A Criterion of Colorectal Cancer Diagnosis Using Exosome Fluorescence-Lifetime Imaging

This study was aimed to investigate the applicability of the exosome fluorescence-lifetime imaging microscopy (FLIM) for colorectal cancer (CRC) diagnosis. Differential ultra-centrifugation was used to extract exosomes from the blood plasma of 11 patients with colon polyps (CPs) and 13 patients with CRC at the T2-4, N0-3, and M0-1 stages. Analysis was performed using a two-photon FLIM device. In total, 165 and 195 FLIM images were recorded for the CP and CCR patient groups, respectively. Two classes of exosomes differentiated by autofluorescence average lifetime tm were discovered in the samples. The first class of exosomes with tm = (0.21 ± 0.06) ns was mostly found in samples from CRC patients. The second class with tm = (0.43 ± 0.19) ns was mostly found in samples from CP patients. The relative number of “CRC-associated” exosomes Nch in the FLIM dataset was shown to be very small for the CP patient group and large for the CRC patient group. This difference was statistically significant. Therefore, the suggested CRS diagnostics criterion can be as follows. If Nch > 0.5, the probability of CRC is high. If Nch < 0.3, the probability of CRC is low.


Introduction
According to the World Health Organization, colorectal cancer (CRC) is the third most common type of cancer and the second leading cause of cancer-associated deaths among men and women worldwide [1,2]. Usually, four stages of CRC are considered, depending on the five-year survival rate (from 90% to less than 1%). CRC is asymptomatic in the early stages and involves the rapid formation of metastases in the late stages [2][3][4]. Early diagnosis can save up to 90% of CRC patients [4][5][6].

•
In clinical practice, the following methods of CRC diagnostics have become the most widespread [7][8][9]: • Finger rectal examination; • Irrigoscopy (X-ray examination of the intestine); • Test for hidden blood in feces; • Capsule endoscopy (evaluation using an endocapsule with a digital video camera that passes through all parts of the intestine); • Colonoscopy (endoscopic diagnostic method using a colonoscope with a video camera, if necessary with a biopsy); • Cancer markers' blood tests.
With the exception of blood tests, similar diagnostic tools do not provide early CRC detection. As an example, in Russia, the diagnosis of CRC is made at stage IV in 27.7% group comprised patients with colon cancer and upper-rectal cancer. Exclusion criteria for the CRC group were: multiple primary CRC, mid-rectal cancers, and low-rectal cancers. Patients with CPs were considered a control group. Colon polyps have a high risk of malignancy and are considered a precancerous stage but still not a cancer. Therefore, it is important to distinguish between these states to provide an adequate therapy.
The study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the Local Ethics Committee of the Cancer Research Institute of Tomsk National Research Medical Center (Tomsk, Russia) (protocol No.2, 15 January 2020). All participants were informed of the purpose and nature of the treatment and gave their informed consent. All patients with Stage II and III CRC underwent radical surgery (hemicolectomy or colon resection). The stage of the disease was established after surgery in accordance with the international TNM Classification of Malignant Tumors (8th Edition).
Standard vacuum K3 tubes with an ethylene-diamine-tetraacetate anticoagulant were used to collect peripheral blood from the ulnar vein. Exosomes from blood plasma were extracted by ultrafiltration followed by ultracentrifugation according to the following protocol [17,28]. Shaped elements of the 9 mL volume blood samples were deposited by centrifugation for 20 min at 290 g (bucket rotor) and 4 • C. The supernatant was recentrifuged for 20 min at 1200 g (bucket rotor) and 4 • C. To remove cellular debris, plasma samples were centrifuged for 20 min at 17000 g (angular rotor) and 4 • C. To remove vesicles over 100 nm, the supernatant was diluted 5 times with a phosphate-salt buffer (PSB) (10 mM phosphate buffer, 0.15 M NaCl, pH 7.5) and filtered through a filter with a pore diameter of 100 nm (Minisart high flow, 16553-K, "Sartorius", Gottingen, Germany). Exosomes were deposited by ultracentrifugation for 90 min at 100,000 g (bucket rotor) and 4 • C. The precipitate was resuspended in 10 mL of PSB and ultracentrifuged twice under the same conditions. Finally, exosomes were resuspended in 200 µL of PSB, frozen in liquid nitrogen and stored at −80 • C [29]. This protocol had been used for the extraction of various exosome types and, according to the results of nano-tracking analysis, provided the following exosome concentrations in term of median and range values × 10 7 /mL: [16 (8; 20)] for healthy volunteers, [24 (20; 138)] for benign breast diseases patients, [21 (10; 180)] for breast cancer patients, and [22 (13; 154)] for ovarian cancer patients [28]. The variation was not very high. Therefore, these values could be used as a estimation of exosomes concentration in our case.
