Retaining matrix dimensions when performing lookup via character subsetting

Question

I would like to use a named vector lookup to map the elements of a matrix to their integer index counterparts. I currently do the lookup via character subsetting but this drops the dimensions of the matrix. Specifying drop = FALSE does not help in this instance.

How can I return an elementwise-mapped matrix using a lookup whilst retaining the matrix dimensions, as per the desired output indicated?

cd3cd28 <- structure(c(26.4, 35.9, 59.4, 73, 33.7, 13.2, 16.5, 44.1, 82.8, 
                       19.8, 8.82, 12.3, 14.6, 23.1, 5.19, 18.3, 16.8, 10.2, 13.5, 9.73, 
                       58.8, 8.13, 13, 1.29, 24.8, 6.61, 18.6, 14.9, 5.83, 21.1, 17, 
                       32.5, 32.5, 11.8, 46.1, 414, 352, 403, 528, 305, 17, 3.37, 11.4, 
                       13.7, 4.66, 44.9, 16.5, 31.9, 28.6, 25.7, 40, 61.5, 19.5, 23.1, 
                       81.3), 
                     .Dim = c(5L, 11L), .Dimnames = list(NULL, c("praf", "pmek", "plcg", 
                                                                 "PIP2", "PIP3", "p44.42", 
                                                                 "pakts473", "PKA", "PKC", 
                                                                 "P38", "pjnk")))

protein_to_idx <- setNames(1:ncol(cd3cd28), colnames(cd3cd28))

proteins <- matrix(c("PKC", "PKA", 
                     "PKA", "P38", 
                     "PKA", "pjnk"), 
                   ncol = 2)

protein_to_idx[proteins]

# Current output
# PKC  PKA  PKA  P38  PKA pjnk 
#   9    8    8   10    8   11

# Desired output
matrix(protein_to_idx[proteins], ncol = 2)

#      [,1] [,2]
# [1,]    9   10
# [2,]    8    8
# [3,]    8   11

Note: Specifying drop = FALSE does not work here (single dimension indexing), i.e. the following

protein_to_idx[proteins, drop = FALSE]