R and Nluc-28a-F/Nluc-28a-R, were ligated. This plasmid was transformed into E. coli BL21(DE3) to generate the expression strain BL21(DE3)/pET-28a-Nluc. All constructed plasmids were verified by Sanger sequencing (Tsingke Biotechnology). Nluc overexpression and purification An overnight culture of E. coli BL21(DE3)/pET-28a-Nluc was diluted 1:100 (v/v) into fresh LB medium and grown at 30 °C with shaking at 150 rpm until the OD600 reached 0.5. Protein expression was induced by adding IPTG to a final concentration of 1 mmol/l, followed by continued incubation for 6 h. Cells were harvested by centrifugation at 3,000 × g for 20 min at 4 °C. The pellet was resuspended in 1/20 of the original culture volume of binding buffer (50 mmol/l of KH₂PO₄, 300 mmol/l of NaCl, pH 8) and subjected to ultrasonic disruption (240 W, 10 s pulse, 20 s pause, 20 cycles). The soluble Nluc protein was purified from the lysate supernatant using BeyoMag™ His-Tag Purification Agarose Magnetic Beads (P2243-2 ml, Beyotime) according to the manufacturer’s instructions. The eluted Nluc protein was dialyzed to remove impurities and ions. Protein concentration was determined using the BCA assay. A standard curve correlating Nluc concentration with chemiluminescence signal was established using serially diluted purified protein (Fig. S3G). Phage titer quantification Displaying heterologous proteins on the phage coat can potentially impair its infectivity [27, 28]. To circumvent any bias introduced by variable infectivity across different fusion constructs, we used a qPCR-based method to quantify phage particles, consistent with previous studies [29, 30]. Helper phage M13KO7 titer was determined using the standard plaque assay [31]. For phage single-stranded DNA (ssDNA) extraction (alkaline lysis method) [32], 10 µl of phage suspension was mixed with 115 µl of sterile water. Then, 125 µl of Buffer P2 (DP103-02, TIANGEN) was added, mixed gently, and incubated for 1 min. Subsequently, 175 µl of Buffer P3 (DP103-02, TIANGEN) was added, mixed by inversion, and centrifuged at 12,000 × g for 15 min. The supernatant was transferred to a new tube, and the ssDNA was purified from the supernatant using the standard ethanol precipitation method. A standard curve relating M13KO7 titer to quantitative PCR (qPCR) cycle threshold (Cq) values was established. Known amounts of M13KO7 ssDNA were serially diluted and used as templates in qPCR reactions with primers q-pIII-F/q-pIII-R. The Cq values were plotted against the corresponding phage titers (Fig. S3A). The titer of recombinant phage preparations was determined by extracting their ssDNA, measuring the Cq value using the same qPCR primers and conditions, and extrapolating from the standard curve. Quantification of phage luminescence For detecting NDP signal, 100 µl of phage suspension was mixed with 100 µl of the Nano-Glo® Luciferase Assay System substrate (N1120, Promega). After incubation at 37 °C for 5 min, chemiluminescence was measured using a SpectraMax i3x microplate reader (Molecular Devices, San Jose, CA, USA) in white opaque plates. Each sample was assayed in triplicate. Luminescence signals were measured across a dilution series of the phage suspension. Based on the linear relationship between luminescence signal and dilution factor (Fig. S3B), the signals were normalized to a titer of 1012 pfu/ml. The signal from CoDP was assessed using a capture assay. Sterile 1.5 ml microcentrifuge tubes were blocked overnight at 4 °C with 1 ml of freshly prepared 3% (w/v) skim milk. After blocking, tubes were washed twice with 1.4 ml of PBS (pH 7.4). A 100 µl aliquot of fresh V. parahaemolyticus suspension (109 cfu/ml) was added to each tube, followed by 100 µl of the respective recombinant phage suspension. Tubes were incubated at 37 °C for 2 h. Unbound phages were removed by centrifugation at 10,000 × g for 10 min, and the pellet was washed twice gently with 1 ml of PBS. The final pellet was resuspended in 100 µl of TBS. Chemiluminescence was measured as described above after adding 100 µl of the Nano-Glo® Luciferase Assay System substrate (test signal). Background signal was determined by replacing the V. parahaemolyticus suspension with 100 µl of PBS and processing identically. The specific CoDP signal was calculated by subtracting the background signal from the test signal. For the four monovalent double-displayed phages, it was assumed that each phage particle displays a maximum of one copy of either Nluc or Nb20 when calculating the percentages of NDP and CoDP. For the three multivalent double-displayed phages, calculations assumed the presence of five copies of Nluc and a maximum of one copy of Nb20 per phage particle. Western blotting Equal volumes (10 µl) of phage suspensions were separated by Tricine-SDS-PAGE (G2159, Servicebio). Proteins were subsequently transferred onto a PVDF membrane using a wet transfer system [33]. The membrane was blocked with protein-free blocking buffer for 30 min at room temperature to prevent non-specific binding. After blocking, the membrane was washed three times for 8 min each with 1× TBST (G0004-1 L, Servicebio). It was then incubated overnight at 4 °C with gentle shaking in a solution containing a mouse monoclonal primary antibody against Nluc (N7000, Promega). Following three TBST washes, the membrane was incubated for 2 h at room temperature with an HRP-conjugated goat anti-mouse IgG secondary antibody (GB23301, Servicebio). After another three TBST washes, proteins were detected using an ECL substrate (G2020-50ML, Servicebio) and visualized with an automated chemiluminescence imaging system (Fusion FX6 Edge, Vilber, Collegien, France). To control for loading, the same membrane was subsequently probed for the major coat protein pVIII using a mouse monoclonal anti-pVIII antibody (ARA828, Arex Biosciences) following the same protocol. Acquired images were quantified using Fiji (ImageJ) software [34]. The