If you're not stuck using R, you might consider using some command-line tools on your FASTA data (say, a file called tfbs.fa), to make a BED file that you can apply set operations on with bedops --range:

$ grep -e ">" tfbs.fa \
    | awk '{ print substr($0, 2); }' - \
    | awk -F ':|-' '{ print $1"\t"$2"\t"$3; }' \
    | sort-bed - \
    | bedops --everything --range 400 - \
    > query.bed

Briefly, this works by filtering the input FASTA file for lines that start with a > (headers that contain your coordinates of interest). This is passed to an awk statement that strips the > character and prints the rest. This result is passed to another awk statement that prints out the chromosome and start and stop positions to tab-delimited lines. This is a minimal BED file format, so we can sort it with sort-bed and symmetrically pad around the edges by 400 bases with bedops --range. Finally, this result is written to a file called query.bed.

Once you have a BED file (here called query.bed) you could convert that back to a FASTA file (e.g. called query.fa) using a tool like bed2fasta.py or with DAS queries to UCSC servers for your build of interest.

With your expanded FASTA file, you could use a tool like FIMO to search query.fa for motifs in the padded regions that match with your PWM.

You might wanna use UCSC's DAS server - that has an API to give you sequences given the co-ordinates.