Serum paraoxonase 1 activity in cats: analytical validation, reference intervals, and correlation with serum amyloid A and alpha-1-acid glycoprotein

Abstract

Paraoxonase 1 (PON1) is an inflammation marker associated with lipid oxidation and is used as a diagnostic marker in people. There is no information about the suitable substrate and analytic performance in cats, or its biological behavior compared with other inflammation markers. Our aims were to validate a paraoxon-based method to measure PON1 activity in feline serum, to assess stability of PON1 under different storage conditions and the impact of interfering elements, to determine a reference interval (RI) for healthy cats, and to correlate PON1 activity with 2 major acute-phase proteins. Intra- and inter-assay precision, accuracy, and RI were assessed using fresh serum. The same specimens were stored at room temperature, refrigerated, or frozen, and retested at defined intervals. Hemolysis, lipemia, and icterus were simulated to study interferences. PON1 results were compared to serum amyloid A (SAA) and alpha-1-acid glycoprotein (AGP) results. Analytical validation yielded precise and accurate results. PON1 activity is stable for up to 24 h at room temperature and up to 48 h at 4°C. Freezing at −20°C results in an increase after 72 h, with return to baseline values after 1 wk, that again increases after 6 mo. Only hyperlipemia interfered with PON1 activity. The RI based on 71 healthy cats was 58–154 U/L. PON1 activity was negatively correlated with AGP, but not with SAA. Serum PON1 activity can be measured accurately in cats, and it acts as a negative acute-phase protein.

Introduction

Acute-phase proteins (APPs) are synthesized in the liver in response to pro-inflammatory cytokines; the use of APPs for diagnostic purposes in animals has increased greatly within the last decade.³¹ Serum amyloid A (SAA) is the major positive APP in cats, given that it may increase up to 1,000-fold during inflammation⁵⁵; alpha-1-acid glycoprotein (AGP) is a moderately positive APP that increases 5- to 10-fold during the acute-phase response.²⁰ Recently, the activity of the enzyme paraoxonase 1 (PON1) has drawn attention from researchers in veterinary medicine.^{14,21,46,47,51}

PON1 is in a family of serum enzymes with PON2 and PON3 (International Union of Biochemistry and Molecular Biology, IUBMB Enzyme Nomenclature EC 3.1.8.1, https://www.qmul.ac.uk/sbcs/iubmb/enzyme/EC3/1/8/), which are found in a variety of mammalian species.^2,46,51,53 The 3 enzymes share some biological roles: they have antioxidative properties,^45,49 inhibit the proliferation of Pseudomonas,^41,53 provide protection against inflammation,¹⁸ and limit lipid peroxidation.³⁷ However, there are remarkable differences among the 3 enzymes: PON2 exerts its activity at a cellular level and is not found in plasma,³⁸ and neither PON2 nor PON3 metabolize organophosphates.^10,11 PON1 is a highly promiscuous enzyme, capable of hydrolyzing a wide range of substrates. Within the paraoxonase family, PON1 is the most studied and clinically relevant. The biological hydrolytic activity of PON1 has been identified as lipolactonase, which catalyzes 3 major reactions: biotransformation of lipid lactones (arylesterase); clearance of toxic organophosphates (such as paraoxon, a metabolite of parathion, which gives the name to the enzyme family, phosphotriesterase); and participation in the metabolism of homocysteine (lactonase).^36,39 Hence, PON1 hydrolyzes a variety of synthetic and natural substrates, and genetic polymorphism of PON1 influences enzyme activity in humans^10,35 and rabbits.⁶⁰ PON1 is synthesized in the liver and intimately related with high-density lipoproteins (HDLs). HDLs facilitate PON1 secretion from the liver, stabilizing and providing it a hydrophobic environment for better functionality.²⁹ Given its ability to hydrolyze peroxide phospholipids, PON1 protects low-density lipoproteins (LDL) from oxidation. Oxidative stress and inflammation are strongly linked: reactive oxygen species can directly attack the lipoid matrix of biological membranes, stimulate arachidonic acid metabolism, and thereby enhance neutrophil accumulation and adherence to the capillary wall. In turn, activated neutrophils and monocytes release inflammatory mediators that catalyze the production of reactive oxidants. Oxidative damage to membrane phospholipids is the starting point for lipid peroxidation that, in turn, may react with proteins, changing their conformation and function, leading to an enhanced inflammatory response.³⁷ Given the antioxidant role of PON1, reduced PON1 activity has been reported in both acute and low-grade inflammatory conditions in both humans^1,32,40 and animals.^9,14,20,46

Although the acute-phase response in cats can be assessed using other APPs, there are several reasons to investigate new inflammation markers. The sensitivity and specificity of acute inflammation markers is variable, and their agreement, with few exceptions, is generally low.⁵⁸ PON1 may have a different response, compared with other APPs, when inflammation is associated with oxidative stress,^1,32 as suggested in investigations in cats using different analytical methods.⁵⁶ Among feline APPs, AGP is usually measured with a radial immunodiffusion technique,⁴³ which is impractical in a diagnostic laboratory given that a maximum of 10 samples can be tested on a plate and the manual technique is time consuming. SAA assays are expensive and not widely offered by diagnostic laboratories.