relative display level of Nluc was calculated by normalizing the Nluc signal to the pVIII signal for each sample. The relative display level for the ppIX sample (harvested at OD600 = 0.6) was set to 1 for normalization across all samples. Absolute quantification of Nluc mRNA A standard curve relating the band intensity on an agarose gel to DNA mass concentration was established. Different volumes of a standard DNA Marker (AG11905, Accurate Biology) were electrophoresed, and the grayscale value of the 200 bp band was plotted against its mass concentration (Fig. S3E). A 167 bp Nluc gene fragment (Nluc167) was amplified via PCR using mpVI ssDNA as the template and primers q-Nluc-F/q-Nluc-R. The PCR product was purified from an agarose gel using the Hipure Gel DNA Mini Kit (D211102, Magen). The grayscale value of the purified Nluc167 fragment was measured from the gel image (Fig. S3C), and its mass concentration was determined using the standard curve. The mass concentration of Nluc167 was converted to copy number concentration using the formula: C (copies/µl) = [ρ (ng/µl) × 10³ × 10⁻¹² × 6.02 × 10²³] / (660 × 167), where ρ is the mass concentration, 660 g/mol is the average mass of a DNA base pair, and 167 is the fragment length in base pairs. A standard curve relating Cq value to Nluc167 copy number concentration was generated by performing qPCR with serially diluted Nluc167 fragments of known copy number, using primers q-Nluc-F/q-Nluc-R and SYBR Green Premix Pro Taq HS qPCR Kits (AG11701, Accurate Biology) on a LightCycler® 480 Instrument II (Roche Diagnostics, Basel, Switzerland) (Fig. S3F). Total RNA was extracted from harvested bacterial cell pellets using the Trizol method [35]. RNA purity and concentration were assessed by agarose gel electrophoresis and spectrophotometry. cDNA was synthesized from total RNA using the PrimeScript™ RT Master Mix (RR036A, Takara). Absolute quantification of Nluc mRNA in the cDNA samples was performed by qPCR using the primers q-Nluc-F/q-Nluc-R and the SYBR Green Premix Pro Taq HS qPCR Kits on the LightCycler® 480 Instrument II system. The copy number of Nluc mRNA in each sample was calculated based on the standard curve derived from the Nluc167 fragment dilutions. Relative quantification of Nluc mRNA The recombinant E. coli strains ER2738/pComb3X-Nb20-gVI-Nluc, ER2738/pComb3X-Nb20-Nluc-gVII, ER2738/pComb3X-Nb20-Nluc-gVIII, and ER2738/pComb3X-Nb20-Nluc-gIX were cultured under non‑induced and induced conditions, respectively. When the OD600 reached 0.6, 1 mL of each bacterial culture was collected by centrifugation at 8,000 × g and 4℃ for 2 min. After discarding the supernatant, the bacterial pellets were subjected to RNA extraction and reverse transcription using the method described in Sect. ""Absolute quantification of Nluc mRNA"". Using 16 S rDNA as an internal reference gene and setting the transcriptional level of the Nluc gene in non‑induced E. coli ER2738/pComb3X-Nb20-gVI-Nluc to 1, the relative expression levels of mRNA transcribed in different recombinant E. coli strains were calculated using the 2−ΔΔCt method [36]. Extracellular lactate dehydrogenase (LDH) activity assay The recombinant E. coli strains ER2738/pComb3X-Nb20-gVI-Nluc, ER2738/pComb3X-Nb20-Nluc-gVII, ER2738/pComb3X-Nb20-Nluc-gVIII, and ER2738/pComb3X-Nb20-Nluc-gIX were cultured under non‑induced and induced conditions, respectively. When the OD600 reached 0.6, 1 mL of each bacterial culture was collected by centrifugation at 8,000 × g and 4℃ for 2 min. The enzymatic activity was determined using a commercial lactate dehydrogenase kit (C0019S, Beyotime) with 100 µL of the supernatant. Probe specificity and interference resistance To assess specificity, 1 ml suspensions of V. parahaemolyticus, E. coli BL21(DE3), E. coli ER2738, E. coli TG1, Salmonella typhimurium, S. enteritidis serotype 3, S. enteritidis serotype 4, and S. enteritidis serotype 21 (each at 108 cfu/ml) were individually mixed with 100 µl of mpVIII (1012 pfu/ml). After incubation at room temperature for 15 min, the mixtures were centrifuged at 12,000 × g for 5 min, and the supernatant was discarded. Pellets were washed twice with a modified Hanks’ Balanced Salt Solution (mHBSS: 8 g/l of NaCl, 0.4 g/l of KCl, 1 g/l of glucose, 60 mg/l of KH₂PO₄, 47.5 mg/l of Na₂HPO₄, 1 g/l Triton X-100, 500 µl/l Tween-20, pH 8.0) and finally resuspended in 100 µl of PBS. Chemiluminescence was measured after adding 100 µl of the NanoLuc substrate and incubating for 5 min at room temperature. For interference tests, 1 ml of V. parahaemolyticus was mixed with 1 ml of E. coli BL21(DE3), 1 ml of S. typhimurium, or a combination of both non-target bacteria (all at 108 cfu/ml). The mixed cultures were centrifuged at 12,000 × g for 5 min. The supernatant was discarded, and the pellet was resuspended in 1 ml of PBS. This 1 ml bacterial suspension was then subjected to the same incubation, washing, and detection procedure with mpVIII as described above. Real sample detection Shrimp samples obtained from a local market were confirmed to be free of endogenous V. parahaemolyticus by plate counting. Each 1 g of shrimp sample was artificially inoculated with 0, 102, 103, 104, 105, 106, 107, or 108 cfu/g of live V. parahaemolyticus. Shrimp samples were then homogenized with 9 ml of sterile water using a homogenizer at low speed for 2 minutes. The homogenate was centrifuged at 500 × g for 5 minutes to remove large debris and the fat layer. The supernatant was transferred to a new tube and centrifuged at 8,000 × g for 10 min to collect bacterial cells. The pellet was washed twice with PBS and resuspended in 100 µl of PBS. This bacterial suspension was combined with 100 µl of mpVIII. Following incubation at 37 °C for 1 h, the mixture was applied to a spin column equipped with a MCE membrane (pore size: 0.45 μm; Fig. S7E) and centrifuged at 1,000 × g for 2 min. The spin column was then washed twice with mHBSS, followed