1. Creation of LLUF Probes
- The polyethersulfone membranes (2 cm2) were sealed with triangle polypropylene paddles (University of California, San Diego) by gluing membranes with epoxy on the borders of paddles. A negatively charged polyethersulfone membrane with a molecular weight cut-off (MWCO) at 30 kDa was used.
- A teflon fluorinated ethylene propylene tube (inner diameter/outer diameter, 0.35/0.50 cm) was attached to a cylinder exit of a triangle polypropylene paddle so the LLUF probe can be connected to a 20 ml syringe.
- After soaking the polyethersulfone membrane into human saliva in a culture dish (50 mm diameter), negative pressure was created by withdrawing a syringe. The syringe with negative pressure drove the ultrafiltration process for sampling proteins from saliva.
- The probes were sterilized with 70% alcohol overnight prior to use. To demonstrate proper sealing, we positioned LLUF probes into a solution containing blue dextran (50 mg/ml) with an average molecular mass of 2,000 kDa for 2 h. The absence of blue dextran in the collected samples indicated that no leaks developed during probe fabrication and sampling.
2. Saliva Collection
- Whole saliva was collected from three healthy volunteers (two males and one female between the ages 20 and 40) taking no medications, with no overt signs of gingivitis or cavities 6.
- After rinsing the mouth with water, the whole saliva samples were collected by spitting, without chemical stimulation, into an ice-cooled vessel.
- All samples were pooled and kept on ice during the collection procedure.
- Immediately after the collection, saliva (200 μl) was applied for the sampling with LLUF probes. The sampling was performed at a 4 °C room.
- Saliva proteins (1.0 μg/μl) with or without LLUF probe collection were directly subjected to nano LC-mass spectrometry analysis without tryptic digestion. Protein concentrations were determined using a Bio-Rad Protein Assay 12.
3. NanoLC-LTQ MS Analysis
- The un-digested saliva samples (5 μl) were directly loaded to the trap column of the Eksigent NanoLC system by the autosampler, using 100% buffer A (2% acetonitrile/0.1% formic acid). The NanoLC was on-line coupled with a Finnigan LTQ mass spectrometer.
- After sample loading and washing, the valve was switched and a 500 nl/min linear gradient was delivered to the trap and the separation column (10 cm in length, 100 μm i.d., in-house packed with Synergi 4 μm C18). The gradient was from 0 to 50% buffer B (80% acetonitrile/0.1% formic acid) in 45 min.
- The nanoLC-LTQ MS instruments were operated in the data dependent mode by Xcalibur. MS/MS spectra of the four strongest MS ions above an intensity of 1 × 105 were collected with dynamic exclusion enabled and the collision energy set at 35%.
- Each sample was run twice by the NanoLC-LTQ MS system. Samples from three separate preparations were used. Those peptides detectable in three separate samples were listed in Supplemental Tables 1 and 2. Representative of NanoLC-LTQ MS spectra was illustrated in Figure 3.
4. Data Analysis and Protein Database Searching
- Each RAW file was converted to a mzXML file using Readw.exe.
- The mzXML file was input into a SEQUEST Sorcerer 2 system and searched against a human database generated from the corresponding National Center for Biotechnology Information (NCBI) protein database using non-enzyme specificity. The mass tolerance of precursor ion was set at 1.5 Da. A molecular mass of 16 Da was added to methionine for differential search to account for oxidation.
- After SEQUEST searching, the results were automatically filtered, validated and displayed by PeptideProphet and ProteinProphet [Institute for Systems Biology (ISB)]. PeptideProphet estimates a comprehensive probability (P) score that a peptide assignment is "correct" vs. "incorrect" on the bases of it's SEQUEST scores (Xcorr, ΔCn, Sp, RSp) and additional information of each peptide sequence identified. ProteinProphet computed a probability score from 0 to 1 for each protein on the basis of peptides assigned to MS/MS spectra.
- To minimize false positive identifications we used stringent filter criteria. First, the minimum P score cutoff is set at 0.8 for any accepted peptide to assure very low error (much less than 3%) and reasonably good sensitivity. Second, all peptides with >0.8 P score must have high cross-correlation (Xcorr) scores at the same time: 1.9, 2.2, and 3.0 for +1, +2, and +3 charge.