The samples were analyzed in the Biophotonics Laboratory of Tomsk State University. Before transportation, the sample storage temperature was slowly increased from −80 • C to −20 • C to avoid sticking [29]. The samples were delivered in a special thermal box containing two exosome samples surrounded by two packages of a Green Glade cold storage battery of a 750 mL volume. The temperature in the container was −20 • C. After the container was opened, the exosome samples were placed on ice and allowed to thaw at a 0 • C temperature.
Immediately before the measurements, an exosome sample was stirred using an OPn-8 centrifuge (Dastan, Kyrgyzstan) for 5 min at 1000 g and a temperature of 4 • C. Exosomes were analyzed using a two-photon MPTflex microscope (Jenlab GmbH, Jena, Germany), which has a spatial resolution of about 0.5 microns (horizontal) and 2 microns (vertical). We used a lens with a 1.3 numerical aperture and a 40× magnification. The pump laser pulse duration was about 200 fs, the repetition rate was 80 MHz, and the wavelength was 760 nm. The detector has a spectral filter in the range of 406-660 nm, suitable for TPAF signal registration. The MPTflex microscope has an option for TPAF registration in an FLIM mode.
In order to avoid the photobleaching of exosomes, the laser power was limited to 5 mW according to the melanin vivo visualization protocol [30].
A glass plate with a thickness of 100-170 µm was placed on a metal ring. This glass was covered by an adhesive tape about 7 µm thick with a cutout in the form of a round hole ( Figure 1a). Next, an exosome sample was applied to this glass in the hole area ( Figure 1b). Then, an analogous glass plate covered this construction from above ( Figure 1c).
A glass plate with a thickness of 100-170 μm was placed on a metal ring. This glass was covered by an adhesive tape about 7 μm thick with a cutout in the form of a round hole ( Figure 1a). Next, an exosome sample was applied to this glass in the hole area ( Figure 1b). Then, an analogous glass plate covered this construction from above ( Figure  1c).
(a) (b) (с) Figure 1. An illustration of an exosome sample process prepared for MPTflex microscope analysis: photographs of a glass plate with adhesive tape (a), the process of the exosome sample's positioning on the glass plate (b), and this construction being covered from above with another glass plate (с).
For each sample, FLIM images from depths of about 3-4 μm with sizes of 18 × 18 μm, 52 × 52 μm, and 118 × 118 μm were recorded in five randomly selected spatial positions. Therefore, in total, 165 and 195 FLIM images were recorded for CP and CRC patients, respectively. The FLIM signal (t) registered by a pixel (or group of pixels) of a matrix detector is usually approximated by a multi-exponential decomposition as follows: where α i and τ are the amplitudes and lifetimes of the TPAF signal components, respectively, and (0) is the intensity of the measured fluorescence signal at an initial moment. As a rule, analysis is limited by two members of the series [31,32]: (t) = a 1 e −t τ 1 ⁄ + a 2 e −t τ 2 ⁄ , Average fluorescence lifetime τ m is often used for TPAF signal characterization: A phasor-plot approach for FLIM data description and analysis was also used [33,34]. The FLIM data projection on a phasor plane is carried out by the Fourier transform of the FLIM signal (t) : where is a frequency. Let us introduce functions ( ) and s( ), characterizing the real and imaginary parts of ( ): A suitable presentation of the FLIM signal (t) was shown to be achieved by using ( ) and s( ) at the frequency 0 , where 0 is the laser pulses' repetition angular frequency [35]. The resulting complex value ( 0 ) -• s ( 0 ) is presented on a two-dimensional graph, where the ordinate represents the imaginary part and the abscissa represents the real one. In this approach, an FLIM curve is assigned a point on the phasor plot.