by the addition of 200 µl of NanoLuc substrate. After 5 min of incubation at room temperature, the reaction solution was collected by centrifugation at 12,000 × g for 5 min and transferred for chemiluminescence detection. Detection using recovered phages Primary detection of various concentrations of V. parahaemolyticus (102 − 108 cfu/ml) was performed using concentrated mpVIII phages (1012 pfu/ml) according to the method described in Sect. 2.11. After incubation of the target bacteria with the mpVIII phages, the suspension was passed through a spin column, and the filtrate was collected as the mpVIII phages for the first round of recovery detection (recycle_1). Similarly, the suspension obtained after incubating the target bacteria with recycle_1 was passed through the spin column, and the filtrate was collected as the mpVIII phages for the second round of recovery detection (recycle_2). The same procedure was repeated using recycle_2 to obtain the mpVIII phages for the third round of recovery detection (recycle_3). The luminescence signals generated from each round of detection were recorded. Data analysis and visualization Statistical significance was determined using analysis of variance (ANOVA) implemented in SPSS software (version 20.0; IBM Corp., Armonk, NY, USA). Bar graphs were generated using GraphPad Prism (version 8.0.2; GraphPad Software, San Diego, CA, USA). Spearman’s rank correlation analysis was performed using Rstudio (version 2021.09.1; Posit Software, PBC, Boston, MA, USA). Schematic diagrams were created, and all figures were composited for publication using Adobe Illustrator CC 2019 (Adobe Inc., San Jose, CA, USA). Results and discussion M13 phage monovalent dual-display capability varies with display site and infection timing To investigate the capacity of M13 phage for the simultaneous display of two large protein molecules, a phage display system utilizing the phagemid pComb3X in conjunction with the helper phage M13KO7 was employed. This system was designed to co-display nanoluciferase (Nluc) [37] and nanobody Nb20 [38]. Specifically, the Nluc-gVII gene fusion cassette was inserted downstream of the Nb20-gIII expression cassette, generating the recombinant phagemid pComb3X-Nb20-Nluc-gVII. E. coli ER2738 harboring pComb3X-Nb20-Nluc-gVII consequently expresses Nb20-pIII and Nluc-pVII. Subsequent infection with helper phage M13KO7 facilitates the synthesis of other wild-type M13 coat proteins and the assembly of recombinant phage particles, designated M13-Nb20-m(Nluc-pVII) (Fig. 1A). The M13-Nb20-m(pVI-Nluc), M13-Nb20-m(Nluc-pVIII), and M13-Nb20-m(Nluc-pIX) constructs were generated following an analogous strategy (Fig. S1). Fig. 1Construction and characterization of monovalent double-displaying phages. (A) Construction strategy for the monovalent double-displaying phage mpVII. CSU characterization and titer measurements for mpVI (B), mpVII (C), mpVIII (D), and mpIX (E). NDP percentage and CoDP percentage for mpVI (F), mpVII (G), mpVIII (H), and mpIX (I). Different letters indicate significant differences (P < 0.05) Construction and characterization of monovalent double-displaying phages. (A) Construction strategy for the monovalent double-displaying phage mpVII. CSU characterization and titer measurements for mpVI (B), mpVII (C), mpVIII (D), and mpIX (E). NDP percentage and CoDP percentage for mpVI (F), mpVII (G), mpVIII (H), and mpIX (I). Different letters indicate significant differences (P < 0.05) Because M13KO7 provides wild-type coat proteins, the recombinant fusion proteins (Nb20-pIII, Nluc-pVII, etc.) compete with their wild-type counterparts during virion assembly. Following infection of recombinant E. coli ER2738 harboring the phagemid pComb3X-Nb20-Nluc-gVII with M13KO7, the harvested phage suspension contained a mixture of four phage types: M13 (wild-type), M13-Nb20 (phage displaying only Nb20), M13-m(Nluc-pVII) (phage displaying only Nluc), and M13-Nb20-m(Nluc-pVII) (dual-display phage). This mixture suspension was marked as the “mpVII suspension” (mpVII). Similarly, the other three monovalent double-displaying phage suspensions were assigned the designations mpVI suspension (mpVI), mpVIII suspension (mpVIII), and mpIX suspension (mpIX). And then, we evaluated the dual-display efficiency using four parameters: phage titer, chemiluminescent signal per unit concentration (1012 pfu/ml) of bacteriophage (CSU), the percentage abundance of Nluc-displaying phage (NDP), and the percentage of NDP that co-displayed both Nluc and Nb20 (hereafter referred to as CoDP). Given that the expression capacity of the phagemid might be influenced by the host growth phase [39, 40], we examined these parameters at four different optical density at 600 nm (OD600) points during M13KO7 infection. Initial validation using a serially diluted mpVII (titer range from 107 to 1012 pfu/ml) confirmed that the luminescent signal decreased proportionally with dilution (Fig. S3B), justifying the use of CSU for signal normalization. Significant variations in CSU and titer were observed for mpVI depending on the infection timing (Fig. 1B). Infection at OD600 = 0.6 yielded the highest CSU (2.19 × 107) with a titer of 2.88 × 1011 pfu/ml. For mpVII, the CSU from the OD600 = 0.6 infection was over 8-fold higher than at other time points, albeit with a lower titer of 1.20 × 109 pfu/ml (Fig. 1C). While the mpVII from OD600 = 0.4 infection achieved a near 1012 pfu/ml titer, its CSU was only 6.62 × 105. Notably, all four infection time points for mpVIII yielded CSU values exceeding 3.10 × 107, with titers ranging from 5.49 × 1010 pfu/ml to 1.95 × 1011 pfu/ml (Fig. 1D). Regarding mpIX, infection at OD600 = 0.4 produced the highest CSU (5.29 × 107), while titers across all time points remained around 1010 pfu/ml (Fig. 1E). Analysis of NDP and CoDP percentages revealed that for mpVI, the highest NDP percentage (3.42%) occurred at OD600 = 0.6, while the lowest (0.27%) was at OD600 = 0.4 (Fig. 1F). Within the NDP population for mpVI, the CoDP percentages were 10.69% (OD600 = 0.2), 19.32% (OD600 = 0.4), 12.75% (OD600 = 0.6), and 22.66% (OD600 = 0.8) (Fig. 1F). For mpVII, the NDP percentage reached 9.43% only at OD600 = 0.6, remaining below 1% at other time points (Fig. 1G). Strikingly, at OD600 = 0.4, CoDP constituted 33.39% of the NDP population for mpVII (Fig. 1G). mpVIII exhibited NDP percentages between 5.45% and 9.69%, with CoDP percentages ranging from 15.75% to 23.29% across infection times (Fig. 1H). Interestingly, the NDP percentage for mpIX at OD600 = 0.4 was notably high (15.03%), exceeding values from other time points by more than eightfold (Fig. 1I). However, within this NDP population at OD600 = 0.4, the CoDP proportion was only 14.92%, less than two-thirds of the value observed at OD600 = 0.6 (Fig. 1I). The systematic evaluation of M13 phage dual-display systems reveals critical dependencies on both the display site and assembly strategy. The CSU of a given phage suspension is significantly influenced by its titer. Correlation analysis revealed that lower phage titers generally correspond to larger CSUs in some cases (Fig. S6, A, B, and G). For instance, the mpVII at OD600 = 0.6, which had the lowest titer among the four time points examined, exhibited the highest CSU (Fig. 1C). This observation is attributed to a higher percentage of NDP in mpVII (Fig. 1G). Notably, a consistent trend in CSU variation with respect to helper phage infection time was not observed across the four types of monovalent double-displaying phages (Fig. 1, B, C, D, and E). This suggests that empirical optimization of the infection time point is necessary to maximize both the recombinant phage titer and display efficiency when displaying the same foreign protein at different capsid protein sites. Moreover, the luminescence signal of the recombinant phage suspensions exhibited only minor fluctuations or remained largely stable within a 60-minute period (Fig. S4 and S5). This stability is a key advantage that enables time-resolved biosensing with high temporal accuracy. It is noteworthy that mpVIII consistently demonstrated the highest CSU levels among all four monovalent double-displaying phages across all infection time points (Fig. 1D). This indicates a uniformly high NDP percentage in mpVIII (Fig. 1H). The M13 phage possesses approximately 2700 copies of the pVIII major coat protein, vastly outnumbering the minor coat proteins pVI, pVII, and pIX with 5 copies [41, 42]. Consequently, during phage assembly, the Nluc-pVIII fusion protein competes more successfully for incorporation into the final phage particle compared to the pVI-Nluc, Nluc-pVII, and Nluc-pIX fusions competing with their respective wild-type counterparts. Regarding phage yield, however, only mpVI (across all infection time points) consistently achieved titers exceeding 1011 pfu/ml (Fig. 1B). Consistent with previous reports, Nluc was fused to the C-terminus of the pVI protein in mpVI phages. Structural analysis of the M13 pVI protein suggests its C-terminus is buried within the phage particle, interacting with the genomic DNA to ensure stable assembly [42]. The pVI-Nluc fusion would likely encounter steric hindrance if it attempted to assemble identically to the wild-type pVI. The detected Nluc catalytic activity in mpVI confirms successful incorporation of the pVI-Nluc fusion. We hypothesize that the pVI-Nluc fusion adopts a distinct assembly conformation compared to the wild-type pVI, one that presents greater steric challenges and makes its incorporation less efficient. Thus, wild-type pVI holds a competitive advantage during assembly, leading to mpVI containing a higher proportion of wild-type M13 phages and a consequently lower NDP percentage (0.27%-3.42%, Fig. 1F). In diagnostic applications, it is the phage co-displayed both Nluc and Nb20 (PNN) that functionally contributes to detection. Therefore, the percentage of PNN within a phage suspension directly impacts the sensitivity of the detection system. The proportion of PNN relative to the total phage particles was generally higher in mpVIII (0.86%-2.20%, Fig. S7C) compared to the other three monovalent double-displaying phages (mpVI: 0.05%-0.44%, Fig. S7A; mpVII: 0.07%-0.38%, Fig. S7B; mpIX: 0.08%-2.24%, Fig. S7D). Theoretical calculations based on prior work by Clackson et al. suggest that the percentage of double-displaying phages in such systems should be < 1% [43]. However, our results deviated from this prediction in several instances. Notably, the PNN percentage for mpVIII at OD600 = 0.8 (2.20%) and for mpIX at OD600 = 0.4 (2.24%) both exceeded the theoretical threshold of 1% (Fig. S7, C and D). We proposed two potential explanations: (1) The nanobody Nb20 displayed at the N-terminus of pIII comprises only 125 amino acids (replacing the native N-terminal 248 amino acids of pIII). The smaller size of the Nb20-pIII fusion likely facilitates its display compared to bulkier proteins like scFvs. This was supported by CoDP percentages exceeding 10% within the NDP fraction for several samples (Fig. 1, F, G, H, and I). (2) The high copy number of pVIII might lead to the incorporation of more than one Nluc-pVIII fusion protein per phage particle in mpVIII. This would cause the theoretically calculated CoDP number (see Materials and methods) to be significantly higher than the actual value. M13 phage multivalent dual-display capability is inferior to its monovalent counterparts Considering that the monovalent display of Nluc might limit the CSU, we next explored multivalent Nluc display by cloning the Nluc gene into the helper phage M13KO7 genome (Fig. 2A and Fig. S2). Specifically, the Nluc gene was fused with gVII, generating recombinant M13KO7-Nluc-gVII. This recombinant phage vector, when propagated in E. coli TG1, produces M13KO7-p(Nluc-pVII). Infection of E. coli TG1/pComb3X-Nb20 with