Therefore, our aims were (1) to validate a paraoxon-based enzymatic method to measure PON1 activity in feline serum; (2) to assess the preanalytical variability resulting from different storage conditions and the presence of interfering substances; (3) to establish reference intervals (RIs) in clinically healthy cats; and (4) to correlate PON1 activity with the concentrations of AGP and SAA.

Materials and methods

Animals and specimens

Specimens (423 feline sera) were sent by veterinary practitioners to the University of Milan Veterinary Teaching Hospital (UM-VTH; Lodi, Italy) for AGP measurement, as part of their diagnostic process. Specimens used to assess interference and effects of storage (n = 13) were processed just after collection from cats referred to the UM-VTH for routine check-up visits or for diagnostic purposes. Specimens used to establish RIs (n = 71) were collected as described below; all other sera were frozen at −20°C immediately after collection, stored no longer than 15 d, and delivered under controlled temperature to the VTH (n = 339). Once received, specimens were kept at room temperature (20–22°C) until they were thawed, vortexed, and used for AGP measurement, then sera were frozen at −20°C for a maximum of 6 mo, before measuring PON1 activity and SAA at the same time to avoid further freeze–thaw cycles. SAA concentration was measured in only 82 specimens because of insufficient specimen volume in several cases. Because this was an analytical validation study focused on methodologic aspects rather than on the diagnostic accuracy of PON1, clinical data and diagnostic information were not needed, except for cats enrolled for the calculation of RIs, which must be healthy.

In order to establish a RI, 71 specimens were collected from client-owned cats admitted to the UM-VTH for routine visits or for diagnostic purposes. Blood was collected from the cephalic or jugular vein and transferred immediately to plain tubes to obtain serum by centrifugation (2,500 × g for 15 min). Serum specimens were processed immediately. The only inclusion criterion for cats in this group was that they must be clinically healthy. Health assessment was based on the lack of clinical or laboratory abnormalities as determined by physical examination and a routine panel of laboratory tests including a complete blood count (XT 2000-iV hematology analyzer; Sysmex) and evaluation of creatinine, urea, glucose, alanine aminotransferase, alkaline phosphatase, and total protein (Daytona chemistry analyzer; Randox). Consequently, we excluded cats from this part of the study if they had a recent history of disease or physiologic conditions potentially affecting blood results (pregnancy, lactation, etc.), or that received any medication or surgical treatments.^28,57

Given that blood was collected for routine diagnostic purposes and the owners had signed a consent form that authorized the use of excess specimens for research purposes, according to the regulations of the UM (EC decision 02-2016), formal approval from the Ethical Committee was not required.

Measurement of PON1 activity

Serum PON1 activity was measured with an automated spectrophotometer (Cobas Mira; Roche Diagnostic), using the organophosphate paraoxon as substrate to ensure that only PON1 and no other enzymes were measured. The assay has been validated previously in dogs,⁴⁴ as a modification of an enzymatic method.¹⁵ The reaction buffer was prepared using glycine buffer (0.05 mM, pH 10.5) containing 1 mM of O,O-diethyl O-(4-nitrophenyl) phosphate as a substrate (paraoxon-ethyl purity > 90%; MilliporeSigma) and 1 mM of CaCl₂. The reaction was initiated with 6 µL of specimen, 89 µL of distilled water, and 100 µL of reaction buffer, at 37°C. The rate of hydrolysis of paraoxon to p-nitrophenol was measured by monitoring the increase in absorbance at 504 nm using a molar extinction coefficient of 18,050 L/mol/cm⁻¹. A PON1 activity of 1 U/L was defined as 1 µmol of p-nitrophenol formed per minute under the assay conditions.

Measurement of AGP and SAA

Feline AGP was measured using a single radial immunodiffusion (SRID) kit (Tridelta Development) as described previously.⁴² Feline SAA was measured using a human immunoturbidimetric assay (LZ test SAA; Eiken Chemical), validated previously in cats,²⁴ on an automated spectrophotometer (ILAB 300 Plus; Instrumentation Laboratory).

Validation of PON1 activity measurement

According to recommended procedures,^16,61 intra-assay precision was determined with 20 consecutive measurements of feline pooled sera with low, medium, and high PON1 activity. After a single PON1 activity measurement in each sample, various volumes of serum from 18 specimens with low PON1 activity were merged to create a pool with low concentration, 22 specimens were merged to create a pool with medium concentration, and 15 specimens were merged to create a pool with high concentration (Suppl. Table 1). The inter-assay precision was determined in triplicate on the same pools aliquoted, frozen, and thawed just before the analyses, on 9 consecutive working days. Mean, standard deviations (SD), and coefficients of variation (CV = SD/mean × 100) were calculated (Table 1).

