Zahir et al. provided the first evidence for a possible clinical benefit of measuring PBMC tacrolimus concentrations in a study of 40 adult liver transplant recipients. They reported that a lower proportion of total blood tacrolimus was associated with the leucocyte fraction in patients with rejection compared to those without (Zahir et al., 2004a; Zahir et al., 2004b). In 90 adult liver transplant recipients studied 7 days post transplantation, Capron et al. (2012) later observed that whilst there was no relationship between C_0Blood and rejection, trough PBMC tacrolimus concentrations (C_0PBMC) were significantly lower in patients with rejection compared to those without, regardless of whether rejection was classified histologically or clinically (Table 1A). Importantly, C_0PBMC measured on days 3 and 5 were also significantly different between recipients who would be classified as rejectors or non-rejectors on day-7, suggesting the potential to predict early rejection, In addition, day-7 C_0PBMC correlated with the histological grading of rejection. To date, this is the only study that has directly and prospectively compared prediction of rejection by C_0Blood and matched C_0PBMC taken on the same day as the protocol liver biopsy used to classify rejection. However, patients received only tacrolimus monotherapy for maintenance immunosuppression, with or without anti-lymphocytic induction therapy, and there was a high incidence (41%) of moderate/severe histological rejection.

Two other clinical studies that measured PBMC tacrolimus concentrations in liver transplants have also included clinical outcome data. These more recent studies recruited patients receiving triple maintenance immunosuppression with tacrolimus, mycophenolic acid and a corticosteroid (Lemaitre et al., 2015; Rayar et al., 2018). In 10 transplant recipients recruited at steady state (Table 1B), only one case of acute rejection was observed precluding any investigation of the relationship between rejection and tacrolimus concentrations in PBMC (Lemaitre et al., 2015). A study of 41 patients (Table 1A) recruited in the first week post transplantation did not collect data on rejection but found no association between C_0PBMC and measures of graft function (Rayar et al., 2018).

Only three studies in kidney transplant recipients provide any data on clinical outcomes and tacrolimus concentrations in blood and PBMC (Francke et al., 2020; Han et al., 2016; Klaasen et al., 2018). Again, transplant recipients received triple maintenance immunosuppression as above. In a small study of 29 patients (Supplementary Table S1), the incidence of rejection 1 week post-transplant was too low to adequately investigate any relationship with PBMC tacrolimus concentrations (Klaasen et al., 2018). Two larger studies (Francke et al., 2020; Han et al., 2016) carried out at steady state assessed clinical outcomes retrospectively (Table 1A). Han et al. reported 15.6% of kidney transplant recipients experienced rejection within the first 6 months post-transplant (Han et al., 2016). However, the authors did not state how rejection was defined and found no association between rejection and C_0PBMC measured in 214 patients up to 14 years post-transplant. Using only for-cause biopsies in 175 recipients, Francke et al. reported 8% developed biopsy proven rejection in the first 3 months of renal transplantation, with no difference in the 3-month C_0PBMC, C_0Blood or their ratio between patients who had and hadn’t experienced rejection (Francke et al., 2020). Both studies (Han et al., 2016; Francke et al., 2020) also investigated tacrolimus-induced nephrotoxicity and found no difference between C_0PBMC in patients who had or hadn’t experienced nephrotoxicity over the first 6 or 3 months of renal transplantation, respectively.

Overall, there is very little evidence for clinical utility of C_0PBMC in the prediction of rejection or tacrolimus-induced nephrotoxicity. This may mostly be due to small sample sizes, the low incidence of rejection with modern triple maintenance immunosuppression, and the often retrospective assessment of rejection and nephrotoxicity. The only clinical evidence for prediction of rejection is from a single well-designed study in liver transplant patients receiving tacrolimus monotherapy, which may not be directly translatable to patients on modern maintenance immunosuppression or to other transplant types.

