Sub-5 nm lanthanide-doped lutetium oxyfluoride nanoprobes for ultrasensitive detection of prostate specific antigen† †Electronic supplementary information (ESI) available: ESI, Tables S1–S4, Fig. S1–S20 and Movie S1. See DOI: 10.1039/c5sc04599a

We demonstrate the successful use of sub-5 nm Lu6O5F8:Eu3+ nano-bioprobes for the ultrasensitive detection of prostate specific antigen in patient serum samples with a limit of detection of 0.52 pg mL–1.


Introduction
Prostate cancer (PCa) is the third most common cancer in men worldwide. 1 Prostate specic antigen (PSA) has been commonly used as a tumour marker for the early diagnosis and monitoring of PCa. 2 For early diagnostic purposes, a PSA level of >4 ng mL À1 is considered as an indicator of suspicion of PCa. For the monitoring of PCa relapse, the PSA level in the serum usually decreases to $1 pg mL À1 aer radical prostatectomy, and a serial increase in the PSA level indicates the recurrence or metastasis of PCa. 3 Thus, ultrasensitive detection of PSA featuring a large linear range from $1 pg mL À1 to >4 ng mL À1 is crucial for the theranostics of PCa.
Tremendous efforts have been devoted to developing sensitive PSA-detecting approaches with a large linear range. Therein, uorescence immunoassays are most commonly employed owing to their good compatibility with currently available analytical platforms in the clinic. 4 However, commercial uorescence bioassays have detection limits around 0.1 ng mL À1 , which are insufficient for the monitoring of PCa relapse aer radical prostatectomy. Therefore, an increased emphasis has been exerted on designing ultrasensitive assays with much better analytical sensitivity than commercial kits. 5 Among uorescence immunoassays previously established for the detection of PSA, time-resolved (TR) photoluminescence (PL) bioassays based on lanthanide-chelate embedded polystyrene or silica nanoparticles (NPs) are capable of achieving improved sensitivity than a commercial kit. 6 Nevertheless, such nanoprobes (usually >40 nm) have a tendency to agglomerate and swell in aqueous solution, and potential chelate leakage concerns have also been documented. 6,7 Compared with lanthanide chelates, lanthanidedoped inorganic NPs have become a research hotspot in biodetection owing to their relatively high stability and greater exibility for bioconjugation. [8][9][10][11][12][13][14][15][16][17][18] In particular, ultrasmall (<5 nm) nano-bioprobes in the size range of biological molecules are desired due to their minimized interference with the antigen-antibody binding process. 19 Hitherto, it is still a challenge to develop ultrasmall lanthanide luminescent nanobioprobes with both a high detection sensitivity and a large linear detection range. Recently, we have developed the dissolution-enhanced luminescent bioassay (DELBA) technique based on inorganic lanthanide uoride NPs. 20 Upon the addition of NPs to the enhancer solution (i.e. Triton X-100 micelle solution), the trivalent lanthanide (Ln 3+ ) ions in the NPs are extracted into micelles containing chelating ligands to form highly luminescent lanthanide complexes. As such, the TRPL signals are amplied, improving the detection sensitivity. However, there is still great demand for exploring inorganic NPs with more highly concentrated Ln 3+ ions and better dissolution performance to further improve the detection sensitivity and linear range of bioprobes. In comparison with uorides, Ln 3+ doped lutetium oxyuorides, like Lu 6 O 5 F 8 , have a much higher molar density of Ln 3+ ions in the matrix (ESI Table S1 †). A larger amount of lanthanide luminescent micelles can be transformed from a single oxyuoride NP. Besides, the Ln-O bond in Lu 6 O 5 F 8 can be more easily loosened than the Ln-F bond via the protonation reaction with H + , 21 which may facilitate Ln 3+ ion extraction from the NP surface into micelles. As a result, Ln 3+ doped Lu 6 O 5 F 8 is expected to be more suitable than uorides as an ultrasmall bioprobe for the detection of PSA.
Herein, monodisperse and ultrasmall Ln 3+ (Ln ¼ Eu, Yb/Er) doped lutetium oxyuoride NPs, which had never been explored before, were synthesized via a modied thermal decomposition route. 22 By utilizing the Lu 6 O 5 F 8 :Eu 3+ nanoprobes featuring a high molar density of Ln 3+ ions and superior dissolution properties in the enhancer solution, we demonstrate the ultrasensitive and accurate detection of PSA in patient serum samples through heterogeneous sandwich bioassays in the TRPL detection mode, as depicted in Fig. 1. Furthermore, we reveal the great potential of ultrasmall Lu 6 O 5 F 8 :Yb 3+ /Er 3+ nanoprobes in upconversion luminescence (UCL) and computed tomography (CT) dual-modal bioimaging.