Table 1.

Intra-assay precision of paraoxonase 1 (PON1) activity determined with 20 consecutive repeat measurements in feline frozen serum pools with low, medium, and high activity. Inter-assay precision of the PON1 activity determined on frozen aliquots of feline serum pools with low, medium, and high PON1, measured in triplicate on 9 consecutive working days.

	Intra-assay			Inter-assay
	Low	Medium	High	Low	Medium	High
Min.–max. (U/L)	20–21	59–61	108–123	19–22	54–60	114–123
Mean (U/L)	21	59	117	21	58	118
SD (U/L)	0.4	0.6	4.4	1.0	2.1	3.3
CV (%)	1.9	1.0	3.7	4.9	3.5	2.8

CV = coefficient of variation; SD = standard deviation.

Neither an accepted gold standard to measure feline PON1 activity, nor purified feline PON1, is available commercially. Therefore, the accuracy was indirectly determined by linearity under dilution (LUD) and spike recovery tests (SRTs). The LUD test was performed by measuring PON1 activity in triplicate on a pooled specimen created by merging 13 specimens with high PON1 activity (Suppl. Table 1), serially diluted with distilled water to cover the 90–10% range of the naive pool (100%). The use of distilled water instead of saline could emphasize the matrix effect because distilled water could modify the electrolyte concentrations and the pH, which may be involved in enzyme activity. The mean of the triplicate measurements, the 95% confidence interval (CI), and prediction interval (PI) were calculated at each point.

The PI is a range of values that predicts the value of a new observation, based on the existing model. The PI must account for both the uncertainty in estimating the population mean and the random variation of the individual values. Thus, the PI is wider than the CI, and it is not affected by the sample size. The PI formula is:

$$\widehat{y}\pm t_{\alpha /2,\mathrm{n-2}}\times S_{\in }\times \sqrt{1+\frac{1}{n}+\frac{n{\left(x-\overline{x}\right)}^2}{n\sum x^2-{(\sum x)}^2}}$$

in which, n is the number of the observations, $\overline{x}$ is the sample mean, $S\in$ is the standard error, and t_α/2 is the value of the student t distribution for n-2 degree of freedom.

Given the lack of a standard, it was not possible to perform a proper SRT. However, to confirm the accuracy of the method, after the LUD test, a modified SRT was run by adding increasing percentages (10–100%) of the pool with high PON1 concentration used for the LUD test into a pool created by merging 4 specimens with low PON1 concentration (Suppl. Table 1). Mean PON1 activity of each spiked specimen was calculated from the measurement in triplicate. Although this modification does not allow estimation of the contribution of the matrix effect (i.e., the effect of other substances dissolved in the high PON1 pool) to possible proportional systematic error, such a modification is recommended when a standard solution is not available for spiking³⁰ and is performed commonly.^34,48

PON1 activity storage stability

Storage effects were evaluated in specimens centrifuged just after collection from cats referred to the UM-VTH for routine check-up visits or for diagnostic purposes. PON1 activity was measured in triplicate immediately after centrifugation, and in aliquots stored for 12 and 24 h at room temperature (RT, 20–25°C; n = 4); 24, 48, and 72 h refrigerated at 4°C (n = 4); 24 h, 72 h, and 1 wk in samples frozen at −20°C (short-term freezing, n = 5); or 1 wk, 1 mo, and 6 mo in samples frozen at −20°C (long-term freezing, n = 8). Before analysis, all specimens were thawed at RT and thoroughly mixed using a vortex mixer before any analysis. Mean values recorded after storage were compared with those recorded in fresh specimens.

Interference studies

The effect of hemolysis, hyperbilirubinemia, and hyperlipemia was assessed as described previously.⁴⁷ Aliquots of the pooled serum with medium PON1 activity were spiked with different concentrations of bilirubin (Bil; Fluka, Sigma-Aldrich), a commercial fat emulsion (Trig; Lipofundin S 20%, Braun Milano), or hemoglobin (HGB; Merck), followed by triplicate measurements of PON1 activity. Interfering substances were added to each aliquot of the pooled specimen to obtain final concentrations consistent with slight, moderate, severe, or extreme icterus (Bil = 2.74 mmol/L; 5.31 mmol/L; 10.8 mmol/L; 21.4 mmol/L), lipemia (Trig = 1.4 mmol/L; 2.8 mmol/L; 5.6 mmol/L; 11.3 mmol/L), and hemoglobinemia (HGB = < 6.9 g/L; 13.8 g/L; 27.5 g/L; 55 g/L). Interferograms were prepared in order to display increasing concentrations of interferents on the x-axes, and on the y-axes the percentage changes (final/original × 100%) of PON1 activity.^22,23 Baseline values were determined by adding the same volume of distilled water as the volume of interfering solutions.

