LAV-BPIFB4 associates with reduced frailty in humans and its transfer prevents frailty progression in old mice

Association with frailty and survival in humans

The baseline characteristics of the cohort are illustrated in Table 1. The association analyses with frailty trait showed that the LAV homozygous haplotype is under-represented in frail subsets of the cohort (p = 0.030 vs. other haplotypes), thus suggesting a potentially protective role of this variant (Table 2 and Figure 1). Conversely, carriers of the RV haplotype are more frequently frail (p = 0.031 vs. other haplotypes), whereas the WT haplotype did not allow to distinguish between frail and not frail subjects (Table 2 and Figure 1).

Looking at the variants influence on lifespan, we analyzed the survival of RV and LAV carriers using a Cox regression. We could see a negative effect on survival by the RV haplotype (adjusted HR = 4.066; p = 0.044) but not by the LAV haplotype (adjusted HR = 0.002; p = 0.97) Figure 2. Likewise, the WT haplotype was uninfluential (data not shown).

Effect of gene therapy with AAV-LAV-BPIFB4 on frailty in aging mice

Systemic gene therapy with LAV-BPIFB4 resulted in a slight but significant delay in the progression of clinical frailty. In fact, mice injected with LAV-BPIFB4 displayed a significant lower frailty index at 7-month follow-up (2 months after the last injection of the gene) as compared with controls (Figure 3A). A subgroup analysis by age groups revealed that only old mice, but not adult mice, treated with LAV-BPIFB4 had a lower frailty index from 5- to 7-month follow-up compared with age-matched controls (Figure 3B). On the other hand, an analysis based on the prevalence of physical frailty could not capture a significant effect of LAV-BPIFB4 gene therapy in old mice at the 7-month assessment (20.0 vs. 44.4% in controls, p = 0.170 by Fisher’s exact test, Supplementary Table 1).

From the time of the first injection (3^rd month) up to the 12^th month assessment, we recorded 24 deaths in controls and 22 deaths in the LAV-BPIFB4-treated group within the old cohort, while the deaths in the adult cohort were 10 and 11, respectively. We found no significant difference in the mortality hazard between the treatment and control groups at the 12^th month, either considering the whole population (control vs. treatment group HR = 1.33, CI = 0.75-2.34; p = 0.33) or the old sub-population (control vs. treatment group HR = 1.28, CI = 0.68-2.41; p = 0.43).

We deemed that any further follow-up for survival after the 12^th month would have been unnecessary as the colony of old mice was numerically exhausted. We also argued that the excess death from the 7-month assessment onwards could had invalidated the power of the physical frailty analysis. Since part of these deaths in the old group may have arisen as an outcome of frailty, we also compared the combined prevalence of physical frailty and deaths in the treated and control mice. Accordingly, the prevalence of this combination was significantly lower in LAV-BPIFB4-treated old mice as compared with control mice (28.6 and 61.5%, respectively, p = 0.03 by Fisher’s exact test, Table 3). Interestingly, the proportion of old mice with reduced grip strength and gait disorders, which are parameters of the clinical frailty index related to the physical phenotype, were improved at the 7^th month by LAV-BPIFB4 (Supplementary Table 2).

Study limitations

Although genetic and molecular evidences support the role of BPIFB4 haplotypes in aging and longevity, additional studies should be carried out to confirm their role in the susceptibility to frailty. Due to the specificity of the studied cohort, replication in different and larger populations should be performed. Furthermore, an evaluation for a longer time is necessary to definitively determine the impact of the LAV haplotype on the risk of death. Likewise, larger cohorts of mice would be necessary to provide enough power in the assessment of survival.