Results and discussion
Monodisperse and ultrasmall Ln 3+ (Ln ¼ Eu, Yb/Er) doped lutetium oxyuoride NPs were synthesized via a modied thermal decomposition route. The as-prepared Eu 3+ (5 mol%) doped Lu 6 O 5 F 8 NPs are readily dispersed in cyclohexane (Fig. 2a). TEM analysis shows that these NPs are nearly spherical with an average diameter of 4.8 AE 0.5 nm (Fig. 2b), as corroborated by the broad powder X-ray diffraction (XRD) peaks, which are well indexed as the orthorhombic phase of Lu 6 O 5 F 8 (ESI Fig. S1 †). The thermogravimetric analysis of the asprepared NPs shows a weight loss of 5.17% between 450 and 750 C (ESI Fig. S2 †), which is consistent with the theoretical weight loss (5.19%) resulting from the decomposition of  Table S2 †). Fig. 2d shows the characteristic Eu 3+ downshiing (DS) emissions assigned to the 5 D 0 / 7 F 0-4 transitions in Lu 6 O 5 F 8 :Eu 3+ (5 mol%) NPs upon excitation at 394 nm ( 7 F 0 / 5 L 6 ). The electric-dipole 5 D 0 / 7 F 2 transitions at $620 nm are much stronger than the magneticdipole 5 D 0 / 7 F 1 transitions at $590 nm, indicating that the doped Eu 3+ ions occupy low-symmetry sites. 23 The appearance of the 5 D 0 / 7 F 0 transitions infers that the site symmetry of Eu 3+ in Lu 6 O 5 F 8 should be restricted to noncentrosymmetric C s , C n , or C nv (n ¼ 1, 2, 3, 4, 6) groups, 23 which is in accordance with the structural analysis that the doped Ln 3+ ions may occupy several sites with symmetries of C 1 , C s or C 2 in Lu 6 O 5 F 8 . 24 Meanwhile, in the corresponding excitation spectrum of the NPs (ESI Fig. S1b †), a broad O 2À -Eu 3+ charge transfer (CT) absorption in the 230-300 nm region was observed, which is typical of Eu 3+ doped oxyuoride hosts. 24 Besides the intense DS luminescence, the colloidal cyclohexane solution of Lu 6 O 5 F 8 :-Yb/Er (10/2 mol%) NPs displayed intense red UCL (inset of    2e). The UCL spectrum shows that the UCNPs exhibited strong red emission at $650 nm and relatively weak green emission at $540 nm with a red-to-green (R/G) ratio of $13, which are assigned to the 4 F 9/2 / 4 I 15/2 and 2 H 11/2 / 4 S 3/2 / 4 I 15/2 transitions of Er 3+ , respectively. [25][26][27] Meanwhile, we observed a much shorter UCL lifetime of 4 S 3/2 (2.3 ms) than that of 4 F 9/2 (15.2 ms) of Er 3+ (ESI Fig. S7 †). The enhanced red UC emission may be attributed to the efficient energy back transfer (EBT) process from Er 3+ to Yb 3+ in the Lu 6 O 5 F 8 :Yb/Er lattice (ESI Fig. S8 †). 28,29 Such a high R/G ratio in UC phosphors is highly desired for in vivo bioimaging because the red UCL exhibits deeper tissue penetration than green light. 29 To evaluate the dissolution performance of the Lu 6 O 5 F 8 :Eu 3+ NPs, we rstly removed the oleate acid (OA) ligands from their surface by an acid-washing treatment to render the NPs hydrophilic. 30 The successful removal of the surface ligands was conrmed by the Fourier transform infrared (FTIR) spectra (ESI Fig. S9 †). Then, we added the solution of ligand-free Lu 6 O 5 F 8 :-Eu 3+ (40 mol%) NPs (20 mg mL À1 ) to the enhancer solution (i.e. the solution of Triton X-100 micelle containing 2-naphthoyltri-uoroacetone (b-NTA) and tri-n-octylphosphine oxide (TOPO) chelating ligands in the inner cavity, pH ¼ 2.76). The appearance of a myriad of tiny NPs (<3 nm) in the TEM images indicates the occurrence of the dissolution reaction of the NPs, which was evidenced by the strong deep pink emission under ultraviolet (UV) lamp illumination upon addition of the NPs to the enhancer solution ( Fig. 3a-c). To elucidate the origin of the dissolution-enhanced PL, we compared the steady-state PL excitation/emission spectra and the PL decays of the NPs before and aer dissolution. The results show unambiguously that the enhanced Eu 3+ PL of the NPs dissociated in the enhancer solution originated from the lanthanide complex, instead of the NPs (ESI Fig. S10 †).