Reference interval

A single measurement of PON1 activity was conducted on each of the 71 serum specimens from healthy cats included in this part of the study. The calculation of RIs is described under the statistical analysis section.

Correlation among PON1, AGP, and SAA

To investigate the role of PON1 activity as a negative APP in cats, PON1, AGP, and SAA were correlated. The AGP concentration is < 0.0125 mmol/L (0.5 mg/mL) in healthy cats, > 0.0125 mmol/L in inflammation, and > 0.0375 mmol/L (1.5 mg/mL) in severe inflammation or feline infectious peritonitis (FIP).^12,20,43 PON1 activity of cats with normal AGP (≤ 0.0125 mmol/L) or high AGP (> 0.0125 mmol/L) were compared to each other. Then, the PON1 activity of cats with normal-to-medium (0.0126–0.0375 mmol/L) or very high AGP (> 0.0375 mmol/L) were compared.

Similarly, results of PON1 activity recorded in specimens with normal and high SAA, using the cutoff of 26.7 mg/L defined in a previous study, were compared.^24,42 Finally, the PON1 activity of cats with normal (< 26.7 mg/L), medium (26.7–135.0 mg/L), or very high SAA (> 135.0 mg/L, ~ 5-fold the cutoff) were compared.

Statistical analyses

Statistical analyses were performed using the Analyse-it software (v.4.90.1; Microsoft); RIs were determined with an Excel spreadsheet (Microsoft) with the Reference Value Advisor (v.2.0)¹⁹ that performs computations according to the International Federation of Clinical Chemistry and Laboratory Medicine (IFCC) and Clinical and Laboratory Standards Institute (CLSI) recommendations as suggested by the American Society for Veterinary Clinical Pathology (ASVCP) guidelines.¹⁷ A p value ≤ 0.05 was considered statistically significant. Independently of statistical significance, we considered values as “clinically acceptable” that were ±15% compared with baseline values in interference and storage studies.

The correlation between observed and expected results of LUD and SRT was assessed using a least squares regression test. This test assesses the probability of the null hypothesis corresponding with the slope of the regression being zero (the regression equation y = ax + b becoming y = b), as if both variables y and x were independent. A p value ≤ 0.05 allows rejection of this null hypothesis, confirming that the variables are related. For each storage condition (RT, 4°C, and −20°C), results obtained at different times were compared with those of fresh specimens (T0) with a Friedman test followed by Wilcoxon signed-rank test to evaluate the differences between T0 and every other time. Results obtained with and without different concentrations of interferents were compared using a Friedman test. The correlations between PON1 and AGP or SAA and between AGP and SAA were assessed using a Spearman correlation test. Significant differences in PON1 activity between cats with normal and high AGP or SAA were assessed using a Mann–Whitney U test. Kruskal–Wallis and Bonferroni post-hoc tests were used to assess differences between PON1 activity recorded in cats with normal, medium, or very high concentrations of AGP or SAA. The possible differences in the proportions of cats with PON1 activity lower than the lower reference limit recorded in cats with normal versus high AGP or SAA concentration, as well as in cats with normal versus medium versus very high AGP or SAA concentration, was assessed using the Pearson chi-square test.

For the establishment of RIs, descriptive statistics, tests of normality according to the Anderson–Darling method with histograms and Q-Q plots, and Box–Cox transformation were calculated. The Dixon–Reed and Tukey tests were used to identify outliers. Following ASVCP guidelines,¹⁷ outliers that were considered “suspicious” by the software were retained. Conversely, far outliers were removed from the analysis. RIs were calculated for the whole cat sample group and for the male subgroup using a nonparametric method after Box–Cox transformation to allow the calculation of the 90% CI of RIs. For the calculation of RIs in the female subgroup, a robust method was used after Box–Cox transformation because the sample size was < 40. Serum PON1 activities recorded in male cats were compared with those recorded in female cats with a Mann–Whitney U test, to assess whether it would be advisable to establish separate RIs for the partitioned groups. Possible age-related differences were investigated by linear regression using the Reference Value Advisor (v.2.0) macroinstructions mentioned above.¹⁹

Results

Validation of PON1 activity

The imprecision of the assay, quantified by the CVs of the intra- and inter-assay measurements, was always < 5%, irrespective of the serum PON1 activity (Table 1). The correlation coefficient for SRT (equation y = 1.007x + 0.5292) was r² = 0.999 and for LUD (equation y = 1.057x − 6.799) was 0.994, with p < 0.001 for both tests (Fig. 1). The ranges of values for the 95% CIs and PIs were narrower for the SRT than for the LUD test (Table 2). The total error was minimal for the SRT at any point but was higher for the LUD test; particularly when the dilution was > 50% (Table 3).

Figure 1.