In the absence of sufficient clinical outcome data, several groups have investigated the relationship between blood or PBMC tacrolimus concentrations and ex vivo calcineurin activity (CNA) or T-cell activation (Table 1B). In 10 de novo liver transplant recipients studied on days 1 and 7 following commencement of tacrolimus therapy Lemaitre et al. (2015) reported that, over a 12-h dosing interval, inhibition of PBMC CNA on day 1 mirrored tacrolimus concentrations in both blood and PBMC, with an average maximum inhibition of 38% occurring slightly after attainment of maximum tacrolimus concentrations in blood (C_maxBlood) and PBMC (C_maxPBMC). On day 7, no change in CNA was observed over the dosing interval, and less peak to trough variability in the tacrolimus concentration vs time curves was apparent in both blood and PBMC compared to day 1. There was no correlation between the 12-h area under the CNA vs time curve (AUC_CNA) on days 1 or 7 and the corresponding 0–12 h areas under the tacrolimus concentration vs time curves in whole blood (AUC_Blood) or PBMC (AUC_PBMC). In this study, although PBMC tacrolimus concentrations and CNA were measured in the whole PBMC fraction, the authors accounted for granulocyte contamination of PBMC by expressing the tacrolimus concentrations per 10⁶ leukocytes, rather than the conventional use of total cells. However, tacrolimus would likely also distribute into granulocytes, potentially biasing the relationship between measured intracellular tacrolimus concentrations and CNA.

In 32 de novo liver transplant recipients studied between day 7–10 post-transplant, Tron et al. (2020) confirmed that maximal inhibition of CNA (CNA_Imax) occurred 2 h post-dose, slightly after attainment of C_maxBlood and C_maxPBMC (1.6 h post-dose). CNA_Imax was correlated with both log-C_maxPBMC and log-C_maxBlood with a median 37% maximal inhibition compared to baseline CNA measured before administration of the first post-transplant tacrolimus dose. The authors calculated C_maxBlood and C_maxPBMC producing 50% inhibition of CNA (IC₅₀) of 18 μg/L and 100 pg/10⁶ cells, respectively. Similar to an in vitro IC₅₀ of 160 pg/10⁶ cells calculated using PBMC isolated from healthy volunteers (Tron et al., 2019). Using a population pharmacokinetic model developed in the same study (discussed below in Whole blood and PBMC tacrolimus pharmacokinetics), Tron et al. (2020) estimated that in recipients with C_0Blood of <4, 4-6 or 6–10 μg/L only 13, 39, and 42%, respectively, were likely to attain C_maxPBMC greater than the IC₅₀. In comparison, Capron et al. reported mean (s.d.) C_0PBMC of 90.9 (41.2) vs 33.8 (16.7) pg/10⁶ cells in liver transplant patients with no/mild vs moderate/severe biopsy graded rejection; and 48.7 (11.9) vs 22.0 (6.1) pg/10⁶ cells in patients without vs with clinically significant rejection (Capron et al., 2012). This may suggest that in vivo prevention of rejection potentially occurs at tacrolimus PBMC concentrations lower than those required for ex vivo inhibition of PBMC CNA. Interestingly, Tron et al. also reported significant inter-individual variability in baseline CNA (coefficient of variation (CV) = 66%), indicating a considerable component of inter-individual pharmacodynamic variability. Most recently, Fontova et al. (2021) also reported an average 29% maximum inhibition of CNA within a 12 h dosing interval and a significant inverse correlation between AUC_Blood and the area under the percentage inhibition of CNA versus time curve in renal transplant recipients. Although they also measured AUC_PBMC a similar analysis was not performed.

Han et al. (2016) investigated C_0Blood and C_0PBMC in 213 stable renal transplant recipients, and quantitated interferon-γ (IFN-γ) and interleukin-2 (IL-2) expressing T-cells in a subset of 39 recipients grouped according to C_0PBMC. They reported that, following ex vivo activation with phorbol-12-myristate 13-acetate and ionomycin, the proportion of CD3⁺CD4⁺IFN-γ⁺, CD3⁺CD4⁺IL-2⁺ and CD3⁺CD8⁺IL-2⁺ T-cells was significantly greater in the low C_0PBMC group. However, a significantly higher proportion of CD3⁺CD4⁺IFN-γ⁺ and CD3⁺CD4⁺IL-2⁺ cells was also observed when the comparisons were based on low versus high C_0Blood.