M13KO7-p(Nluc-pVII) yields the multivalent dual-display phage M13-Nb20-p(Nluc-pVII) (Fig. 2A). In this configuration, the Nluc-gVII fusion in the recombinant helper phage genome is the sole source for the pVII coat protein. Given that pVII assembles into the virion in 5 copies per phage, the Nluc display on M13-Nb20-p(Nluc-pVII) phage is theoretically pentavalent. Accordingly, the three multivalent dual-display phage suspensions obtained via this strategy (Fig. 2A and Fig. S2) were designated as ppVI suspension (ppVI), ppVII suspension (ppVII), and ppIX suspension (ppIX). Notably, despite multiple attempts, we were unable to obtain the M13KO7-Nluc-gVIII recombinant construct, as no transformants grew post-transformation. Fig. 2Construction and characterization of polyvalent double-displaying phages. (A) Construction strategy for the polyvalent double-displaying phage ppVII. CSU characterization and titer measurements for ppVI (B), ppVII (C), and ppIX (D). NDP percentage and CoDP percentage for ppVI (E), ppVII (F), and ppIX (G). Different letters indicate significant differences (P < 0.05) Construction and characterization of polyvalent double-displaying phages. (A) Construction strategy for the polyvalent double-displaying phage ppVII. CSU characterization and titer measurements for ppVI (B), ppVII (C), and ppIX (D). NDP percentage and CoDP percentage for ppVI (E), ppVII (F), and ppIX (G). Different letters indicate significant differences (P < 0.05) The multivalent dual-display capability was similarly assessed using the four aforementioned metrics. Contrary to expectations, the CSU levels for all multivalent dual-display phage suspensions, regardless of infection time, were lower than their monovalent counterparts (Fig. 2, B, C, and D). Specifically, no luminescent signal was detected for ppVII and ppIX with infection at OD600 = 0.2 and OD600 = 0.8 (Fig. 2, C and D). Titers of the three multivalent dual-display phage suspensions exhibited inconsistent trends with varying infection times. The titer of ppVI showed an “increase-decrease-increase” pattern with delayed infection (Fig. 2B). Conversely, ppVII titer displayed a “decrease-increase-decrease” trend (Fig. 2C), whereas ppIX titer declined initially before a slight subsequent rise (Fig. 2D). Surprisingly, the NDP percentages in all multivalent dual-display suspensions across infection time points were exceedingly low, remaining below 0.1% (Fig. 2, E, F, and G). Furthermore, the percentage of CoDP within the NDP population for ppVI, ppVII, and ppIX showed no consistent pattern with infection timing (Fig. 2, E, F, and G). The strikingly low efficiency of the multivalent display strategy necessitated an investigation into its underlying causes. Unlike monovalent Nluc display, fusing the Nluc gene to the helper phage genomic copies of the minor coat protein genes gVII, gVI, or gIX should, in theory, result in phages displaying 5 copies of Nluc per particle (Fig. 2A and Fig. S2). Surprisingly, the CSUs of the three polyvalent double-displaying phages were significantly lower than their monovalent counterparts (Fig. 1, B, C, and E; Fig. 2, B, C, and D). Furthermore, the NDP percentages for all samples of the three polyvalent phages were even below 0.1% (Fig. 2, E, F, and G), a finding starkly inconsistent with theoretical expectations. We speculated two primary reasons for this low NDP percentage: (1) Unlike in monovalent display, the enzymatic activity of Nluc in the pVI-Nluc, Nluc-pVII, and Nluc-pIX fusions might be substantially reduced when incorporated multivalently, due to steric crowding and constrained conformation within the assembled phage particle (Fig. 2A and Fig. S2). (2) Viral particles could assemble into non-standard configurations when capsid proteins were mutated or fused to heterologous proteins, a plasticity that facilitated viral adaptation and immune evasion [44–47]. We hypothesized that the pVI, pVII, and pIX proteins might assemble with fewer than the standard 5 copies, perhaps only 4 or less. This non-standard assembly might become particularly extreme under certain specific physiological conditions of the host. For instance, in the cases of ppVII and ppIX at OD600 = 0.2 and OD600 = 0.8, the assembly efficiency of Nluc-fused capsid proteins was zero (Fig. 2, F and G). Additionally, we were unable to successfully construct the recombinant E. coli TG1/M13KO7-Nluc-gVIII. Although the signal peptide sequence of pVIII—responsible for directing the pVIII coat protein to the cell membrane [48]—was preserved in our design, the steric hindrance imposed by approximately 2700 copies of the Nluc-pVIII fusion protein likely impeded proper phage assembly. The accumulation of unincorporated Nluc-pVIII fusion proteins in the periplasmic space of the host cells was presumed to obstruct normal material and energy exchange between the host and its environment, ultimately leading to host lethality. IPTG induction enhances CoDP percentage in monovalent dual-display phage Data from the monovalent dual-display experiments indicated that the majority of phage particles in the suspensions lacked displayed Nluc (Fig. 1, F, G, H, and I). Moreover, the actual Nluc display rate (< 0.1%) in multivalent experiments fell far short of the theoretical expectation (100%). We therefore assessed the relative displaying level (RDL) of Nluc via Western blot for samples from the OD600 = 0.6 infection point. In comparison to RDL for ppVII (1.05) and ppIX (1.00), ppVI RDL was lower (0.57) (Fig. 3A). Besides, mpVI and mpIX exhibited RDL values greater than 1 (1.73 and 1.18, respectively); RDL values for mpVII (0.86) and mpVIII (0.71) were not significantly different. After correcting the CSU of each phage sample by its RDL, we obtained the luminescent signal per unit phage concentration, which was normalized to the same display level. This analysis revealed that the corrected signals of mpIX, ppVI, ppVII, and ppIX were