A. Linearity under dilution (LUD) and B. spike recovery test (SRT) for feline paraoxonase 1 (PON1). LUD was measured on a pool of 13 feline sera with high PON1 activity (122 U/L) progressively diluted (100–0%) with distilled water. SRT was conducted on a pool of 4 feline sera with low PON1 activity (11 U/L) spiked with increasing amounts of the LUD pool. Each data point indicates the mean of a triplicate measurement. The solid lines indicate the linear fit, the dotted lines represent the 95% confidence interval. The dashed lines represent the 95% prediction interval.

Table 2.

Linearity under dilution (LUD) results, with upper and lower 95% confidence and prediction intervals, evaluated in triplicate measurements of pooled feline paraoxonase 1 (PON1) activity of 13 frozen (−20°C) specimens.

LUD	PON1 activity (U/L)					Total error (U/L)
	Mean	CI		PI
		Lower 95%	Upper 95%	Lower 95%	Upper 95%
100%	121	119	123	112	130	1.5
90%	109	107	111	100	118	1.3
80%	97	96	97	89	105	0.8
70%	82	80	83	76	93	4.2
60%	70	68	72	64	81	4.5
50%	58	52	65	52	68	11
40%	43	42	45	40	56	7.0
30%	31	20	43	28	44	34
20%	15	13	16	16	32	18
10%	1.1	0.2	2.0	3.5	21	77
0%	0.8	−1.5	2.6	−8.8	8.8	314

CI = confidence interval; PI = prediction interval.

Table 3.

Spike recovery test (SRT), with upper and lower 95% confidence and prediction intervals, for triplicate measurement of pooled frozen feline paraoxonase 1 (PON1) activity, performed by adding increasing percentages (10–100%) of the high concentration pool to a low PON1 concentration pool.

SRT	PON1 activity (U/L)					Total error (U/L)
	Mean	CI		PI
		Lower 95%	Upper 95%	Lower 95%	Upper 95%
0%	21	20	21	19	24	1.5
10%	32	31	33	29	34	0.2
20%	41	41	42	39	44	1.1
30%	52	50	53	49	54	2.1
40%	61	60	63	59	65	−0.3
50%	72	72	73	70	75	−0.4
60%	84	82	87	80	85	−1.0
70%	94	92	96	90	95	0.3
80%	103	101	105	100	105	3.1
90%	112	107	117	110	116	0.5
100%	122	121	122	120	126	1.5

CI = confidence interval; PI = prediction interval.

PON1 storage stability

Although median values tended to increase, no significant differences were recorded in specimens kept at RT for 12 and 24 h (Table 4). Moreover, changes were < 15% compared with T0. Serum PON1 activity increased in refrigerated specimens (p < 0.001), specifically with an increase of 18% after storage of 72 h. Short-term freezing evidenced a significant (p < 0.01) increase (19%) after 72 h, followed by a return to baseline values at 1 wk. A significant difference was present also during long-term freezing (p < 0.01), with a significant increase (25%) after 6 mo.

Table 4.

Assessment of feline paraoxonase 1 (PON1) activity (U/L) stability under different storage conditions.

Temperature/Time	n	Median	I–III quartiles	Min.–max.
RT
T0	4	77	62–101	57–112
12 h	4	78	59–101	55–109
24 h	4	82	68–110	64–124
4°C
T0	4	77	62–101	57–112
12 h	4	77	60–98	57–106
24 h	4	82	66–109	61–123
48 h	4	83	68–112	63–127
72 h	4	96*	82–128	79–144
−20°C (short-term)
T0	5	85	64–107	57–112
24 h	5	86	68–110	60–119
72 h	5	101*	79–113	73–128
1 wk	5	84	64–111	56–113
−20°C (long-term)
T0	8	70	61–85	55–112
1 wk	8	74	59–83	54–113
1 mo	8	77	59–112	53–120
6 mo	8	87*	72–109	69–131

Long-term = storage at freezing temperature up to 6 mo; RT = room temperature; short-term = storage at −20°C up to 1 wk; TO = initial time (time zero).

^*PON1 activity median values > 15% compared with initial value (T0).

Effects of interfering substances

The addition of hemoglobin, bilirubin, and lipids induced significant differences compared with the baseline values (Table 5). However, increasing concentration of hemoglobin and bilirubin induced a fluctuation around the baseline values and despite significant differences, deviations never exceeded ±15% (Fig. 2). Conversely, lipids induced a progressive increase of PON1 activity, with significant differences at the maximum concentration, when the deviation was > 15% (Fig. 2).

Table 5.

Results of interference studies performed by spiking a pool of 22 frozen (−20°C) feline sera with increasing amounts of hemoglobin, bilirubin, or triglycerides.