Since the lanthanide complexes are formed through the coordination reaction of the chelating ligands with the Ln 3+ ions in the NPs (Fig. 3d), the dissolution-enhanced PL intensity of the Lu 6 O 5 F 8 :Eu 3+ NPs depends critically on the molar concentration of Eu 3+ in the NPs. Specically, we found that Lu 6 O 5 F 8 NPs doped with 40 mol% Eu 3+ proved the most satisfactory (ESI Fig. S11 †), due to the fact that the inert Lu 3+ ion complex may decrease the concentration quenching of Eu 3+ and enhance the luminescence of Eu 3+ complex through a co-uorescence effect. 31 Similarly, the pH value of the enhancer solution has a complicated effect on the dissolution process of the NPs and the PL intensity of the Eu 3+ complex (Fig. 3e). 31 To gain deep insight into the dissolution dynamics of the NPs and to screen out optimal pH conditions for NP dissolution, we derived a kinetic equation (equation (S7) in ESI †) to simulate the dissolution process of the NPs. As shown in Fig. 3e and ESI  Fig. S12, † the PL signal was well tted by the kinetic equation. The kinetic parameters derived from the tting are listed in ESI Table S3. † The results show that the equilibrium constant K eq is the highest at pH 2.76. The higher value of K eq indicates that a larger amount of NPs is dissolved when reaction equilibrium is reached. Consistently, the dissociation/association rate constants (K d /K a ) are relatively high at pH 2.76, which means only a short time is needed to reach equilibrium in the dissolution reaction. As such, 2.76 was chosen as the optimal pH value for the dissolution of Lu 6 O 5 F 8 . Under these optimized conditions, we measured the linear dynamic range of dissolution-enhanced PL intensity versus the NP concentration. The PL signal increased linearly with the NP concentration in the large range of 0-12.5 mg mL À1 (i.e. 0-61.1 nmol Ln 3+ per mL) (Fig. 3f). As shown in Fig. 3e, the dissolution-enhanced PL of NPs with a concentration even as high as 50 mg mL À1 was effectively stable within 5 minutes under optimum conditions. These results reveal the excellent dissolution performance of ultrasmall Lu 6 O 5 F 8 :Eu 3+ nanoprobes for bioassays.
Prior to the luminescent bioassay, the OA-capped Lu 6 O 5 F 8 :-Eu 3+ NPs were surface-modied via ligand exchange with citrate, and then coupled with avidin following the EDC/NHS protocol. 32,33 The successful conjugation of avidin to the surface of the NPs was corroborated by the appearance of amide bands in the FTIR spectrum, the increase in dynamic light scattering (DLS) size and the decrease of the zeta-potential for the NPs aer surface modication (ESI Fig. S13 and S14 †). By utilizing the bicinchoninic acid (BCA) protein assay kit, the number of avidin molecules conjugated to each NP was estimated to be $1.8 (ESI Fig. S15 and S16 †).
By virtue of the specic recognition of the anti-PSA antibody with PSA, we employed avidin-conjugated Lu 6 O 5 F 8 :Eu 3+ nanoprobes in a sandwich-type bioassay for the detection of PSA. As illustrated in Fig. 4a and ESI Movie S1, † the capture antibody was rst bound to the microplate well, and the biotinylated detection antibody (which was bound to PSA) was coupled to the nanoprobe through biotin-avidin interactions. PSA was quantied by measuring the dissolution-enhanced TRPL signal of the oxyuoride nanoprobes upon addition of the enhancer solution on a microplate reader. For comparison, control experiments by replacing PSA with bovine serum albumin (BSA) under otherwise identical conditions, were also conducted. These control experiments showed negligible TRPL signal (Fig. 4b), verifying the high specicity. The limit of detection (LOD), dened as the concentration that corresponds to three times the standard deviation above the signal measured in the control experiment, was determined to be 0.52 pg mL À1 (15.2 fM). The derived LOD exhibits an almost 200-fold improvement relative to that of a commercial dissociationenhanced lanthanide uoroimmunoassay (DELFIA) kit (0.1 ng mL À1 ), 34 and is the lowest among lanthanide nanoprobes ever reported for the luminescent bioassay of PSA (ESI Table S4 †). Such a low LOD is attributed to the high molar density of Eu 3+ ions in a single oxyuoride NP and the high specicity of nanoprobes in the bioassays.