similar and significantly lower than those of mpVI, mpVII, and mpVIII (Fig. 3B). Furthermore, the corrected signal of mpVIII (5.12 × 107) was 4.28 and 6.03 times higher than that of mpVI and mpVII, respectively (Fig. 3B). Given potential differences in antibody affinities or epitope accessibility across the various fusions (e.g., pVI-Nluc versus Nluc-pVII), this may affect the accuracy of the RDL measurements in this study. In the future, more precise methods could facilitate the quantification of RDL or even absolute display levels, such as mass spectrometry or high-resolution cryo-electron microscopy, provided that sufficiently concentrated and homogeneous dual-display phage particles can be isolated. Fig. 3Recombinant phage RDL and induction expression studies. (A) Western blot results for seven phage suspensions at OD600 = 0.6. (B) CSU of the seven phage suspensions at OD600 = 0.6 after RDL correction. Rank correlations for four monovalent double-displaying phage suspensions at OD600 = 0.6: between titer and absolute host Nluc mRNA levels (C), between CSU and absolute host Nluc mRNA levels (D), between CSU and titer (E). Rank correlation for three polyvalent double-displaying phage suspensions at OD600 = 0.6: between titer and absolute host Nluc mRNA levels (F), between CSU and absolute host Nluc mRNA levels (G), between CSU and titer (H). CSU and titer (I), NDP percentage and CoDP percentage (J) for four monovalent double-displaying phage suspensions at OD600 = 0.6 after IPTG induction. Fold change in CSU and titer (K), and fold change in NDP percentage and CoDP percentage (L) for the four monovalent double-displaying phage suspensions after IPTG induction compared to pre-induction levels. Relative expression levels of the Nluc gene (M) and extracellular LDH content (N) in recombinant E. coli ER2738 harboring different phagemids at OD600 = 0.6. Different letters indicate significant differences (P < 0.05). Values under non‑induced and induced conditions were analyzed by paired t‑test; n.s., not significant (P > 0.05), *, P < 0.05, **, P < 0.01, ***, P < 0.001 Recombinant phage RDL and induction expression studies. (A) Western blot results for seven phage suspensions at OD600 = 0.6. (B) CSU of the seven phage suspensions at OD600 = 0.6 after RDL correction. Rank correlations for four monovalent double-displaying phage suspensions at OD600 = 0.6: between titer and absolute host Nluc mRNA levels (C), between CSU and absolute host Nluc mRNA levels (D), between CSU and titer (E). Rank correlation for three polyvalent double-displaying phage suspensions at OD600 = 0.6: between titer and absolute host Nluc mRNA levels (F), between CSU and absolute host Nluc mRNA levels (G), between CSU and titer (H). CSU and titer (I), NDP percentage and CoDP percentage (J) for four monovalent double-displaying phage suspensions at OD600 = 0.6 after IPTG induction. Fold change in CSU and titer (K), and fold change in NDP percentage and CoDP percentage (L) for the four monovalent double-displaying phage suspensions after IPTG induction compared to pre-induction levels. Relative expression levels of the Nluc gene (M) and extracellular LDH content (N) in recombinant E. coli ER2738 harboring different phagemids at OD600 = 0.6. Different letters indicate significant differences (P < 0.05). Values under non‑induced and induced conditions were analyzed by paired t‑test; n.s., not significant (P > 0.05), *, P < 0.05, **, P < 0.01, ***, P < 0.001 To explore potential factors influencing phage titer and CSU, we performed correlation analyses with the absolute transcript levels of the Nluc gene within host cells at the time of helper phage infection. For monovalent dual-display phages harvested from OD600 = 0.6 infection, both titer and CSU showed strong positive correlations with the absolute quantity of Nluc mRNA (Fig. 3, C and D). A positive correlation was also observed between CSU and titer for these samples (Fig. 3E, p = 0.04). For multivalent dual-display phages from the same infection point, a significant positive correlation was found only between CSU and Nluc mRNA absolute quantity (ρ = 0.90, p = 0.00, Fig. 3G). In light of the positive correlations between phage yield/CSU and Nluc mRNA levels in monovalent display, we sought to increase the CoDP percentage by inducing overexpression of the Nb20-gIII fusion gene from the phagemid with IPTG (isopropyl β-D-thiogalactoside). IPTG (final concentration 1 mmol/l) was added at the time of inoculating the host culture. When the induced culture reached OD600 = 0.6, helper phage M13KO7 was added for infection. Among the four harvesting monovalent dual-display phages after induction, only mpVII and mpVIII achieved CSU values above 106 (Fig. 3I). mpVI CSU was merely 334.59, and no CSU was detected for mpIX (Fig. 3I). Regarding phage yield, mpVI titer exceeded 1012 pfu/ml, whereas mpVII, mpVIII, and mpIX titers were all below 1010 pfu/ml (Fig. 3I). The NDP percentages of mpVII (48.15%) and mpVIII (11.24%) were significantly higher than that of mpVI (Fig. 3J). Conversely, the CoDP percentages for mpVI (31.19%) and mpVIII (30.02%) significantly surpassed that for mpVII (9.19%) (Fig. 3J). Upon IPTG induction, the CSU for mpVII increased 10.78-fold compared to the uninduced control, whereas CSU values for mpVI, mpVIII, and mpIX dropped to less than 0.10-fold of their uninduced levels (Fig. 3K). Induced mpVI titer was 7.57 times higher than its uninduced counterpart, and induced mpVII titer remained nearly identical (0.94-fold) to the uninduced control (Fig. 3K). In contrast, induced titers for mpVIII and mpIX decreased by over 100-fold relative to uninduced levels (Fig. 3K). Regarding percentages, induced NDP values for mpVII and mpVIII were 5.11-fold and 2.04-fold higher, respectively, than their uninduced values (Fig. 3L). Remarkably, the induced NDP for mpVI was drastically reduced to less than 0.001-fold of its uninduced level (Fig. 