Hemoglobin	0.00 g/L	6.9 g/L	13.8 g/L	27.5 g/L	55 g/L
PON1 (U/L)	53 ± 0.7	52 ± 0.5* (−2.3%)	51 ± 0.2* (−3.9%)	48 ± 0.9* (−10.2%)	57 ± 0.7* (+8.8%)
Bilirubin	0.00 mmol/L	2.7 mmol/L	5.3 mmol/L	10.8 mmol/L	21.4 mmol/L
PON1 (U/L)	52 ± 0.5	52 ± 0.9 (+1.2%)	47 ± 0.4 (−8.8%)	59 ± 0.4* (+14.4%)	52 ± 0.5 (0.0%)
Triglycerides	0.00 mmol/L	1.41 mmol/L	2.82 mmol/L	5.65 mmol/L	11.30 mmol/L
PON1 (U/L)	53 ± 0.7	53 ± 1.6 (+1.1%)	54 ± 1.3 (+2.4%)	57 ± 0.3 (+7.9%)	64 ± 1.2* † (+19.3%)

PON1 (paraoxonase 1) values are expressed as mean ± SD, with percent deviation reported in parentheses.

^*Significant difference (p ≤ 0.05) compared with the baseline value.

^†Deviation > 15% from the baseline value.

Figure 2.

Interferograms of feline paraoxonase 1 (PON1) activity in pooled serum with medium concentration (from 22 frozen samples) spiked with bilirubin, lipids, and hemoglobin. Data are expressed on y-axis as percentage changes of PON1 activity (mean of triplicate readings) compared with the baseline value. The x-axis refers to the increasing concentrations of interfering substances. The gray area shows the ±15% acceptance criterion.

Reference intervals

Information about sex were available for 69 of 71 cats: 43 were male and 26 were female. Twenty-five cats were < 1 y old, 13 cats were 1–2 y old, and 33 cats were > 2 y old and up to 13 y old. The RI determined in 70 clinically healthy cats (1 outlier was removed before analysis) was 58–154 U/L (Table 6, Suppl. Fig. 1).¹⁷ Significant age-related differences in PON1 activity were not found by linear regression analysis (p = 0.822). PON1 activity was significantly higher in male than in female cats (Mann–Whitney U test, p = 0.025). However, the RIs calculated separately for male and female cats largely overlapped (Table 6) and, more importantly, the lower reference limit of male cats was similar to that of females.

Table 6.

Reference interval for paraoxonase 1 (PON1) activity in a healthy cat population (first row), with sex-based intervals (sex unknown in 2 cats).

Analyte (U/L)	n	Mean ± SD	Median	Min.–max.	Outlier	LRL 90% CI	URL 90% CI	D	M
PON1 activity	70	91 ± 23	87	47–174	2 (S); 1 (R)	58 (47–65)	154 (132–174)	NG	NP
Female	26	85 ± 17	81	62–148	1 (S)	63 (59–68)	138 (110–211)	NG	Rb, T
Male	42	96 ± 24	91	61–174	1 (S)	61 (61–66)	172 (137–174)	NG	NP

CI = confidence interval; D = data distribution; LRL = lower reference limit; M = method; NG = non-gaussian; NP = nonparametric; R = removed outlier; Rb = robust; S = suspected outlier; SD = standard deviation; T = transformed; URL = upper reference limit.

Correlation among PON1, AGP, and SAA

PON1 activity showed a moderate (r = −0.533) significant negative correlation (p < 0.001) with AGP, but a very weak (r = −0.189) and not significant correlation with SAA (p = 0.093). SAA and AGP were strongly correlated to each other (r = 0.666; p < 0.001; Fig. 3). PON1 activity was significantly higher (p < 0.001) in cats with normal AGP (mean ± SD = 83 ± 33 U/L; median: 84 U/L; min.–max.: 1.2–178 U/L; n =100) than in cats with high AGP (53 ± 31 U/L; 50 U/L; 0.2–160 U/L; n = 239), as well as in cats with normal AGP compared with cats with medium AGP (64 ± 31 U/L; 60 U/L; 0.2–160 U/L; n = 120) and very high AGP (42 ± 28 U/L; 39 U/L; 4.3–152 U/L; n = 119; p < 0.001 for all the comparisons; Fig. 4). However, PON1 activity in cats with normal and medium AGP markedly overlapped; the overlap between groups with normal and high AGP was minimal. A portion of cats with normal AGP concentration (20 of 100, 20%) as well as a portion of cats with high AGP (144 of 239, 60.3%) had PON1 activity below the lower reference limit, but there were no significant differences (p = 0.657). Conversely, a significant difference (p < 0.001) was found between the proportions recorded among cats with normal AGP, cats with medium AGP concentration (53 of 120; 44.2%), and cats with very high AGP concentration (91 of 119; 76.5%).

Figure 3.

Scatter diagrams of feline paraoxonase 1 (PON1) activity (n = 339) and concentrations of alpha-1-acid glycoprotein (AGP; n = 339) and serum amyloid A (SAA; n = 82) measured in frozen sera.

Figure 4.

Box and whisker dot plots of paraoxonase 1 (PON1) activity in 339 feline frozen sera versus different concentration of alpha-1-acid glycoprotein (AGP). The ends of the box are the upper and the lower quartiles; the box spans the interquartile range, and the central line is the median. The whiskers are the 2 lines outside the box that extend to the highest and lowest observation. ***p < 0.001 significant difference compared to the group with normal AGP concentration. †††p < 0.001 significant difference between group with medium and high AGP concentration. The gray area is the reference interval for PON1 activity.