It is worth emphasizing that the luminescent bioassay based on ultrasmall oxyuoride nanoprobes demonstrates an excellent linear dependence of the PL intensity on PSA concentration over a rather wide range, namely, 8.5 Â 10 À4 to 5.6 ng mL À1 (R 2 ¼ 0.998) (Fig. 4b and ESI Fig. S17 †). According to Diamandis et al., the PSA levels in $60% of patients suffering from PCa decreased to <5 pg mL À1 aer radical prostatectomy, 35 and a postoperative increase of serum PSA indicates the recurrence or metastasis of cancer. 3 Beneting from the ultrahigh sensitivity of our nanoprobe, such a large linear dynamic range spanning four orders of magnitude for a PSA bioassay might full the critical requirement for tracing the serum PSA level for the early diagnosis of PCa and for monitoring PCa relapse aer prostatectomy.
To further verify the applicability of the Lu 6 O 5 F 8 :Eu 3+ nanoprobes in practical bioassays for PSA, we rstly studied the response of the oxyuoride probes towards potential interfering proteins coexisting in serum samples. PSA (300 pM) and noncognate proteins (10 nM), including human serum albumin (HSA), carcinoembryonic antigen (CEA), alpha fetoprotein (AFP), and beta-human chorionic gonadotropin (b-HCG), were added to prepared serum samples from a healthy male with an original PSA level of 0.54 ng mL À1 (15.88 pM) determined by a commercial DELFIA kit, respectively. In the blank control, no targets were added to the samples. As shown in ESI Fig. S18, † the PL signals of all the samples with noncognate proteins added were indistinguishable from that of the blank control. By contrast, when the standard solution of PSA was added, even with its concentration more than 30-fold lower than that of the noncognate protein, a greater than 7 times stronger PL signal was observed. These results indicate that nonspecic binding is negligible, thus conrming the great specicity of our proposed method for the detection of PSA in complex human serum samples. Then, we carried out in vitro detection of PSA in 23 serum samples of patients with PSA levels from 12.00 to 140.00 ng mL À1 . The PSA levels determined were compared with those independently detected using a commercial DELFIA kit. As shown in Fig. 4c and Table 1, the PSA levels determined from the DELBA assay based on Lu 6 O 5 F 8 :Eu 3+ NPs are consistent with those from the DELFIA. The correlation coefficient between both kinds of assays was determined to be 0.97, indicating that the NPs-based assay results are as reliable as that of the commercial DELFIA. The coefficients of variation (CV) of the Lu 6 O 5 F 8 :Eu 3+ -based assay ranged from 0.2% to 8.5% (Table 1), showing the assay's good reproducibility. Moreover, we evaluated the analytical accuracy and precision of the Lu 6 O 5 F 8 :Eu 3+ bioprobe through the determination of PSA levels, CV, and the recovery of two serum samples from healthy males upon the addition of PSA standard solutions with different concentrations. The CVs of all the assays are below 8% and the analytical recoveries are in the range of 92-108% ( Table 2), both of which are within the acceptance criteria (CVs # 15%; recoveries in the range of 90-110%) set for bioanalytical method validation. 20 These results verify that the Lu 6 O 5 F 8 :Eu 3+ bioprobe has high reliability and practicability for PSA detection.
To demonstrate the multifunctionality of the oxyuoride nanoprobes, we also employed ultrasmall Lu 6 O 5 F 8 :Yb/Er NPs in proof-of-concept UCL and CT bioimaging, in view of the high R/ G UC emissions and the high mass density of the Lu 6 O 5 F 8 :Yb/Er NPs ($9 g cm À3 ). Aer incubation with human lung cancer (H1299) cells, strong red UC emissions of Er 3+ were clearly visualized in the cells upon 980 nm irradiation, in sharp contrast to the weak green luminescence observed in the dark (ESI Fig. S19a †), which agrees well with the corresponding nearly single-band UC spectrum (Fig. 2e). In addition, as a proof-of-concept experiment, we recorded in vitro color-mapped CT images and measured the CT values using an aqueous solution of Lu 6 O 5 F 8 :Yb/Er NPs. The CT value of NPs at 12.5 mg mL À1 was 319 Hounseld units (HU), which is much higher than that of LiLuF 4 NPs (181 HU) and commercial iopromide (186 HU) at the same concentration (ESI Fig. S19b and c †). 36 To further evaluate the cytotoxicity of the citrate-capped Lu 6 O 5 -F 8 :Yb/Er NPs for potential bioimaging applications, the dark/ photo toxicity of the NPs were measured against HELF cells using a standard methylthiazolyltetrazolium (MTT) assay. There is no signicant cytotoxicity of the NPs (ESI Fig. S20 †), even at a high concentration of 1 mg mL À1 , either in the dark or upon 980 nm irradiation for 2 min. These results demonstrate the potential of the UC NPs as bioprobes for UCL/CT dual-modal bioimaging.