In general, these studies show that both blood and PBMC tacrolimus concentrations correlate with inhibition of CNA or measures of lymphocyte activation, although the correlations appear stronger using PBMC concentrations. The reports are also consistent with other studies indicating that tacrolimus does not completely inhibit lymphocyte CNA (Fukudo et al., 2005) and support the role of triple maintenance immunosuppression in allowing lower tacrolimus exposures whilst still maintaining a relatively low risk of rejection.

Tacrolimus PBMC concentrations have been measured in recipients of liver, kidney and heart transplants, with mean C_0PBMC ranging from 22.5–266 pg/10⁶ cells and corresponding mean C_0Blood between 3.4–10.5 μg/L (Capron et al., 2010; Capron et al., 2012; Lemaitre et al., 2013; Lemaitre et al., 2015; Han et al., 2016; Klaasen et al., 2018; Romano et al., 2018; Francke et al., 2020; Tron et al., 2020; Fontova et al., 2021) (Supplementary Table S1). Most studies report greater inter-individual variability in PBMC compared to blood tacrolimus concentrations, with CVs for C_0PBMC and C_0Blood ranging from 40 to 110% and 20–57%, respectively (Capron et al., 2010; Capron et al., 2012; Lemaitre et al., 2013; Lemaitre et al., 2015; Han et al., 2016; Klaasen et al., 2018; Romano et al., 2018; Tron et al., 2020). Whether this reflects greater physiological variability or analytical variability is unclear. Only one study has assessed variability of the PBMC preparation step and reported a CV of 7.3%, which was greater than the intra-assay imprecision but similar to inter-assay imprecision (Klaasen et al., 2018). This provides some confidence that inter-individual variability in tacrolimus PBMC pharmacokinetics may indeed be greater than in blood, supporting a potential benefit of measuring concentrations in PBMC.

In keeping with a greater variability of tacrolimus PBMC pharmacokinetics, a lack of statistically significant correlation between C_0Blood and C_0PBMC has been reported in some studies (Capron et al., 2010; Capron et al., 2012; Lemaitre et al., 2013; Romano et al., 2018). However, others have found significant, albeit weak, correlations between blood and PBMC tacrolimus concentrations at C₀ (Pensi et al., 2015; Han et al., 2016; Klaasen et al., 2018; Francke et al., 2020; Tron et al., 2020), 1.5 h post dose (Klaasen et al., 2018) and C_max (Tron et al., 2020), and between AUC_Blood and AUC_PBMC (Lemaitre et al., 2015; Tron et al., 2020; Fontova et al., 2021). Tron et al. (2020) also reported that C_0Blood correlated with both C_maxPBMC and AUC_PBMC. Fontova et al. recently demonstrated circadian variability in both blood and PBMC tacrolimus pharmacokinetics, with higher blood and PBMC tacrolimus exposures following the morning (compared to evening) dose; and stronger correlations between the 0–12 h tacrolimus AUC in blood or PBMC and the corresponding 12 h (C₁₂₎, rather than the pre-dose (C₀) trough concentrations (Fontova et al., 2021). Previous studies have also reported better correlations between C_12Blood and AUC_Blood (Barraclough et al., 2011; Marquet et al., 2018).

Since T-cells can comprise between 12–92% of total cell numbers in PBMC preparations, one study has compared tacrolimus C₀ in whole blood, PBMC, purified CD4⁺ T-cells and purified CD19⁺ B-cells of kidney transplant recipients (Romano et al., 2018). In this study, tacrolimus C₀ were higher in T- and B-cells compared to PBMC, and there was a significant correlation between C₀ in blood and T-cells, but not between C₀ values in any of the other matrices. A study in healthy volunteers administered tacrolimus also found a significant correlation between tacrolimus concentrations in blood and T-cells (but not PBMC) (In 't Veld et al., 2019), and in contrast to the study in transplant recipients, tacrolimus concentrations in T-cells were lower than those in PBMC. Taken together, all studies indicate that across liver, renal and heart transplant groups, the relationship between blood and PBMC tacrolimus concentrations is relatively modest and may be affected by the composition of the PBMC fraction. This may, in part, contribute to the poor correlation between blood and PBMC tacrolimus concentrations, as rejection and inflammation are likely to alter the cellular composition of this matrix.