No significant differences (p = 0.203) in PON1 activity were found between cats with normal SAA (PON1 activity: 71 ± 40 U/L; 74 U/L; 1.2–160 U/L; n = 45) and high SAA (60 ± 35; 56 U/L; 3.8–149 U/L; n = 37) or between cats with normal-to-medium (64 ± 39 U/L; 61 U/L; 3.8–149 U/L; n = 20) and very high SAA (55 ± 29 U/L; 50 U/L; 7.9–126 U/L; n = 17; p = 0.332; Fig. 5). No significant differences were found between the proportions of cats with PON1 activity lower than the lower reference limit recorded in cats with normal SAA (17 of 45, 38%) and in cats with high SAA (19 of 37, 51%; p = 0.218), or between the proportions recorded among cats with normal SAA, cats with medium SAA (9 of 20, 45%) and cats with very high AGP (10 of 17, 59%; p = 0.328).

Figure 5.

Box and whisker dot plots of paraoxonase 1 (PON1) activity in 82 frozen feline sera with different concentration of serum amyloid A (SAA). The ends of the box are the upper and the lower quartiles; the box spans the interquartile range. The line across the box is the median. The whiskers are the 2 lines outside the box that extend to the highest and lowest observation. The gray area is the reference interval for PON1 activity.

Discussion

Given the lack of a “natural” substrate for PON1, non-physiologic substrates are used to measure PON1 activity: paraoxon (paraoxonase activity), non-phosphorus arylesters (arylesterase activity), or lactones (lactonase activity).^5,6 The rationale of our selection of paraoxon as substrate was based on technical reasons, as well as on previous publications. Phenylacetate (PA), p-nitrophenyl acetate (pNA), and 5-thiobutil butyrolactone (TBBL) have been validated previously in cats. However, although paraoxon is more toxic than other substrates, we believe it offers some advantages: the use of PA or TBBL is time consuming, given that it can only be performed using manual methods or on 96-well microplates and ultraviolet wavelength. On the other hand, even though the measurement with pNA can be automated, results in healthy cats are very low (median: 3.4 U/mL),⁵⁶ and it may be difficult to appreciate decreased values expected for a negative APP such as PON1. To our knowledge, paraoxon-based methods have not been validated previously in cats, and the results showed higher values than other substrates,^56,59 so it may be more useful in practice. A limitation of paraoxon as a substrate is its toxicity.⁶ Paraoxon has acute oral, dermal, and inhalation toxicity (MilliporeSigma safety information); personal protective equipment (PPE; such as nitrile gloves, eye shields, and face masks) must be worn, and it must be handled in a fume hood. In some countries (e.g., United States) additional PPE could be required (e.g., full-face respirator). Moreover, paraoxon is very toxic to aquatic life, and appropriate disposal of contaminated consumables is mandatory.

We demonstrated that our method is precise and accurate given that both intra-assay and inter-assay imprecision were < 5% (an acceptable imprecision for most biochemical analytes).⁴⁴ The LUD and SRT confirmed the excellent accuracy of this method, already recorded in other species.^21,47,51 A limitation is represented by the use of distilled water instead of saline as diluent for the LUD test, which can modify the pH and electrolyte concentrations, which in turn can affect PON1 activity. Conversely, the use of distilled water provided some evidence of a matrix effect, as demonstrated by a higher total error when the serum-to-water ratio was ≥ 50%. The matrix effect was also supported by the wider CIs and PIs in the LUD test compared with the SRT.

RIs for PON1 activity in our study were lower than those reported for dogs⁴⁶ and humans,⁵⁴ and similar to those of cattle²¹ or horses.⁵¹ The peculiar lipoprotein metabolism of cattle may explain their low PON1 activity; however, a similar finding in cats is surprising given that cats are metabolically more similar to dogs than to cattle. The lack of significant age-related differences seems to contrast with reports in cattle²¹ and people.¹⁸ Nevertheless, these latter studies were focused on PON1 activity in newborns, in which the biosynthetic ability of the liver is immature. In our study, the youngest animals were already weaned and could be considered “young adults,” with PON1 activity not significantly different from adults, as also found in people¹⁸ or in horses.⁵¹ It is therefore possible that newborn cats (not included in our study) may have significantly lower PON1 activity than adults. Conversely, the higher PON1 activity recorded in male cats contrasts with reports in dogs, in which no significant sex-related differences were found,⁴⁷ or in mice, cattle, or horses, in which PON1 activity is higher in females than in males.^3,21,51 The sex-related difference recorded in our study may suggest that different RIs should be used in male or female cats. However, the 2 RIs largely overlap and, more importantly, the lower reference limit, which in practice may correspond to the clinical decision limit, was very similar with a difference quantitatively lower than the intrinsic imprecision of the method. Therefore, although the decision to use separate RIs should depend on statistical approaches,^25,33 the current guidelines for the establishment of de novo RIs suggest the use of non-statistical approaches based on practical considerations.¹⁷ From this perspective, it would not be advisable to differentiate the RIs of male versus female cats because it would likely not affect clinical decisions.