Human sample description

The Calabria cohort involved in this study is a subset of a larger population already described by Montesanto et al [10]. This subset includes a total of 237 unrelated individuals (106 men and 131 women) 65–90 years old (median age 72 years), participated in the present study. All the subjects lived in Calabria (southern Italy) and their origin in the area have been verified up to the grandparent’s generation, as previously described [10]. Health status was ascertained by medical visit carried out by a geriatrician through a structured interview including physical and cognitive tests, as well as questions on common diseases occurred in the past. At the same time, it was performed DNA extraction and hematological analyses on peripheral venous blood samples.

For analyzing the correlation with quality of aging of the genetic variants investigated, we used the frailty classification of this sample, as obtained in a previous work [10]. In brief, according to this approach, each individual can be classified respect to his/her frailty level, determined by applying a hierarchical cluster analysis (HCA) on specific geriatric parameters, including Mini Mental State Examination (MMSE), Self-Reported Health Status (SHRS), Activity of Daily Living (ADL) and Hand Grip (HG) strength. For this population, two clusters were considered: non frail (the cluster with subjects showing the best scores for the classification variables) and frail (the clusters with subjects showing the worst scores for the classification variables). Furthermore, Calabria cohort has been followed-up for 10 years.

Ethics statement

Investigation has been conducted in accordance with the ethical standards and according to the Declaration of Helsinki and according to national and international guidelines and has been approved by the authors' institutional review board. Each subject, before the visit, signed an informed consent, for the permission to collect blood samples and usage of register-based information for research purposes

Genotyping

Samples were genotyped using Taqman assays for SNPs rs2070325 and rs11699009, to identify haplotypes. Alleles of rs2889732 and rs11696307 were imputed, given that are in total LD with the previous named respectively. It was performed data analysis with QuantStudio software 1.1 (ThermoFisher Scientific).

Gene therapy with AAV-LAV-BPIFB4 in mice

Constructs and vectors used in this study

We used LAV-BPIFB4- and green fluorescent protein–encoding adeno-associated viral vectors (AAV serotype 9 with a TBG promoter) to transduce mice. Details on the construction of these constructs have been previously described [9].

Animal study

All experiments were performed according to the European Community Council Directives of 2010/63/UE and the protocol was approved according to current Italian law (D.Lgs. n. 26/2014) by the General Direction of Animal Health and Veterinary Drugs of the Italian Ministry of Health with the authorization n° 130/2018-PR. We used C57BL/6J mice housed under specific pathogen-free (SPF) conditions in a room with controlled temperature (22 ± 2°C) and a 12-h light–dark cycle, with ad libitum access to food and water.

A total of 103 mice (71 males and 32 females) were used in the study. Three mice died before any treatment was performed and were excluded from the study. The mice were assigned to two age-matched experimental groups: a treatment group (AAV-LAV-BPIFB4; 50 mice) and a control group (AAV-GFP; 50 mice). The experimental groups were further subdivided into 4 subgroups based on the age of the mice at the start of the study. The first group consisted of “adult controls” (24 mice, 11 females and 13 males) aged 16-17 months (mean age ± SD = 16.8 ± 0.7 months). The second group consisted of “treated adults” (22 mice, 9 females and 13 males) aged 16-17 months (mean age ± SD = 16.7 ± 0.7 months). The third group consisted of “old controls” (26 mice, 4 females and 22 males) aged 18-23 months (mean age ± SD = 21.6 ± 1.9 months). The fourth group consisted of “treated old” (28 mice, 8 females and 20 males) aged 18-23 months (mean age ± SD = 21.0 ± 2.0 months). We performed non-invasive measurements of clinical frailty once a month in all mice. Physical frailty data were recorded at the 3^rd and at the 7^th month from the start of the study. After recording the frailty data at the 3^rd month, we injected (i.v.) into the tail vein 1*10¹⁴ viral particles of AAV-LAV-BPIFB4 or AAV-GFP in the treatment and control groups, respectively. The same treatment was repeated at the 5^th month from the start of the study. We also recorded the time to death for each mouse until the 12^th month since the beginning of the experiment. Mortality occurred when animals died suddenly or euthanized due to illness. A detailed design of the study is reported in Supplementary Figure 1.