As previously discussed, only unbound tacrolimus distributes from plasma into erythrocytes and the other cells contained in whole blood. Thus, changes in haematocrit, red cell number and plasma proteins may impact the proportion of tacrolimus in whole blood that is distributed within the PBMC fraction. In addition, leucocyte uptake and efflux of tacrolimus may involve carrier mediated processes which may be subject to saturability, induction, inhibition and genetic polymorphisms. All these processes may also affect the rate and extent of tacrolimus distribution within the different compartments in whole blood. The C_0PBMC/C_0Blood ratio is an indication of the proportion of whole blood tacrolimus that is distributed within PBMC. Identifying covariates that determine this ratio may assist in the prediction of tacrolimus C_0PBMC from C_0Blood.

In kidney transplant recipients, Capron et al. (2010) used multiple linear regression to assess pharmacogenetic and other clinical variables (Table 2A) as predictors of C_0PBMC, C_0PBMC/dose and C_0PBMC/C_0Blood. They reported that recipient ABCB1 SNPs (1199GA, 3435TT), CYP3A5 non-expressor genotype, a CYP3A5*3—ABCB1 1199GA interaction, and the log of mean corpuscular volume (MCV) were independent determinants of C_0PBMC/Dose one week after renal transplantation, whilst at steady-state ABCB1 1199GA was no longer significant. The effect of CYP3A5 most likely reflected the TDM guided lower doses of tacrolimus in non-expressors. In contrast, ABCB1 SNPs (1199GA, 3435CT, 3435TT) and total plasma protein concentrations were independent determinants of C_0PBMC/C_0Blood one week post-transplantation, whilst at steady-state ABCB1 3435TT was no longer significant (Table 2A). In this study the variants of ABCB1 (the gene coding for P-glycoprotien) were independent predictors of a higher C_0PBMC/C_0Blood ratio (Table 2A), consistent with reduced efflux of tacrolimus from PBMC. In addition, high total plasma protein was an independent predictor of a lower ratio (Table 2A), consistent with increased binding of tacrolimus to plasma proteins and therefore less unbound tacrolimus available for distribution into PBMC. Unfortunately, haematocrit does not appear to have been tested as a covariate in this analysis.

In contrast, in stable renal transplants, Han et al. (2016) found no association between recipient ABCB1 SNPs and C_0PBMC/C_0Blood, but did find a significant association with sex, haematocrit and transplant duration using analysis of covariance (Table 2A). In addition, haematocrit and transplant duration were also significantly associated with C_0PBMC. Unfortunately, tacrolimus dose was not investigated. More recently, using multiple linear regression in kidney transplants, Francke et al. (2020) also found no effect of recipient ABCB1 1199/3435 variant alleles, CYP3A5*3 or CYP3A4*22 on C_0PBMC/C_0Blood ratio but age, albumin and haematocrit were independent predictors of the ratio at 3 months post-transplant (Table 2A), similar to (Han et al., 2016).

Tron et al. (2020) developed a 2-compartment population pharmacokinetic model describing the relationship between blood and PBMC tacrolimus concentrations approximately 1 week after liver transplantation. Although they did not find any demographic or pharmacogenetic covariates that significantly improved their model, univariate analyses of model-derived AUCs and observed C₀ and C_max revealed a significant but weak inverse correlation between haematocrit and AUC_PBMC and between haematocrit and the ratio of AUC_PBMC/AUC_Blood (Table 2B), similar to the studies in renal transplants. High haematocrit is consistent with a larger erythrocyte binding compartment and therefore less unbound tacrolimus available for distribution into PBMC. In the above renal and hepatic transplant studies, haematocrit was the only consistent determinant of the proportion of whole blood tacrolimus that is present within PBMC. Haematocrit is also a significant covariate in many solid organ transplant population models of whole blood tacrolimus pharmacokinetics (Brooks et al., 2016).