5. Removal of Oral Bacteria with LLUF Probes
- To determine the capability of LLUF probes to remove the oral bacteria, saliva before and after LLUF probe collection (Section 3) was spread on agar plates for bacterial detection.
- Aerobic bacteria were grown on an antibiotic-free Lauria-Bertani (LB) agar plate at 37 °C for one day.
- Anaerobic bacteria were grown on an antibiotic-free Brucella broth agar plate (BD, Sparks, MD) under anaerobic conditions using Gas-Pak (BD Biosciences, San Jose, CA) at 37 °C for one day.
6. Representative Results
1. Fabrication of LLUF probes and sampling saliva in an imitated oral environment
If sampling saliva can be performed as sucking a lollipop, the procedure will avoid the degradation of spitted saliva in collection devices 13-14. Importantly, it will also become possible to monitor patients dynamically and locally from oral cavities. In addition, if sample preparation using saliva could be simplified, clinicians would easily facilitate the procedure to expedite their decision on the next clinical operation. Mass spectrometry is one of the most sensitive techniques to detect and even sequence proteins in a very short period of time. However, complicated procedures for sample preparation have hampered using this technique in clinic. Furthermore, it is known that higher abundance proteins or proteins with high molecular weights in clinical samples (e.g. amylase in saliva) mask low abundance proteins in mass spectrometric analysis 15-16. To overcome the hurdles mentioned above, we developed a lollipop-like ultrafiltration device named LLUF probes (Figure 1). A negatively charged polyethersulfone membrane with a MWCO at 30 kDa (Figure 1A, a) was glued to a polypropylene paddle (Figure 1A, b). It was positioned in front of the LLUF probe with the intention of filtering out larger proteins in saliva. To mimic the human oral environment (Figure 1A, g), a sponge (Figure 1A, e) was soaked into saliva in a culture dish (Figure 1A, f). After fully withdrawing the syringe (Figure 1A, d), filtered saliva started moving along a connected tube (Figure 1A, c) and was collected.
2. Identification of indigenous saliva peptidome by NanoLC-LTQ MS
Comparing LC chromatograms, we found distinct chromatograms of saliva before and after LLUF sampling (Figure 2), indicating that there are different protein compositions in saliva after LLUF sampling. To determine the protein compositions, we employed NanoLC-LTQ mass spectrometry that is known to be able to promptly sequence peptides from a multiple protein mixture. More importantly, to simplify sample preparation for clinical purposes, whole saliva without chemical or enzymatic digestion was applied for NanoLC-LTQ MS analysis. Unexpectedly, 131 peptides were identified in undigested saliva (Supplemental Table 1). These peptides are fragments derived from various proline rich proteins, actin, alpha amylase, alpha 1 globin, beta globin, histain 1, keratin 1, mucin 7, polymeric immunoglobulin receptor, satherin, and S100A9. Twenty-six unique peptides were identified in saliva after filtering with LLUF probes (Supplemental Table 2). These peptides are fragments mainly derived from various proline-rich proteins. Peptides derived from proteins such as the polymeric immunoglobulin receptor (83.24 kDa) and alpha amylase, were undetectable, demonstrating the capability of LLUF probes in removal of larger and abundant proteins. A MS/MS spectrum of the PFIAIHAEAESKL peptide corresponding to an internal peptide of alpha-amylase is illustrated in Figure 3A. Most intriguingly, after removing larger proteins, 18 of 26 sequenced peptides became detectable in the LLUF probe-sampled saliva (Table 1). These 18 peptides were derived from proline-rich proteins or hypothetical proteins that ended with a proline (P)- glutamine (Q) (-PQ), -SR, -SP or -PP C-terminus. Figure 3B showed a MS/MS spectrum of the PQGPPQQGGHPRPP peptide that was detected in the LLUF probe-sampled saliva. The peptide could be derived from proline-rich protein HaeIII subfamily 1 and 2 (Table 1).

