How to automatically install Bioconductor package on which a CRAN package depends through a long chain of dependencies?

I have a package metagam which is on CRAN, and it passes all of CRAN's automated tests. However, the package depends on the Bioconductor package multtest through the chain of dependencies metagam <- metap <- mutoss <- multtest. This is visualized below for the dependencies of the metap package.

# Code for creating dependency graph
library(miniCRAN)
plot(makeDepGraph("metap", suggests = FALSE))

The answer to this question suggests adding biocViews: to DEPENDENCIES. Accordingly, the relevant parts of my DESCRIPTION look like this:

biocViews: 
Imports: 
    dplyr,
    furrr,
    ggplot2,
    knitr,
    metafor,
    metap,
    purrr,
    rlang,
    stringr,
    tidyr
RoxygenNote: 7.1.0
Suggests: 
    future,
    mgcv,
    gamm4,
    gratia,
    roxygen2,
    rmarkdown,
    devtools,
    covr,
    viridis,
    testthat (>= 2.1.0)

However, adding biocViews appears to only work for first-order dependencies. In my case, there are no Bioconductor packages listed in DESCRIPTION, and hence automatic package installation fails. This is shown by the example below.

# Remove packages 'metap', 'mutoss', and 'multtest' if they are installed
# If, any of these are installed, the dependency 'multtest' will not be attempted to be installed
pkgs <- installed.packages()[, "Package", drop = TRUE]
if("metap" %in% pkgs) remove.packages("metap")
#> Removing package from '/Library/Frameworks/R.framework/Versions/4.0/Resources/library'
#> (as 'lib' is unspecified)
if("mutoss" %in% pkgs) remove.packages("mutoss")
#> Removing package from '/Library/Frameworks/R.framework/Versions/4.0/Resources/library'
#> (as 'lib' is unspecified)
if("multtest" %in% pkgs) remove.packages("multtest")

# Install 'metagam', which trigges installation of the dependency 'metap'
# The dependencies are 'metagam' <- 'metap' <- 'mutoss' <- 'multtest'
install.packages("metagam")
#> Warning: dependency 'multtest' is not available
#> also installing the dependencies 'mutoss', 'metap'
#> 
#> The downloaded binary packages are in
#>  /var/folders/sz/q9lc1ggd66n5k5x_yp094hgh0000gn/T//Rtmphvg646/downloaded_packages

^{Created on 2020-06-20 by the reprex package (v0.3.0)}

Furthermore, this causes the package to fail, because the dependency on multtest is real:

# The code below is from the examples of the metagam::metagam() function
# The last line fails because the dependency 'multtest' is not available
library(metagam)
library(mgcv)
#> Loading required package: nlme
#> This is mgcv 1.8-31. For overview type 'help("mgcv-package")'.

## Create 5 datasets
set.seed(1234)
datasets <- lapply(1:5, function(x) gamSim(scale = 5, verbose = FALSE))

## Fit a GAM in each dataset, then use strip_rawdata() to remove
## individual participant data
models <- lapply(datasets, function(dat){
  ## This uses the gam() function from mgcv
  model <- gam(y ~ s(x0, bs = "cr") + s(x1, bs = "cr") + s(x2, bs = "cr"), data = dat)
  ## This uses strip_rawdata() from metagam
  strip_rawdata(model)
})

## Next, we meta-analyze the models.
## It is often most convenient to analyze a single term at a time. We focus on s(x1).
meta_analysis <- metagam(models, terms = "s(x1)", grid_size = 30)
#> Error in loadNamespace(j <- i[[1L]], c(lib.loc, .libPaths()), versionCheck = vI[[j]]): there is no package called 'multtest'

^{Created on 2020-06-20 by the reprex package (v0.3.0)}

I know I can inform the users, either on the GitHub repository or via a startup message, that they need to run the following lines before attempting to install metagam, but it does not provide the best user experience.

if (!requireNamespace("BiocManager", quietly = TRUE))
    install.packages("BiocManager") 
BiocManager::install("multtest")

What seems to me as the solution, is adding multtest to Imports. However, wouldn't this give a warning about unused imports, since no functions from multtest is directly used by metagam, but rather through the chain of dependencies specified above. I could not create such a warning on my own system, running R CMD check --as-cran, so maybe CRAN is fine with this? Alternatively, I could add a line of code somewhere using multtest, but this seems very hacky.

To summarize, my question is what I should do for the multtest package to be automatically installed when metagam is installed.

Because Bioconductor has twice-yearly releases, and this differs from CRAN release practices, the 'right' thing to do is to use Bioconductor tools to install your package, so BiocManager::install("metap"). It does not matter that metap is a CRAN package. BiocManager installs the correct version of the Bioconductor package for the user's version of R.

If this solution isn't palatable, then the next-best right thing to do is to adjust your dependencies to avoid direct or indirect dependencies on Bioconductor packages.