3L). Encouragingly, IPTG induction led to increases in the CoDP percentage for mpVI, mpVII, and mpVIII, reaching 2.45, 2.31, and 1.90 times higher than their respective uninduced values (Fig. 3L). To explore the reasons for the divergent responses of the four display systems to IPTG induction, we examined the growth status, Nluc gene expression levels, and total extracellular LDH levels (a surrogate marker of membrane integrity and cytotoxicity [49, 50] of the recombinant E. coli strains harboring these four systems under both uninduced and induced conditions. Although no obvious differences in growth status were observed between uninduced and induced conditions for the four recombinant strains (Fig. S6H–S6K), both Nluc gene expression levels (Fig. 3M) and extracellular LDH activity (Fig. 3N) were higher upon induction. These results indicate that IPTG induction significantly elevated Nluc gene transcription while also exerting certain cytotoxic effects. Notably, the recombinant E. coli ER2738/pComb3X-Nb20-Nluc-gIX exhibited the highest extracellular LDH activity upon induction, suggesting that this strain suffered from more severe cytotoxicity, which likely accounts for the low CoDP percentage observed in this system (Fig. 3J). Beyond cytotoxicity, the divergent responses of the four display systems to IPTG induction probably indicate that the underlying assembly mechanisms are not universal. The RDL analysis for phages with different display formats aligned with the non-standard assembly hypothesis discussed above. For example, ppVI exhibited a significantly lower RDL than ppVII and ppIX, whereas mpVI showed an RDL considerably greater than 1 (Fig. 3A). The high RDL for mpVI might be partially attributable to the larger size of the pVI-Nluc fusion (295 amino acids) compared to the Nluc-pIX fusion (213 amino acids). This size discrepancy was not corrected in our RDL calculations, as Western blot band sizes did not show obvious differences (Fig. 3A). Nonetheless, we used the RDL to normalize the CSUs, allowing comparison of Nluc activity across display formats. The normalized Nluc activity was markedly higher in monovalent displays than in their polyvalent equivalents (Fig. 3B), further supporting the notion that Nluc activity in the polyvalent context was likely hampered by a crowded assembly conformation. Interestingly, a positive correlation was observed between the intracellular abundance of Nluc mRNA in the host at the time of helper phage infection and the gained CSU, for both monovalent and polyvalent display experiments (Fig. 3, D and G). However, upon IPTG induction, among the monovalent double-displaying phages, only mpVII showed an increase in CSU (Fig. 3K), and only mpVI exhibited a titer increase (Fig. 3K). Notably, IPTG induction effectively increased the percentage of CoDP within the NDP fraction for all phages except for mpIX (Fig. 3L), indicating that induction might be a viable strategy to enhance the proportion of functional phages. Regrettably, no consistent pattern emerged in the changes of CSU, titer, NDP percentage, and CoDP percentage across the four monovalent phages harvested after the induction treatment. This lack of consistency implied that the assembly of double-displaying phages involving different site combinations was governed by distinct and intricate regulatory mechanisms. Elucidating these mechanisms will be crucial for further improving phage yield and the effective phage ratio in suspensions, thereby accelerating the application of monovalent double-displaying phages in biosensing. Detection performance of monovalent dual-display phage mpVIII Considering both signal intensity and phage yield, the mpVIII from the OD600 = 0.6 infection offered a favorable combination of high titer and the highest CSU (Fig. 1D). Consequently, this phage suspension was selected for assessing detection performance in real shrimp samples (Fig. 4A). For single-bacterium detection assays, a concentration of mpVIII (1012 pfu/ml) was used to evaluate antibody specificity. The signal-to-noise ratio for detecting V. parahaemolyticus was 3.97 (Fig. 4B), substantially higher than ratios obtained for other non-target bacteria (1.10–1.34), indicating the excellent detection specificity of mpVIII. In mixed-bacteria detection assays, a high concentration of the target sample (108 cfu/ml) was employed to test the system’s anti-interference capability. Results demonstrated no significant difference between signals from mixed-bacteria and single-bacterium detection (Fig. 4C), confirming the robust anti-interference capacity of the mpVIII-based system. For detection in shrimp samples, we employed a custom-made spin column fitted with a mixed cellulose esters (MCE) membrane. The column assembly consisted of four layers of MCE membrane securely compressed by an inverted, truncated 200 µl pipette tip placed above the membrane (Fig. S7E). During the detection assay, the mixture of sample and mpVIII after incubation, washing buffer mHBSS, and NanoLuc substrate solution were all applied through the opening of the truncated pipette tip (Fig. S7E). The detection results demonstrated a strong linear relationship between the signal intensity and the logarithm of V. parahaemolyticus concentration (R2 = 0.9701, Fig. 4D), with a limit of detection of 102 cfu/ml (Table 1 and S3). Additionally, unbound phages in the filtrate and washing solutions were recovered. The recovered mpVIII exhibited chemiluminescence signals ranging from 46.97% to 72.34% of the original mpVIII, and this recovery showed a significant negative correlation with the logarithmic concentration of V. parahaemolyticus (p = 4.768 × 10− 10, Fig. 4E). Finally, the detection performance of the recovered mpVIII phages was evaluated. When the target bacterial concentration was high (105 − 108 cfu/ml), the detection signals from the recovered phages consistently fell below the level of the original mpVIII (Fig. 4F). In contrast, at target bacterial concentrations of 102, 103, and 104 cfu/ml, the recovered phages achieved luminescent signals comparable to those of the original mpVIII after two, one, and one round(s) of recovery, respectively (Fig. 4F). Fig. 4Detection performance of mpVIII. (A) Schematic of the detection assay using mpVIII. Specificity (B), anti-interference capability (C), and detection range (D) of the assay using mpVIII. (E) Luminescent signals of the recycled mpVIII from the filtrate and wash solutions after detection. (F) Luminescent signals from the detection of various concentrations of V. parahaemolyticus using three sequential rounds of recycled mpVIII phages. VVAL, a mixed solution of the four bacteria: V. alginolyticus, V. harveyi, A. hydrophila, and L. monocytogenes, n. s., not significant (paired t-test, P > 0.05), *, P < 0.05, **, P < 0.01, ***, P < 0.001 Detection performance of mpVIII. (A) Schematic of the detection assay using mpVIII. Specificity (B), anti-interference capability (C), and detection range (D) of the assay using mpVIII. (E) Luminescent signals of the recycled mpVIII from the filtrate and wash solutions after detection. (F) Luminescent signals from the detection of various concentrations of V. parahaemolyticus using three sequential rounds of recycled mpVIII phages. VVAL, a mixed solution of the four bacteria: V. alginolyticus, V. harveyi, A. hydrophila, and L. monocytogenes, n. s., not significant (paired t-test, P > 0.05), *, P < 0.05, **, P < 0.01, ***, P < 0.001 Table 1Recovery tests for V. parahaemolyticus in spiked shrimp (n = 3)SamplesSpiked (cfu/g)Found (cfu/ml)Recoveries (%)RSDs (%)Shrimp 12 × 102176 ± 688.00 ± 2.783.16Shrimp 22 × 1031766 ± 7988.32 ± 3.954.47Shrimp 32 × 10419,840 ± 1320106.52 ± 18.746.65Shrimp 42 × 105201,849 ± 26,114100.92 ± 13.0612.94Shrimp 52 × 1062,220,102 ± 237,152111.01 ± 11.8610.68Shrimp 62 × 10719,719,711 ± 2,224,62698.60 ± 11.1211.28Shrimp 72 × 108214,645,893 ± 22,710,802107.32 ± 11.3610.58Data are presented as mean (± standard deviation, where applicable) with units indicated in parentheses Table 1Recovery tests for V. parahaemolyticus in spiked shrimp (n = 3)SamplesSpiked (cfu/g)Found (cfu/ml)Recoveries (%)RSDs (%)Shrimp 12 × 102176 ± 688.00 ± 2.783.16Shrimp 22 × 1031766 ± 7988.32 ± 3.954.47Shrimp 32 × 10419,840 ± 1320106.52 ± 18.746.65Shrimp 42 × 105201,849 ± 26,114100.92 ± 13.0612.94Shrimp 52 × 1062,220,102 ± 237,152111.01 ± 11.8610.68Shrimp 62 × 10719,719,711 ± 2,224,62698.60 ± 11.1211.28Shrimp 72 × 108214,645,893 ± 22,710,802107.32 ± 11.3610.58Data are presented as mean (± standard deviation, where applicable) with units indicated in parentheses Recovery tests for V. parahaemolyticus in spiked shrimp (n = 3) Data are presented as mean (± standard deviation, where applicable) with units indicated in parentheses Because mpVIII carried both the targeting agent (Nb20) and the signal generator (Nluc), the detection protocol was remarkably straightforward, requiring only minimal sample pretreatment (Fig. 4A). mpVIII demonstrated excellent target specificity and resistance to interference (Fig. 4, B and C). The detection limit achieved using mpVIII was 102 cfu/ml (Table S3), which is comparable to that of several established immunological detection methods [38, 51–54]. Notably, unbound phages could be efficiently recovered from both the filtrate and washing solutions (Fig. 4E). These recovered phages remain suitable for subsequent detection of V. parahaemolyticus, thereby offering a potential reduction in overall detection costs. Unexpectedly, the recovered mpVIII phages did not exhibit restored signals when detecting high concentrations of the target bacterium (Fig. 4F). It is speculated that this is because a substantial amount of CoDP in the mpVIII phages was bound and consumed during the initial detection, leaving insufficient CoDP to support subsequent detection signals. Furthermore, even at a low target bacterium concentration (102 cfu/ml), the third round of recovered phages failed to show restored signals (Fig. 4F), indicating that some degree of CoDP loss also occurred during the recovery process. Nevertheless, as the detection protocol relies on membrane filtration to retain the “V. parahaemolyticus-dual-display phage” complex, the detection system is inherently suited for targets of relatively large size, such as bacterial cells. Furthermore, mpVIII used for detection in this study had a titer of 1012 pfu/ml, a concentration at which the solution turbidity is already substantial (Fig. S7F). Attempts to concentrate the preparation further to achieve a lower detection limit were hampered by the apparent saturation solubility of the phage at titers exceeding 1012 pfu/ml, leading to precipitate formation upon standing. Future efforts to enhance the detection performance of monovalent double-displaying phages should focus on optimizing buffer systems to increase phage solubility and/or reduce background noise. Conclusion In summary, this study provides a systematic comparison of M13 phage-based dual-display systems, revealing that display efficiency varies considerably depending on the capsid protein site, infection timing, and induction conditions. Among the constructs evaluated, the mpVIII system demonstrated favorable performance in terms of specific luminescent activity, functional co-display proportion, and detection capability for V. parahaemolyticus in spiked shrimp samples. These findings suggest that mpVIII represents a promising candidate for further development as a biosensing probe. Future efforts should focus on optimizing phage suspension buffer formulations to enhance solubility and minimize background interference, which will be crucial for further improving the detection limit and practical applicability of these versatile biosensing probes in complex real-world samples. Supplementary Information Supplementary Material 1 Supplementary Material 1