Figure 1. Compositions of LLUF probes and sampling of whole saliva from mimic human oral cavity. Panel A: (a) a semi-permeable polyethersulfone with a MWCO of 30 kDa; (b) a polypropylene paddle; (c) a teflon fluorinated ethylene propylene tube; (d) a 20 ml syringe. Panel B: a sponge (e) (a mimicked tongue) was soaked into a culture dish (f) containing human saliva to create an artificial human oral cavity (g). The resulting negative pressure created by fully withdrawing a syringe drives the collected fluid to move along a connected tube (arrow) and towards a created space (arrowhead) within syringe. Bar: 2.0 cm.

Figure 2. Differential LC/MS/MS chromatograms of saliva before and after LLUF sampling. Proteins (1.0 μg/μl) in human whole saliva were filtered without or with a LLUF probe. Proteins without tryptic digestion were directly subjected to NanoLC- LTQ MS that was conjugated with an Eksigent Nano LC system as described in Material and Methods. The base-peak chromatograms (with 44-mim retention times) of saliva before (A) and after (B) LLUF sampling were illustrated.

Figure 3. Detection of saliva peptidome by NanoLC- LTQ MS sequencing. Saliva proteins before (A) and after (B) sampling with LLUF probes were analyzed by NanoLC- LTQ MS as described in Experimental Procedures. Saliva peptidome derived from natural human saliva without tryptic digestion were demonstrated in Supplemental Tables 1 and 2. A peptide (PFIAIHAEAESKL) derived from alpha-amylase was exclusively detected in a saliva sample without collection with LLUF probes (A), illustrating that the capability of LLUF probes in removing large proteins. Many peptides with -PQ, -SR, -SP or -PP C-termini were solely in samples after ultrafiltration LLUF probes. One (PQGPPQQGGHPRPP) of peptides derived from various proline rich proteins were shown (B). MS/MS spectra with characteristic "y" and "b" series ions confirmed the identities of both peptides.

Figure 4. Removal of oral bacteria using during LLUF probes. A LLUF probe was constructed as described in Materials and methods. The probe was positioned into the human saliva within an imitated oral environment (Figure 1). The syringe at the end of LLUF probe was withdrawn to create a negative pressure that drove the ultrafiltration process for saliva sampling. During sampling, whole saliva crossed selectively through the polyethersulfone membrane and accumulated inside a syringe. Whole saliva before the sampling with a LLUF probe served as a control. Saliva (10 μl) before and after LLUF probe sampling was spread on agar plates for bacterial detection. Panel A: Saliva with (+LLUF) and without (+LLUF) LLUF probe sampling was spread side-by-side on an antibiotic-free LB agar plate at 37 °C for one day. Panel B: Saliva with and without LLUF probe sampling was spread on an antibiotic-free Brucella broth agar plate under anaerobic conditions using Gas-Pak (BD Biosciences, San Jose, CA) at 37 °C for one day. Bacteria did not grow on agar plates spread with LLUF probe-sampled saliva, demonstrating the capability of LLCF probes in eliminating aerobic as well as anaerobic oral bacteria. Bar: 1.0 cm.