Statistical analysis was performed using the nonparametric Mann-Whitney U tests [36]. For each sample, FLIM images from depths of about 3-4 µm with sizes of 18 × 18 µm, 52 × 52 µm, and 118 × 118 µm were recorded in five randomly selected spatial positions. Therefore, in total, 165 and 195 FLIM images were recorded for CP and CRC patients, respectively. The FLIM signal I(t) registered by a pixel (or group of pixels) of a matrix detector is usually approximated by a multi-exponential decomposition as follows: where α i and τ i are the amplitudes and lifetimes of the TPAF signal components, respectively, and I(0) is the intensity of the measured fluorescence signal at an initial moment. As a rule, analysis is limited by two members of the series [31,32]: Average fluorescence lifetime τ m is often used for TPAF signal characterization: A phasor-plot approach for FLIM data description and analysis was also used [33,34]. The FLIM data projection on a phasor plane is carried out by the Fourier transform of the FLIM signal I(t) : where ω is a frequency. Let us introduce functions g(ω) and s(ω), characterizing the real and imaginary parts of T(ω) : A suitable presentation of the FLIM signal I(t) was shown to be achieved by using g(ω) and s(ω) at the frequency ω 0 , where ω 0 is the laser pulses' repetition angular frequency [35]. The resulting complex value g(ω 0 ) − j · s(ω 0 ) is presented on a two-dimensional graph, where the ordinate represents the imaginary part and the abscissa represents the real one. In this approach, an FLIM curve is assigned a point on the phasor plot.
Statistical analysis was performed using the nonparametric Mann-Whitney U tests [36].

Results
Examples of the average TPAF lifetime t m image for randomly selected CP and CRC patients are shown in Figure 2.

Results
Examples of the average TPAF lifetime t m image for randomly selected CP and CRC patients are shown in Figure 2. The TPAF lifetime distribution functions for images presented in Figure 2 are shown in Figure 3. The small values of the distribution functions presented in Figure 3 correspond to spatial areas that emitted a small number of photons. To focus on spatial areas with large values of emitted TPAF photons and which definitely corresponded to exosomes, we applied a cutting filter with a threshold of 800 photons per pixel (see Figure 4).  The small values of the distribution functions presented in Figure 3 corresp spatial areas that emitted a small number of photons. To focus on spatial areas wi values of emitted TPAF photons and which definitely corresponded to exosom applied a cutting filter with a threshold of 800 photons per pixel (see Figure 4). The small values of the distribution functions presented in Figure 3 correspond to spatial areas that emitted a small number of photons. To focus on spatial areas with large values of emitted TPAF photons and which definitely corresponded to exosomes, we applied a cutting filter with a threshold of 800 photons per pixel (see Figure 4).
The results of the cutting filter application presented in a phasor plane are shown in Figure 5.  The results of the cutting filter application presented in a phasor plane are shown in Figure 5.   The results of the cutting filter application presented in a phasor plane are shown in Figure 5.  We see that phasor-plot data points for the CRC patient form two spatially separated areas and that there is only one analogous area for the CP patient. To quantitatively describe these areas, the k-means clustering method [37] was applied to the image shown in Figure 5c. The number of classes equal to two was used as an input parameter. To estimate the radii of classes, we used the average value of the Euclidean distance from each point of the set to the center plus its tripled standard deviation (the 3σ rule): where r i,j = g c,j − g i,j 2 + s c,j − s i,j 2 , g i,j and s i,j are coordinates of a data point in the j − th class, and g c,j , s c,j , and N j are coordinates of the center and number of data points in j − th class. The class parameters are: g c,2 =0.53, s c,2 = 0.23, R 2 = 0.71.
The class areas on the phasor plots presented in Figure 5 are shown in Figure 6. We see that phasor-plot data points for the CRC patient form two spatially separated areas and that there is only one analogous area for the CP patient. To quantitatively describe these areas, the k-means clustering method [37] was applied to the image shown in Figure 5c. The number of classes equal to two was used as an input parameter. To estimate the radii of classes, we used the average value of the Euclidean distance from each point of the set to the center plus its tripled standard deviation (the 3σ rule): where , = √( , − , ) 2 +( , − , ) 2 , , and , are coordinates of a data point in the − ℎ class, and , , , , and are coordinates of the center and number of data points in − ℎ class. The class parameters are: с,1 = 0.80, ,1 = 0.14, 1 = 0.51, с,2 = 0.53, ,2 = 0.23, 2 = 0.71.