Tron et al. also found that recipient (but not donor) 2677TT and 1236/2677/3435 homozygous triple variant ABCB1 SNPs were associated with a lower AUC_PBMC/AUC_Blood ratio, whilst the recipient ABCB1 1199A variant allele had no effect (Table 2B) (Tron et al., 2020). In this study the lower AUC_PBMC/AUC_Blood ratio in carriers of the ABCB1 3435T variant contradicts its association with higher C_0PBMC/C_0Blood ratios reported by Capron et al. (2010) 1 week after renal transplantation. In addition, the association between recipient ABCB1 1199A variant and C_0PBMC/C_0Blood ratios reported by Capron et al. (2010) 1 week after renal transplantation was not observed for the AUC ratio reported by Tron et al. (2020). In kidney transplants, recipient genotypes would relate to hepatic, intestinal and PBMC enzyme/transporter activities (Capron et al., 2010), whilst in liver transplants (Tron et al., 2020) recipient genotypes would relate to intestinal (not hepatic) and PBMC activities, potentially explaining some of the pharmacogenetic discordance between studies. Additionally, relatively small sample sizes and very small numbers of patients who were carriers of variant alleles may also have contributed to discordant observations.

Although ABCB1 genetic polymorphisms are not major determinants of tacrolimus blood clearance, the ABCB1 1199G>A SNP has been shown to increase in vitro intra-cellular accumulation of tacrolimus in HEK293 and K562 recombinant cell lines (Dessilly et al., 2014). However, its relative role in the net efflux of tacrolimus from PBMC is unknown and currently its effects on the PBMC:blood concentration ratio is contradictory. Unfortunately, direct comparison of these studies is difficult due to differences in the preparation of PBMC; cellular composition of PBMC, sample size; ethnicity of transplant recipients; exclusion of drugs that interact with CYP3A and P-glycoprotein; transplant duration; covariates investigated; statistical analyses and corrections for multiple comparisons.

Most of the above studies have addressed inter-individual variability in the C_0PBMC/C_0Blood ratio but not intra-individual variability. Han et al. (2016) measured C_0PBMC/C_0Blood on two occasions in a small subset of renal transplant recipients and reported that the ratio measured >1 year post-transplant was significantly lower than that measured in the first year. In contrast, two later studies with larger sample sizes found no effect of time post-transplant on C_0PBMC/C_0Blood when subjects were repeatedly sampled at different time-points after transplantation (Klaasen et al., 2018; Francke et al., 2020). They reported median or mean intra-individual CVs of 45% (range 5.9–88%) (Klaasen et al., 2018) and 39.0% (range 3.5–173.2%) (Francke et al., 2020), which were lower than the CVs for inter-individual variability (Francke et al., 2020). Interestingly, in the patients with the greatest intra-individual variability, the variability could not be explained by changes in haematocrit (Francke et al., 2020).

One important determinant of tacrolimus distribution (and activity), which has not been addressed, is intra-cellular binding capacity (Figure 1). FKBP12 is the major erythrocyte cytoplasmic protein to which tacrolimus binds (Nagase et al., 1994). Whilst haematocrit is an estimate erythrocyte numbers, it does not address variability in erythrocyte expression of FKBP12. Although inhibition of calcineurin in lymphocytes is mediated by the tacrolimus-FKBP12 complex, tacrolimus also binds to other FKBPs whose expression differs between tissues and cell types (Baughman et al., 1997). Thus, differences or changes in the expression of FKBPs are likely to affect both inter- and intra-individual variability in whole blood and PBMC tacrolimus pharmacokinetics. In addition, variability in FKBP expression within lymphocytes may also affect the degree of calcineurin inhibition by tacrolimus (Kung and Halloran, 2000).