Measurement of frailty in mice

We measured both clinical and physical frailty in mice. We measured the clinical frailty index (FI) in mice based on the validated murine clinical FI tool described previously [14]. Details on the measurement of weight, body surface temperature and grip strength (which are included in the FI tool) have been also described previously [12]. FI data were recorded the second week of each month from 10 am to 2 pm. All measurements of frailty were performed within the SPF animal facility of INRCA in a dedicated area. The clinical FI score for each mouse was calculated using the checklist published previously [14]. Clinical assessment included evaluation of the integument, musculoskeletal system, vestibulocochlear and auditory systems, ocular and nasal systems, digestive system, urogenital system, respiratory system, signs of discomfort, as well as the body weight and body surface temperature. For each parameter, a score of 0 was given if there was no sign of a deficit, a score of 0.5 denoted a mild deficit and a score of 1 indicated a severe deficit. Deficits in body weight and body surface temperature were scored based on their deviation from average reference values obtained from the entire cohort. Values that differed from reference values by less than 1 SD were scored as 0. Values that were ±1 SD with respect to the reference value were given a frailty value of 0.25; values that differed by ±2 SD scored 0.5, those that differed by ±3 SD scored 0.75 and values that were >3 SD above or below the mean received the maximal frailty value of 1. The sum retrieved from the values assigned to the 31 items on the checklist was then divided by 31 to yield a FI score between 0 and 1 for each animal.

The measurement of physical frailty in mice was performed following the same procedure described to translate the physical frailty screening performed in humans [1] to mice [11, 20, 31]. Performance testing was performed in both old and adult mice at the 3^rd month from the start of the study (before the treatment period) and at the 7^th month (after the treatment period). In order to ensure testing reliability, we adapted the mice to the tests for at least 2 months before the start of the study and performed multiple measurements for each criterion of the frailty assessment. The results from the multiple measurement were combined in a unique score for each criterion. The same testers performed all the measurements of frailty. All measurements performed to define the Physical Frailty phenotype are schematically described in Supplementary Table 3.

Our frailty phenotype included the following physical components:

1. Shrinking (weight loss). Shrinking was assessed by recording the current body weight and changes of body weight (these last measurements were obtained by comparing the current weight with the one measured in the previous 1 and 2 months).

2. Weakness. This criterion was assessed by measuring forelimb grip strength with 3 different tests: grip strength meter (Ugo Basile, Varese, Italy) measurement [32], dynamometer force measurement and increasing weights lift test [33].

3. Endurance. This criterion was measured by treadmill distance (program: starting at 5 rpm for 2 min and increasing speed from 5 to 50 m/s in 2700 s), mean time to fall at rotarod test (program: starting at 5 m/s for 2 min and increasing speed from 5 to 40 rpm in 300 s) and max weight reached at the increasing weights lift test. This test includes an endurance component due to the continuous increasing of the weight to be lifted by the mouse [33].

4. Slowness. We assessed this criterion by analyzing the distribution of the time spent by the mouse in different speed intervals in an Open Field test (whole test duration 5 min). The speed intervals considered where: I1 (0-1 cm/s), I2 (1-5 cm/s), I3 (5-10 cm/s), I4 (10-15 cm/s), I5 (15-20 cm/s), I6 (20-25 cm/s), I7 (25-30 cm/s), I8 (30-35 cm/s), I9 (35-40 cm/s), I10 (40-90 cm/s). We recorded the highest speed interval that the mouse run for at least 3 s and assigned as value of the test the mean speed of the interval (e.g. 12.5 for I4 and 37.5 cm/s for I9). The threshold of 3 s was established based on association with mortality data obtained from other cohort of mice (data not shown). Locomotor activity was conducted by a 5-min open field test on a white wood-chamber (72×72×30 cm) surmounted by a Xiaomi Yi Camera 16MP 1080 P 60FPS (YI Technology) controlled WI-Fi by a Smartphone. Videos were collected in a microSD disk and the tracking was performed offline with Biobserve Viewer3 (Biobserve GmbH, Germany) as previously described [12]. An additional measurement for slowness was obtained by recording the max speed recorded at rotarod test. Furthermore, we assessed slowness by also including the measurement of the mean stride length of the mice following a previously established protocol [34]. Indeed, there is a strong rationale in support of the relationship between walking speed and stride length, especially in older individuals [35].