The class areas on the phasor plots presented in Figure 5 are shown in Figure 6.
(a) (b) (c) (d) Figure 6. The class areas on phasor plots of the average TPAF lifetime t m images of the exosome samples presented in Figure 5: the first class position on the phasor plot for the CRC patient (a) and the CP patient (b), the second class position on the phasor plot for the CRC patient (c) and the CP patient (d). The first class parameters: с,1 = 0.80, ,1 = 0.14, and 1 = 0.51; the second class parameters: с,2 = 0.53, ,2 = 0.23, and 2 = 0.71. Here, , , , , and are coordinates of the center and radius of the − ℎ class.
The differentiation of exosomes presented in Figure 4 on the classes is shown in Figure 7.  Figure 5: the first class position on the phasor plot for the CRC patient (a) and the CP patient (b), the second class position on the phasor plot for the CRC patient (c) and the CP patient (d). The first class parameters: g c,1 = 0.80, s c,1 = 0.14, and R 1 = 0.51; the second class parameters: g c,2 = 0.53, s c,2 = 0.23, and R 2 = 0.71. Here, g c,j , s c,j , and R j are coordinates of the center and radius of the j − th class. The differentiation of exosomes presented in Figure 4 on the classes is shown in Figure 7.  (6) and (7). Here, color data points correspond to the exosomes with short average TPAF lifetimes and gray data points correspond to the exosomes with long average TPAF lifetimes. These images are presented unprocessed in Figure 4.
We see that exosomes with short TPAF lifetimes are practically absent on the FLIM image for the CP patient, whereas the opposite situation is true for exosomes with long TPAF lifetimes. Therefore, the class parameters described by Formulas (6) and (7) could differentiate exosomes specific for CRC and CP patients. To test this hypothesis, the sequential application of the cutting filter and the circle masks Formulas (6) and (7) in a phasor plane were implemented for the whole dataset. The corresponding TPAF lifetime distribution functions are shown in Figure 8. Their parameters are presented in Table 1.   (6) and (7). Here, color data points correspond to the exosomes with short average TPAF lifetimes and gray data points correspond to the exosomes with long average TPAF lifetimes. These images are presented unprocessed in Figure 4.
We see that exosomes with short TPAF lifetimes are practically absent on the FLIM image for the CP patient, whereas the opposite situation is true for exosomes with long TPAF lifetimes. Therefore, the class parameters described by Formulas (6) and (7) could differentiate exosomes specific for CRC and CP patients. To test this hypothesis, the sequential application of the cutting filter and the circle masks Formulas (6) and (7) in a phasor plane were implemented for the whole dataset. The corresponding TPAF lifetime distribution functions are shown in Figure 8. Their parameters are presented in Table 1.  (6) and (7). Here color data points correspond to the exosomes with short average TPAF lifetimes and gray data points correspond to the exosomes with long average TPAF lifetimes. These images are presented unprocessed in Figure 4.
We see that exosomes with short TPAF lifetimes are practically absent on the FLIM image for the CP patient, whereas the opposite situation is true for exosomes with long TPAF lifetimes. Therefore, the class parameters described by Formulas (6) and (7) could differentiate exosomes specific for CRC and CP patients. To test this hypothesis, the se quential application of the cutting filter and the circle masks Formulas (6) and (7) in a phasor plane were implemented for the whole dataset. The corresponding TPAF lifetime distribution functions are shown in Figure 8. Their parameters are presented in Table 1.   Figure 8.

Mean Value, ns Standard Deviation, ns
The short average TPAF lifetime distribution 0.21 0.06 The long average TPAF lifetime distribution 0.43 0. 19 We constructed interval estimates for the distributions described by the parameters presented in Table 1, and the results are shown in Figure 9.
Diagnostics 2022, 12, x FOR PEER REVIEW 9 of 12 We constructed interval estimates for the distributions described by the parameters presented in Table 1, and the results are shown in Figure 9.  Figure 8. These curves were calculated using Gaussian approximation of the data presented in Table 1. Notations on the legend: CPs-the exosome samples from the whole CRC patient group; CRC-the exosome samples from the whole CP patient group.