In 1992 Sandborn et al. reported that liver transplant recipients with cellular rejection had lower graft tissue ciclosporin concentrations than those without (Sandborn et al., 1992). They later expanded these observations to tacrolimus in a group of 17 de novo liver transplant recipients administered maintenance immunosuppression consisting of tacrolimus and prednisolone (Sandborn et al., 1995). Seven of the 17 patients developed nine episodes of histological rejection and had graft tacrolimus concentrations (measured using clinical biopsy samples) that were significantly lower than those in the patients without rejection. However, they found no difference in plasma tacrolimus concentrations between patients with and without rejection. These early reports (Table 1A) provided much of the impetus for better understanding the role of graft tacrolimus concentrations in determining risk of rejection. In these early studies both plasma and tissue tacrolimus concentrations were measured by an immunoassay with significant cross-reactivity to tacrolimus metabolites (Wallemacq et al., 2009). Since the metabolite/parent tacrolimus concentration ratio changes over a dosing interval and may also differ between plasma and other tissues, these early observations may have been subject to significant analytical bias. Capron et al. next investigated the relationship between graft tacrolimus concentrations and rejection in 146 de novo liver transplant recipients administered tacrolimus and corticosteroid maintenance immunosuppression (Table 1A) (Capron et al., 2007). They reported that day 7 graft tacrolimus concentrations (measured by a specific LC-MS/MS method) were significantly lower in patients with moderate/severe histological rejection compared to those with no (or mild) rejection, and that there was a strong first-order exponential correlation between Banff histology score and hepatic tacrolimus concentrations (r² = 0.98 p = 0.002). A cut-off hepatic tacrolimus concentration of 30 pg/mg of tissue predicted clinically significant rejection with 89% sensitivity and 98% specificity. In comparison, there was no difference in C_0Blood in patients with or without mild/moderate rejection. However, C_0Blood were measured with an immunoassay also associated with significant metabolite cross-reactivity (Wallemacq et al., 2009). This was followed by another study (also discussed in Rejection and nephrotoxicity) in liver transplant recipients (Table 1A) again reporting significantly lower liver tacrolimus concentrations (and C_0PBMC) in patients with moderate/severe histological rejection compared to those with no/mild rejection, and a significant relationship between liver tacrolimus concentrations (and C_0PBMC) and Banff scores (Capron et al., 2012). Even though C_0Blood were measured by a relatively specific immunoassay there was still no association between C_0Blood and rejection.

Two studies (Table 1A) have recently investigated potential relationships between renal graft tacrolimus concentrations and clinical outcomes (Zhang et al., 2020; Sallustio et al., 2021). In 52 renal transplant recipients there was no difference in renal tacrolimus concentrations between patients with or without histologically classified subclinical acute rejection at 3 months or 1 year post transplantation (Zhang et al., 2020). In a larger study of 132 renal transplant recipients, biopsy-proven rejection was similarly not associated with renal tacrolimus concentrations (Sallustio et al., 2021). In both studies, patients received triple maintenance immunosuppression with tacrolimus, mycophenolic acid and prednisolone and rejection episodes were observed in 21% of patients (Zhang et al., 2020) and 44% of biopsy samples (Sallustio et al., 2021). These later studies contrast with the earlier results in hepatic transplantation. However, the renal transplant recipients were administered current triple maintenance immunosuppression, whereas the hepatic transplant recipients were administered maintenance immunosuppression of tacrolimus monotherapy or tacrolimus and a steroid (Table 1A). Therefore, the utility of hepatic tacrolimus concentrations as predictors of rejection with modern triple therapy is yet to be determined.