5. Activity. Activity was recorded automatically by Biobserve Viewer3 (Biobserve GmbH, Germany) as the % the mice walked or run (speed above 0.45 cm/s) in a 5-min open field test. We additionally recorded the total distance run by the mouse in the same test.

All variables obtained by the measurements described above were standardized (transformed into Z-scores) and the variables assigned to the same criterion were averaged to create a composite Z-score. Following the percentiles used by Fried et al. in humans [1] and by others in mice [20], mice that fell in the bottom 20% of our old cohort for the composite Z-score computed for each criterion (Shrinking, Weakness, Endurance, Slowness and Activity), were considered positive for frailty for that given criterion. Mice with three or more positive frailty criteria were identified as frail.

Statistical analysis

Human study

For all the analyses, we re-codified the classification at BPIFB4 locus taking into account the haplotypic phase, as previously reported [23]. Since the LAV is represented by the minor allele of the two SNPs in haplotypic phase, the association with frailty trait was tested assuming a recessive model for this allele (i.e., LAV homozygotes – defined as rs2070325 = G and rs11699009 = T on both chromosomes – vs. LAV heterozygotes – defined as rs2070325 = G and rs11699009 = T on one chromosome – plus remaining haplotype carriers pooled). For the remaining two haplotypes with frequency >1% (i.e., RV and WT), dominant and recessive genetic models were assumed respectively (i.e., for RV, RV haplotype carriers – defined as rs2070325 = A and rs11699009 = T on at least one chromosome – vs. non-carriers; and for WT, WT homozygotes – defined as rs2070325 = A and rs11699009 = C on both chromosomes – vs. WT heterozygotes – defined as rs2070325 = A and rs11699009 = C on one chromosome – plus remaining haplotype alleles pooled, respectively).

Haplotype frequencies and phases were estimated from the observed genotypes. Fisher's Exact test were used for the comparison of frequencies of analyzed haplotypes. In order to evaluate if the detected effects of the analyzed polymorphisms on frailty status might finally result in differential patterns of survival of the different relevant genotypes, the survival after 11 years from the baseline visit was estimated. Kaplan–Meier survival curves were estimated for carriers vs non carriers of the relevant haplotype; in order to evaluate their predictive value with respect to mortality risk, the obtained survival curves were then compared by log-rank test. Hazard Ratios (HR) and 95% Confidence Intervals (95% CI) were estimated by using Cox proportional hazard models taking also into account possible confounder variables (age, gender, frailty status). Subjects alive after the follow-up time were considered as censored, and this time was used as the censoring date in the survival analyses. All the analyses were performed in R environment [36] and SPSS 25.0 (SPSS Inc., Chicago, IL). A significance level (α) of 0.05 was chosen in all the tests.

Animal study

Generalized linear mixed model analysis (SPSS 25.0) was used to take into account the longitudinal design of the study in mice. The identifier of each mouse, age group, gender, age of mouse at inclusion and time was indicated in the model. The linear model was developed assuming normal distribution with identity link function for data of the Clinical Frailty Index. The Satterthwaite approximation and robust estimator were used to take into account unbalanced data and violation of the assumptions. Fisher exact test was used to compare the prevalence of the physical frailty phenotype between control and treated animals. Differential patterns of survival due to the treatment were estimated by Cox-regression taking also into account possible confounder variables (age, age group and gender).