As in other species, the paraoxon-based method suffers from the presence of interfering substances. Hemoglobin and bilirubin induced clinically relevant differences in other species; in cats, they induced non-clinically relevant fluctuations around the baseline values rather than homogeneous trends to increased or decreased values. Conversely, the interference with PON1 activity caused by the addition of lipids, previously documented in other species,^47,51 may be clinically relevant. PON1 activity must not be measured in lipemic specimens, whereas it can be measured in hemolytic and icteric specimens, although slight deviations from the true value may occur. This is important because inflammatory conditions of cats (FIP, cholangitis, triaditis, inflammation associated with hemoplasma infection) may induce icterus or hemolysis.^4,13

Refrigeration and long-term freezing may increase serum PON1 activity. Our study did not allow us to investigate the possible cause of the increase. However, PON1 in human serum is not affected by long-term storage or repeated freeze–thaw cycles, regardless of the substrate used.²⁷ Further investigations of temperature-dependent effects are warranted to explain the differences between our study and previous studies. We advise measuring PON1 just after sampling or after short-term freezing. This may be a limitation of our study, in which PON1 and other APPs s were compared on stored specimens. Another limitation of our study is the 2 freeze–thaw cycles; however, a study performed on dogs demonstrated that PON1 activity significantly decreased only after the third cycle.⁵⁰ So, it is unlikely that 2 freeze–thaw cycles were enough to induce a significant decrease in PON1 activity. However, the magnitude of the differences in PON1 activity of specimens with normal or very high AGP was higher than the deviation related to long-term storage. Speculatively, the lack of correlation with SAA, as well as the lack of significant differences between groups with low and high SAA, is the result of factors other than the storage-induced increase of PON1. For example, the type of substrate may influence the magnitude of changes and the subsequent lack of correlation. In a prior study, a strong negative correlation was demonstrated using pNA and phenyl acetate as substrates (which both assess arylesterase activity), whereas only a weak negative correlation was present using TBBL (which assesses lactonase activity) as substrate.⁵⁶ More studies are necessary to better explain the relationship between PON1 determined using paraoxon as substrate and SAA, using fresh specimens and clinically well-defined healthy and diseased groups. Another limitation of our study is the lack of information about PON1 genetic polymorphisms in cats. In people, polymorphism induces high inter-individual variability.^7,8,35 Polymorphism can also explain the lower PON1 activity observed in cats compared to dogs.

APP increases are nonspecific, with the exception of AGP for the diagnosis of FIP; although very high values can be found in cats with other viral diseases, an increase over 0.0375 mmol/L may be diagnostic for FIP.^12,26 Therefore, we assessed the ability of PON1 activity to identify cats with very high AGP, demonstrating minimal overlap with results of cats with normal AGP. This finding encourages further studies on cats with FIP to assess whether PON1 can be included in the diagnostic approach to this disease. As a negative APP, a decrease of PON1 activity is expected when AGP or SAA concentrations increase, whereas a positive correlation is expected between AGP and SAA. However, the negative correlation was moderate with AGP and absent with SAA. A partial overlap of results in cats with normal, medium, or high AGP values was found, and PON1 activity was not significantly different in cats with different concentrations of SAA. A high proportion of cats with PON1 activity lower than the lower reference limit was found also in cats with normal AGP or SAA. It can be hypothesized that the group of cats with normal AGP concentration included cats with diseases able to induce a decrease of PON1 activity without a concurrent increase in AGP. Conversely, SAA increases dramatically and very early during several diseases.⁵² PON1 liver synthesis is decreased by inflammatory stimuli, but it is also consumed by an oxidative burst,^15,40 whereas AGP or SAA are not affected by oxidative stress. Therefore, the different responses of PON1, AGP, or SAA can depend on different underlying diseases in the cats included in our study, and this may justify the lack of correlation with SAA. This result is similar to a report in dogs, in which PON1 activity and C-reactive protein (CRP) were not always correlated.⁴⁷ Moreover, ours was a methodologic study focused on PON1 measurement and on the establishment of RIs, whereby comparison with positive APPs was performed for preliminary information about the possible role of PON1 as a negative APP in cats. The lack of clinical information may lead to possible inaccuracy in classifying the health status of cats based on AGP or SAA. APPs may also increase in non-inflammatory conditions,^42,52 or they may be normal in cats with diseases not associated with inflammation but that could affect PON1 activity. Despite these considerations, the moderate negative correlation with AGP and the trend to decreased PON1 activity in cats with increasing SAA encourages the design of further studies based on a rigorous clinical classification of sick cats according to the presence or absence of inflammatory diseases.