| | Peptide sequence/ Measured peptide mass | Accession number | Name |
| 1 | AGNPQGPSPQGGNKPQ GPPPPPGKPQ
2485.3 | gi|41349484 | Proline-rich protein BstNI subfamily 1 isoform 2 precursor |
| | | gi|41349482 | Proline-rich protein BstNI subfamily 1 isoform 1 precursor |
| | | gi|60301553 | Proline-rich protein BstNI subfamily 2 |
| 2 | GGHQQGPPPPPPGKPQ
1576.9 | gi|4826944 | Proline-rich protein HaeIII subfamily 2 |
| | | gi|9945310 | Proline-rich protein HaeIII subfamily 1 |
| 3 | GPPPAGGNPQQPQAPPA GKPQGPPPPPQGGRPP
3126.2 | gi|37537692 | Proline-rich protein BstNI subfamily 4 precursor |
| 4 | GPPPPGGNPQQPLPPPAGKPQ
2028.3 | gi|113423660 | PREDICTED: hypothetical protein |
| 5 | GPPPPGKPQGPPPQGDKSRSP
2077.8 | gi|113423262 | PREDICTED: hypothetical protein isoform 5 |
| | | gi|41349482 | Proline-rich protein BstNI subfamily 1 isoform 1 precursor |
| | | gi|60301553 | Proline-rich protein BstNI subfamily 2 |
| | | gi|113423663 | PREDICTED: hypothetical protein |
| | | gi|41349484 | Proline-rich protein BstNI subfamily 1 isoform 2 precursor |
| | | gi|41349486 | Proline-rich protein BstNI subfamily 1 isoform 3 precursor |
| 6 | GPPPPPPGKPQGPPPQ GGRPQGPPQGQSPQ
2918.5 | gi|9945310 | Proline-rich protein HaeIII subfamily 1 |
| 7 | GPPPQEGNKPQRPPPPGRPQ
2131.3 | gi|113423660 | PREDICTED: hypothetical protein |
| 8 | GPPPQGGNKPQGPPPPGKPQ
2030 | gi|113423663 | PREDICTED: hypothetical protein |
| | | gi|41349482 | Proline-rich protein BstNI subfamily 1 isoform 1 precursor |
| | | gi|113423262 | PREDICTED: hypothetical protein isoform 5 |
| | | gi|60301553 | Proline-rich protein BstNI subfamily 2 |
| 9 | GPPPQGGRPQGPPQGQSPQ
1866.6 | gi|4826944 | Proline-rich protein HaeIII subfamily 2 |
| 10 | GPPQQGGHPPPPQGRPQ
1713.8 | gi|9945310 | Proline-rich protein HaeIII subfamily 1 |
| | | gi|4826944 | Proline-rich protein HaeIII subfamily 2 |
| 11 | GPPQQGGHQQGPPPPPPGKPQ
2083.1 | gi|9945310 | Proline-rich protein HaeIII subfamily 1 |
| | | gi|4826944 | Proline-rich protein HaeIII subfamily 2 |
| 12 | GRPQGPPQQGGHQQGP PPPPPGKPQ
2512.6 | gi|4826944 | Proline-rich protein HaeIII subfamily 2 |
| | | gi|9945310 | Proline-rich protein HaeIII subfamily 1 |
| 13 | NKPQGPPPPGKPQGPP PQGGSKSRSSR
2720.2 | gi|113423262 | PREDICTED: hypothetical protein isoform 5 |
| | | gi|113423663 | PREDICTED: hypothetical protein |
| 14 | PQGPPQQGGHPRPP
1450.1 | gi|9945310 | Proline-rich protein HaeIII subfamily 1 |
| | | gi|4826944 | Proline-rich protein HaeIII subfamily 2 |
| 15 | QGRPQGPPQQGGHPRPP
1791.1 | gi|4826944 | Proline-rich protein HaeIII subfamily 2 |
| | | gi|9945310 | Proline-rich protein HaeIII subfamily 1 |
| 16 | SPPGKPQGPPPQ
1186.9 | gi|113423262 | PREDICTED: hypothetical protein isoform 5 |
| | | gi|41349482 | Proline-rich protein BstNI subfamily 1 isoform 1 precursor |
| | | gi|60301553 | Proline-rich protein BstNI subfamily 2 |
| | | gi|113423663 | PREDICTED: hypothetical protein |
| | | gi|41349484 | Proline-rich protein BstNI subfamily 1 isoform 2 precursor |
| | | gi|41349486 | Proline-rich protein BstNI subfamily 1 isoform 3 precursor |
| 17 | SPPGKPQGPPPQGGNQ PQGPPPPPGKPQ
2720.3 | gi|113423663 | PREDICTED: hypothetical protein |
| | | gi|41349482 | Proline-rich protein BstNI subfamily 1 isoform 1 precursor |
| | | gi|113423262 | PREDICTED: hypothetical protein isoform 5 |
| | | gi|60301553 | Proline-rich protein BstNI subfamily 2 |
| 18 | SPPGKPQGPPQQEGNKPQ
1870.9 | gi|37537692 | Proline-rich protein BstNI subfamily 4 precursor |
Table 1. Peptides were exclusively detectable in LLUF-collected samples.
Supplemental Table 1. Click here to view supplemental Table 1.
Supplemental Table 2. Click here to view supplemental Table 2.