Using these interval approximations, we could construct a CRS diagnostics criterion as follows: 1. For every exosome in an FLIM image, we calculated the TPAF average lifetime t m . If the t m value fell within the intersection of the intervals corresponding to short and long lifetime interval approximations, then this exosome was ignored. Otherwise, we assigned this exosome the appropriate index: "h", corresponding to a CP-associated exosome, or "c", corresponding to a CRC-associated exosome. 2. For a whole FLIM dataset corresponding to a definite participant, we calculated the ratio as follows: where сℎ is the relative number of CRC-associated exosomes in the FLIM dataset, is the number of CRC-associated exosomes, and ℎ is the number of "non-cancerous" (CP-associated) exosomes. If the ℎ value is large, it is a sign of possible CRC presence.
The results of applying this criterion to CP and CRC patient groups are presented in Figure 10.  Figure 8. These curves were calculated using Gaussian approximation of the data presented in Table 1. Notations on the legend: CPs-the exosome samples from the whole CRC patient group; CRC-the exosome samples from the whole CP patient group.
Using these interval approximations, we could construct a CRS diagnostics criterion as follows: 1.
For every exosome in an FLIM image, we calculated the TPAF average lifetime t m . If the t m value fell within the intersection of the intervals corresponding to short and long lifetime interval approximations, then this exosome was ignored. Otherwise, we assigned this exosome the appropriate index: "h", corresponding to a CP-associated exosome, or "c", corresponding to a CRC-associated exosome.

2.
For a whole FLIM dataset corresponding to a definite participant, we calculated the ratio as follows: where N ch is the relative number of CRC-associated exosomes in the FLIM dataset, N c is the number of CRC-associated exosomes, and N h is the number of "non-cancerous" (CP-associated) exosomes. If the N ch value is large, it is a sign of possible CRC presence.
The results of applying this criterion to CP and CRC patient groups are presented in Figure 10.
The relative number of CRC exosomes in the FLIM dataset was very small in the CP patient group, whereas the opposite situation was observed in the CRC patient group. In order to evaluate the statistical difference between the CP and CRC patient groups in terms of CP-and CRC-associated exosome content in blood plasma samples, the empirical value of the Mann-Whitney criterion U c was calculated [35]. Consider the hypothesis H 0 about the insignificance of differences between groups. The hypothesis is accepted if the critical value U c is less than the empirical U e . In our case, for the probability α = 0.01, the critical value was U c = 34 and the empirical value was U e = 1, i.e., the H 0 hypothesis was not accepted.
where сℎ is the relative number of CRC-associated exosomes in the FLIM dataset, is the number of CRC-associated exosomes, and ℎ is the number of "non-cancerous" (CP-associated) exosomes. If the ℎ value is large, it is a sign of possible CRC presence.
The results of applying this criterion to CP and CRC patient groups are presented in Figure 10.

Conclusions
Two classes of exosomes differentiated by average TPAF lifetime t m were discovered in blood plasma samples from CP and CRC patients. Exosomes with short TPAF lifetimes were practically absent in blood plasma samples from the CP patients, whereas the opposite situation was the case for exosomes with long TPAF lifetimes. The TPAF average lifetime t m was shown to be (0.2445 ± 0.0054) ns for CRC-associated exosomes and (0.4071 ± 0.0114) ns for CP-associated exosomes.
Therefore, the CRS diagnostics criterion can be implemented as follows. For an exosome sample from a studied participant, the TPAF average lifetime t m can be measured and the relative number of CRC-associated exosomes in the FLIM dataset N ch can be calculated considering the t m intervals for CRC-associated exosomes and "non-cancerous" (CP-associated) exosomes mentioned above. According to the data presented in Figure 10, if N ch > 0.5, the probability of CRC is high, and if N ch < 0.3, the probability of CRC is low. As N ch changes continuously in the interval [0, 1], collecting larger datasets potentially allows one to construct a scale of CRC severity in terms of N ch values. The nature of the suggested criterion based on peculiarities of cancer-associated exosomes' chemical cargo, reflected in FLIM data, needs aditional research.  Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.