Despite a reduction in the target C_0Blood therapeutic range, evidence of chronic tacrolimus induced nephrotoxicity is still present in 34 and 72% of renal allograft biopsies by 5 and 10 years post-transplantation (Nankivell et al., 2016). Although high C_0Blood are associated with increased risk of nephrotoxicity, it is unclear whether renal CNI concentrations may be better predictors. Of the two studies that have measured intra-renal tacrolimus concentrations in clinical allograft biopsies (Table 1A), only one has investigated nephrotoxicity, reporting that the relationship between blood and renal tacrolimus concentrations may be different (steeper) in patients with acute nephrotoxicity compared to those without (Sallustio et al., 2021). However, these results were based on a very small incidence of acute nephrotoxicity and require confirmation.

Measurement of allograft tacrolimus concentrations is ethically limited to the use of biopsies collected for clinical assessment of graft dysfunction or as part of established routine clinical monitoring protocols. Thus, there are no clinical data on graft tacrolimus AUCs. However, similar to the studies of PBMC tacrolimus exposures, Capron et al. found no correlation between C_0Blood and graft tacrolimus concentrations in liver transplant recipients, using a relatively non-specific immunoassay to measure C_0Blood (Capron et al., 2007) and also in a later publication in which C_0Blood were measured with a more specific immunoassay (Capron et al., 2012). In contrast, the latter study reported a good correlation (r² = 0.55, p = 0.001) between C_0PBMC and hepatic tacrolimus concentrations (Capron et al., 2012), possibly indicating that both PBMC and hepatic tacrolimus concentrations are more closely related to unbound plasma tacrolimus concentrations than those in whole blood. In renal transplant recipients, weak correlations have been reported between C_0Blood and graft tacrolimus concentrations with r² values of 0.13 (p = 0.01) (Zhang et al., 2020) and 0.19 (p = 7.4 × 10⁻¹⁰) (Sallustio et al., 2021). In addition, a better correlation between dose and renal tacrolimus concentrations than between dose and C_0Blood (Sallustio et al., 2021) has also been reported, again potentially indicating that renal tacrolimus concentrations better reflect unbound plasma tacrolimus concentrations, hence, dose.

Similar to PBMC, there appears to be greater inter-individual variability in intra-graft tacrolimus concentrations than C_0Blood (Supplementary Table S1), with mean (s.d.) concentrations of 91.3 (52.2) pg/mg of tissue and 8.9 (3.0) μg/L, respectively, and CVs of 57 and 34%, respectively in liver transplant recipients (Capron et al., 2012). In renal transplantation, C_0Blood and graft tacrolimus concentrations ranged from 2.6 to 52.3 μg/L and 33–828 pg/mg of tissue, respectively (Sallustio et al., 2021). Analytical variability in measurement of intra-renal tacrolimus concentrations appears relatively small, with one study reporting intra- and inter-assay CVs between 5.9 and 14.1% for replicate analyses of in vivo renal cortical tissue tacrolimus concentrations (Noll et al., 2013).

Only one group (Elens et al., 2007) has investigated the effect of ABCB1 and CYP3A5 genetic polymorphisms on hepatic tacrolimus concentrations in liver transplantation (post hoc analysis of (Capron et al., 2007) in Table 1A). Using multiple linear regression analysis, donor ABCB1 2677 G/T, T/T and G/A, 1199G/A and day 7 log plasma bilirubin concentrations were independent predictors of day 7 hepatic tacrolimus concentrations, whist the same genotypes (but not bilirubin) were independent predictors of dose-corrected hepatic tacrolimus concentrations (Elens et al., 2007). In renal transplants, donor or recipient ABCB1 and CYP3A5 genetic polymorphisms had no effect on renal tacrolimus concentrations (Zhang et al., 2020; Sallustio et al., 2021), but C_0Blood (p = 1 × 10⁻⁸), dose (p = 0.02) and acute nephrotoxicity (main effect p = 0.01 and first-order interaction with C_0Blood p = 0.002) were independent predictors of renal tacrolimus concentrations (Sallustio et al., 2021). Interestingly, a greater role of P-glycoprotein in determining hepatic versus renal tacrolimus concentrations is supported by animal work showing that knockout of P-glycoprotein expression in mice results in increased tissue/blood tacrolimus concentration ratios in liver but not in kidneys (Yokogawa et al., 1999).